Magnetic Resonance Neuroimaging: Methods and Protocols (Methods in Molecular Biology 711)

ME T H O D S

IN

MO L E C U L A R BI O L O G Y

Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK

For other titles published in this series, go to www.springer.com/series/7651

TM

Magnetic Resonance Neuroimaging Methods and Protocols

Edited by

Michel Modo Department of Neuroscience, Institute of Psychiatry, King’s College London, UK

Jeff W.M. Bulte Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA

Editors Michel Modo Department of Neuroscience Institute of Psychiatry King’s College London London, SE5 9NU, UK [email protected]

Jeff W.M. Bulte Russell H. Morgan Department of Radiology and Radiological Science Division of MR Research Cellular Imaging Section Institute for Cell Engineering The Johns Hopkins University School of Medicine Baltimore, MD, 21205, USA [email protected]

ISSN 1064-3745 e-ISSN 1940-6029 ISBN 978-1-61737-991-8 e-ISBN 978-1-61737-992-5 DOI 10.1007/978-1-61737-992-5 Springer New York Dordrecht Heidelberg London © Springer Science+Business Media, LLC 2011 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of going to press, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Printed on acid-free paper Humana Press is part of Springer Science+Business Media (www.springer.com)

Preface To visualize the inside of a living human brain has been the goal of physicians since ancient times. The advent of noninvasive imaging technology, such as magnetic resonance imaging (MRI), during the latter half of the twentieth century has allowed for the opening of new vistas of the inner workings of the brain to biologists and clinicians on a daily basis. Great strides in unraveling the secrets of the brain have been achieved since the widespread implementation of imaging protocols in universities and hospitals. The gradual merging of molecular biology and imaging techniques at the beginning of the twenty-first century now affords a detailed investigation of the molecular underpinning of a working brain. The 30 chapters in this book contain experimental MRI protocols that can be used to noninvasively interrogate the healthy and diseased brain. The protocols are divided into general techniques (e.g., measuring relaxivity, magnetic resonance spectroscopy, diffusion tensor imaging, MR reporter genes) and specific applications in brain imaging (e.g., phenotyping transgenic animals, detecting amyloid plaques, fMRI in psychiatry). Most of these methods can be applied to both animal and human studies and may therefore provide a great resource for translational efforts. Clinical neurologists, psychiatrists, and radiologists will find these protocols useful, as will basic scientists working in the field of neuroscience. Michel Modo Jeff W.M. Bulte

v

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v

Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

SECTION I 1.

INTRODUCTION

From Molecules to Man: The Dawn of a Vitreous Man . . . . . . . . . . . . . . Michel Modo and Jeff W.M. Bulte

SECTION II

3

GENERAL TECHNIQUES

2.

Magnetic Resonance Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Andrew Simmons and Kristina Hakansson

17

3.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI . . . Nadim Joni Shah, Veronika Ermer, and Ana-Maria Oros-Peusquens

29

4.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain . . . Sean C.L. Deoni

65

5.

Magnetic Resonance Brain Image Processing and Arithmetic with FSL . . . . . . 109 William R. Crum

6.

Diffusion Tensor Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 Derek K. Jones and Alexander Leemans

7.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) . . . . . . . . . . 145 Cynthia A. Massaad and Robia G. Pautler

8.

Sodium MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Ronald Ouwerkerk

9.

MR Spectroscopy and Spectroscopic Imaging of the Brain . . . . . . . . . . . . . 203 He Zhu and Peter B. Barker

10. Amide Proton Transfer Imaging of the Human Brain . . . . . . . . . . . . . . . 227 Jinyuan Zhou 11. High-Field MRI of Brain Iron . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 Jozef H. Duyn 12. Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications . . . 251 Manisha Aggarwal, Jiangyang Zhang, and Susumu Mori 13. CEST MRI Reporter Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 Guanshu Liu, Jeff W.M. Bulte, and Assaf A. Gilad

vii

viii

Contents

14. Longitudinal Functional Magnetic Resonance Imaging in Animal Models . . . . . 281 Afonso C. Silva, Junjie V. Liu, Yoshiyuki Hirano, Renata F. Leoni, Hellmut Merkle, Julie B. Mackel, Xian Feng Zhang, George C. Nascimento, and Bojana Stefanovic 15. Combining EEG and fMRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Karen Mullinger and Richard Bowtell 16. MR Angiography and Arterial Spin Labelling . . . . . . . . . . . . . . . . . . . 327 David Thomas and Jack Wells SECTION III SPECIFIC APPLICATIONS 17. MRI Phenotyping of Genetically Altered Mice . . . . . . . . . . . . . . . . . . . 349 Jason P. Lerch, John G. Sled, and R. Mark Henkelman 18. Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications . . . . . . . 363 Philip K. Liu and Christina H. Liu 19. Molecular MRI Approaches to the Detection of CNS Inflammation . . . . . . . . 379 Nicola R. Sibson, Daniel C. Anthony, Sander van Kasteren, Alex Dickens, Francisco Perez-Balderas, Martina A. McAteer, Robin P. Choudhury, and Benjamin G. Davis 20. Brain Redox Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Ken-ichiro Matsumoto, Fuminori Hyodo, Kazunori Anzai, Hideo Utsumi, James B. Mitchell, and Murali C. Krishna 21. Systems Biology Approach to Imaging of Neural Stem Cells . . . . . . . . . . . . 421 Li Hua Ma, Yao Li, Petar M. Djuri´c, and Mirjana Maleti´c-Savati´c 22. MRI of Transplanted Neural Stem Cells . . . . . . . . . . . . . . . . . . . . . . 435 Stacey M. Cromer Berman, Piotr Walczak, and Jeff W.M. Bulte 23. MRI of Experimental Gliomas . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 Frits Thorsen 24. MRI in Experimental Stroke . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473 Timothy Q. Duong 25. Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of Parkinson’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487 Anthony C. Vernon and Michel Modo 26. Detecting Amyloid-β Plaques in Alzheimer’s Disease . . . . . . . . . . . . . . . . 511 Christof Baltes, Felicitas Princz-Kranz, Markus Rudin, and Thomas Mueggler 27. Assessing Subtle Structural Changes in Alzheimer’s Disease Patients . . . . . . . . 535 Jennifer L. Whitwell and Prashanthi Vemuri 28. Pharmacological Application of fMRI . . . . . . . . . . . . . . . . . . . . . . . 551 Mitul A. Mehta and Owen G. O’Daly

Contents

ix

29. MRI of Neuronal Plasticity in Rodent Models . . . . . . . . . . . . . . . . . . . 567 Galit Pelled 30. MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579 Yuexi Huang and Kullervo Hynynen Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595

Contributors MANISHA AGGARWAL • The Russell H. Morgan Department of Radiology and Radiological Science, Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA DANIEL C. ANTHONY • Department of Pharmacology, University of Oxford, Oxford, UK KAZUNORI ANZAI • Radiation Modifier Research Team, Heavy-Ion Radiobiology Research Group, National Institute of Radiological Sciences, Research Center for Charged Particle Therapy, Chiba, Japan CHRISTOF BALTES • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland PETER B. BARKER • The Kennedy Krieger Institute, Baltimore, MD, USA; Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA RICHARD BOWTELL • School of Physics and Astronomy, Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, UK JEFF W.M. BULTE • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA ROBIN P. CHOUDHURY • Department of Cardiovascular Medicine, University of Oxford, Oxford, UK STACEY M. CROMER BERMAN • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA WILLIAM R. CRUM • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK BENJAMIN G. DAVIS • Department of Chemistry, University of Oxford, Oxford, UK SEAN C.L. DEONI • Division of Engineering, Brown University, Providence, RI, USA ALEX DICKENS • Department of Pharmacology, Department of Chemistry, CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK PETAR M. DJURI C´ • Department of Computer and Electrical Engineering, Stony Brook University, Stony Brook, NY, USA TIMOTHY Q. DUONG • Research Imaging Institute, University of Texas Health Science Center at San Antonio, San Antonio, TX, USA JOZEF H. DUYN • Advanced MRI Section, National Institute for Neurological Disease and Stroke, National Institute of Health, Bethesda, MD, USA VERONIKA ERMER • Institute of Neuroscience, Medicine (INM-4), Research Centre Juelich, Juelich, Germany ASSAF A. GILAD • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA

xi

xii

Contributors

KRISTINA HAKANSSON • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK R. MARK HENKELMAN • The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada YOSHIYUKI HIRANO • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA YUEXI HUANG • Sunnybrook Health Sciences Centre, Toronto, ON, Canada KULLERVO HYNYNEN • Sunnybrook Health Sciences Centre, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada FUMINORI HYODO • Innovation Center for Medical Redox Navigation, Kyushu University, Fukuoka, Japan DEREK K. JONES • School of Psychology, Cardiff University Brain Research Imaging Centre (CUBRIC), Cardiff University, Cardiff, UK MURALI C. KRISHNA • Radiation Biology Branch, National Cancer Institute, National Institutes of Health, Center for Cancer Research, Bethesda, MD, USA ALEXANDER LEEMANS • Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands RENATA F. LEONI • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JASON P. LERCH • The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada YAO LI • Department of Computer and Electrical Engineering, Stony Brook University, Stony Brook, NY, USA JUNJIE V. LIU • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA PHILIP K. LIU • Department of Radiology, AA Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, USA CHRISTINA H. LIU • Department of Radiology, AA Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, USA GUANSHU LIU • Kennedy Krieger Institute, F.M. Kirby Research Center for Functional Brain Imaging, The Johns Hopkins University School of Medicine, Baltimore, MD, USA LI HUA MA • Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA JULIE B. MACKEL • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA MIRJANA MALETI C´ -SAVATI C´ • Department of Pediatrics, Baylor College of Medicine, Houston, TX, USA CYNTHIA A. MASSAAD • Department of Molecular Physiology and Biophysics, Baylor College of Medicine, Houston, TX, USA KEN-ICHIRO MATSUMOTO • Radiation Modifier Research Team, Heavy-Ion Radiobiology Research Group, National Institute of Radiological Sciences, Research Center for Charged Particle Therapy, Chiba, Japan

Contributors

xiii

MARTINA A. MCATEER • Department of Cardiovascular Medicine, University of Oxford, Oxford, UK MITUL A. MEHTA • Institute of Psychiatry at King’s College London, Centre for Neuroimaging Sciences (PO89), London, SE5 8AF, UK; Medical Research Council Clinical Sciences Centre, Imperial College London, Hammersmith Hospital, London, UK HELLMUT MERKLE • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JAMES B. MITCHELL • Radiation Biology Branch, National Cancer Institute, National Institutes of Health, Center for Cancer Research, Bethesda, MD, USA MICHEL MODO • Department of Neuroscience, Kings College London, Institute of Psychiatry, Centre for the Cellular Basis of Behaviour, The James Black Centre, London, UK SUSUMU MORI • The Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA THOMAS MUEGGLER • Pharmaceuticals Division, F. Hoffmann-La Roche Ltd, Basel, Switzerland KAREN MULLINGER • School of Physics and Astronomy, Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, UK GEORGE C. NASCIMENTO • Department of Biomedical Engineering, Federal University of Rio Grande do Norte, Campus Universitario, Natal, RN, Brazil OWEN G. O’DALY • Institute of Psychiatry at King’s College London, Centre for Neuroimaging Sciences (PO89), London, UK ANA-MARIA OROS-PEUSQUENS • Institute of Neuroscience, Medicine (INM-4), Research Centre Juelich, Juelich, German; King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK RONALD OUWERKERK • Cardiovascular Imaging, National Institute of Diabetes and Digestive and Kidney Disease, National Institute of Health, Bethesda, MD, USA ROBIA G. PAUTLER • Department of Molecular Physiology and Biophysics, Department of Neuroscience, Department of Radiology, Baylor College of Medicine, Houston, TX, USA GALIT PELLED • Department of Radiology, Kennedy Krieger Institute, Johns Hopkins School of Medicine, Baltimore, MD, USA FRANCISCO PEREZ-BALDERAS • CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK FELICITAS PRINCZ-KRANZ • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland MARKUS RUDIN • Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland; Institute of Pharmacology and Toxicology, University of Zurich, Zurich, Switzerland NADIM JONI SHAH • Institute of Neuroscience and Medicine (INM-4), Research Centre Juelich, Juelich, Germany; RWTH Aachen University, Aachen, Germany NICOLA R. SIBSON • CR-UK/MRC Gray Institute for Radiation Oncology and Biology, University of Oxford, Oxford, UK AFONSO C. SILVA • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA ANDREW SIMMONS • King’s College London, Institute of Psychiatry, Centre for Neuroimaging Sciences, London, UK

xiv

Contributors

JOHN G. SLED • Department of Medical Biophysics, The Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada; University of Toronto, Toronto, ON, Canada BOJANA STEFANOVIC • Imaging Research, Sunnybrook Health Science Centre, Toronto, ON, Canada DAVID THOMAS • Department of Medical Physics and Bioengineering, University College London, London, UK FRITS THORSEN • Department of Biomedicine, University of Bergen, Bergen, Norway HIDEO UTSUMI • Innovation Center for Medical Redox Navigation, Kyushu University, Fukuoka, Japan SANDER VAN KASTEREN • Department of Chemistry, University of Oxford, Oxford, UK PRASHANTHI VEMURI • Department of Radiology, Mayo Clinic, Rochester, MN, USA ANTHONY C. VERNON • Department of Neuroscience, Institute of Psychiatry, Centre for the cellular basis of behaviour, The James Black Centre, Kings College London, London, UK PIOTR WALCZAK • Russell H. Morgan Department of Radiology and Radiological Science, Division of MR Research, and Cellular Imaging Section, Institute for Cell Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD, USA JACK WELLS • Centre for Advanced Biomedical Imaging, University College London, London, UK JENNIFER L. WHITWELL • Department of Radiology, Mayo Clinic, Rochester, MN, USA XIAN FENG ZHANG • Cerebral Microcirculation Unit, Laboratory of Functional and Molecular Imaging, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD, USA JIANGYANG ZHANG • The Russell H. Morgan Department of Radiology and Radiological Science, F.M. Kirby Research Center for Functional Brain Imaging, Johns Hopkins University School of Medicine, Baltimore, MD, USA JINYUAN ZHOU • Department of Radiology, Johns Hopkins University, Baltimore, MD, USA HE ZHU • Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA

Section I Introduction

Chapter 1 From Molecules to Man: The Dawn of a Vitreous Man Michel Modo and Jeff W.M. Bulte Abstract One of the greatest challenges to study the structure, function, and molecules in the living brain is that it is enclosed within the skull and difficult to access. Although biopsies are feasible, they are invasive, could lead to functional impairments, and in any case will only provide a small regional sample that is not necessarily reflecting the entire brain. Since the beginning of the twentieth century, in vivo imaging has gradually, and steadily, matured into non-invasive techniques that enable the repeated investigation of the structural, functional, cellular, and molecular composition of the brain. Not only is this information of great importance to scientists aiming to understand how the brain works, but these techniques are also essential to physicians who use imaging to diagnose and treat disease. The current book is a collection of 29 cutting-edge methods and protocols that are used in the current field of neuroimaging. Key words: Neuroimaging, magnetic resonance imaging, computer tomography, positron emission tomography.

1. Looking Inside the Living Brain: A Historical Primer

Correct functioning of the brain is central to our everyday lives. Developmental problems or damage to the brain can interfere with someone’s ability to take care of basic live functions. However, the study of the brain is hampered by it being enclosed by the skull that prevents us from seeing or studying the brain directly. Initial theories and studies of the brain were based on ill-conceived ideas, but these have gradually paved the way for a thorough scientific study of the brain and mind. Since the beginning of modern medicine, these studies have been highly dependent on technological developments.

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_1, © Springer Science+Business Media, LLC 2011

3

4

Modo and Bulte

Arguably, the eighteenth-century pseudo-scientific ideas of Franz Gall’s cranioscopy, i.e., the appearance (scopos in Greek) of the skull (cranium), are at the origin of modern brain science. Gall professed that the external appearance of the skull mirrors the enclosed brain (1). Importantly, phrenology (phrenos for mind and logos for study) further developed this concept and implied that these external anatomical characteristics are indicative of particular behaviors or personality traits (2). Although even during its day doubts about this approach arose and were considered spurious by some (3), phrenology retained a large following and in some cases was used to justify prejudice and political agendas (4). However, in the second half of the nineteenth century, linking damage of particular brain areas with functional impairments surpassed the scientifically unfounded rhetoric of phrenology. Notably, the seminal studies of Phineas Gage by John Harlow (5) and Tan’s aphasia by Paul Broca (6) ushered in a new era of science that attempted to directly link an anatomical brain region with its contribution to behavior. Ever since these seminal studies, postmortem neuropathology has been the foundation of brain science against which in vivo imaging has been compared. The cellular and molecular compositions of the brain are the gold standard of evidence that damage or aberrations occurred in that region. Although postmortem histopathology can be very informative about the locale of damage or abnormalities, it only provides a means to study conditions after someone has passed away. It is hence of no diagnostic value. In contrast, being able to visualize aspects of pathology in vivo not only allows a more rigorous study of the temporal and spatial progression of these pathologies but potentially also provides a means to diagnose particular conditions and establish a differential diagnosis for an appropriate treatment. At present, both specialties of psychiatry and neurology depend on the study of the brain in vivo to elucidate the underlying causes of behavioral dysfunction. The first step in achieving this in vivo visualization of the brain has been taken at the beginning of the twentieth century with elementary X-ray-based imaging techniques, such as pneumoencephalography (PEG) (7, 8). Table 1.1 provides a time line of milestones in neuroimaging. Although these early techniques provided insights into the living brain, they were also often causing damage to the patient’s brain (e.g., injection of air into the lateral ventricles to provide contrast). Gradually technological developments, such as tomography, where X-rays are rotated around the patient to record (graphens) the signal on single sections (tomos) (9, 10), heralded new innovations that are still in use today. Already in the 1920s, Edgar Moniz used imaging to noninvasively visualize blood vessels in the brain to identify the location of brain tumors (11). Despite these early pioneering advances,

From Molecules to Man: The Dawn of a Vitreous Man

5

Table 1.1 Time line of technological and methodological milestones in (neuro) imaging. MRI milestones are in bold Year

Researcher

Milestone

1895

Roentgen (43)

X-ray image of skull

1910

Bachem and Gunther (44)

First use of contrast media

1916

Dandy (7)

Pneumoencephalography

1924

Hevesy (45)

Radiotracer use in animals

1927

Moniz (11)

Angiography

1931

Vallebona (22)

Stratigraphic imaging

1935

Grossman (9, 10)

Tomographic imaging

1936

Gorter (46)

Paramagnetic relaxation

1938

Rabi (47)

Nuclear magnetic resonance

1942

Bloch (48) and Purcell (49)

Measured NMR signal

1953

Brownell and Sweet (50)

Positron imaging in brain tumors

1956

Kuhl (51)

Recorder for radionuclide scanning

1958

Anger (52)

Scintillation camera

1962

Rankowitz and Robertson (53)

PET transverse section instrument

1963

Kuhl (54)

Emission reconstruction tomography

1965

Harper and Lathrup (55)

Tc-99m radiotracer for brain

1971

Damadian (56)

Hydrogen density in tumors measured by NMR

1973

Hounsfield (13) and Cormack (12)

Computer tomography (CT)

1973

Mansfield (57) and Lauterbur (58)

Magnetic resonance imaging (MRI)

1974

Budinger and Gullberg (59)

SPECT

1974

Hoult (60)

Magnetic (MRS)

1975

Ter-Pogossian (61) and Phelps (62)

Positron emission tomography (PET)

1975

Kuhl (63)

First quantitative cerebral blood volume measurement

1975

Ernst (64)

Phase encoding for MRI

1977

Jaszcak (65)

First head SPECT

1977

Ido and Alavi (67)

FDG-PET

1977

Damadian (66)

First MRI scan of patient

1977

Mansfield (68)

Echo planar imaging (EPI)

1980

Redpath (69)

Spin-warp technique for MRI

1981

Bydder (70)

MR contrast agent

resonance

spectroscopy

1983

Wagner (14)

First neuroreceptor imaging using PET

1984

Weinmann (71)

Gd-DTPA

1986

Nishimura (72)

MR angiography

6

Modo and Bulte

Table 1.1 (continued) Year

Researcher

Milestone

1987

Kornguth (30)

T-cell tracking by MRI

1989

Friston (15)

Statistical parametric mapping

1989

Koretsky (36)

Creatine kinase reporter gene for 1 P NMR

1990

Kalender (73)

Spiral CT

1990

Wolff and Balaban (32)

Magnetization transfer

1990

Weissleder (74)

USPIO

1992

Ogawa (17)

fMRI

1994

LeBihan (23)

Diffusion tensor imaging

1995

Tjujavev (75)

Thymidine kinase reporter gene for PET

1995

Wright (76)

Voxel-based morphometry

1997

Cherry (42)

PET/MRI

1998

Kinahan (77)

PET/CT

1999

Pruessman (78)

Parallel imaging

2000

Ward and Balaban (79)

CEST

2000

Louie (31)

Gene expression by MRI

2003

Dronkers (26)

Voxel lesion-symptom mapping

2005

Gleich and Weizenecker (40)

Magnetic particle imaging (MPI)

it was only in the late 1970s that the advent of computers heralded new technological developments in noninvasive brain imaging that resulted in the spread of scanners beyond a few dedicated academic research centers. The advent of X-ray-based computer tomography (CT) (12, 13) undoubtedly provided the single most important step forward in neuroimaging. CT rapidly became a core assessment tool in neurology during the 1980s and has remained the most commonly used imaging modality in a day-to-day clinical setting. As the importance of clinical neuroimaging grew, the importance of technological developments and its impact on clinical practice and science increased as well. Developments and implementations of novel techniques, such as positron emission tomography (PET) and magnetic resonance imaging (MRI), further increased the diagnostic accuracy. Especially the sensitivity of PET for specific receptors opened up the opportunity to determine particular neurochemical deficits in vivo (14). Sophisticated computational models were required though to generate accurate maps of neuroreceptor binding of radioligands (15). The models could also be used to investigate regional brain activity as

From Molecules to Man: The Dawn of a Vitreous Man

7

18-fluorodeoxyglucose (FDG), a glucose analogue used in PET imaging, is only taken up by metabolically active cells. Functional activity was hence measurable and localizable using FDG-PET. Although this was an important step forward for scientists interested in regional activation during particular behaviors, the low temporal resolution of the technique makes it more suitable as a diagnostic tool for brain tumors. As of today, cost, subject throughput, and the required resources (such as a cycloctron, which is needed to generate 18-FDG) still limit the widespread use of PET imaging. As PET does not provide detailed anatomical information about the exact substructural brain area of interest, additional scans using CT or MRI are often necessary.

2. The Anatomy of Behavior: The Birth of Functional Neuropathology

Developments in nuclear magnetic resonance (NMR) during the 1980s rapidly developed diverse applications ranging from structural and spectroscopic imaging to angiography and mapping of cerebral blood flow. Although CT and PET were initially most commonly employed, since the start of the 1990s MRI has increasingly become the modality of choice to study the brain (Fig. 1.1a). Most importantly in this predominance of MRI is the use of “functional” MRI (Fig. 1.1b). In 1990, Seji Ogawa (16) described the blood oxygen level-dependent (BOLD) contrast that takes advantage of the paramagnetic relaxivity of deoxygenated blood. As areas in the brain consume oxygen during increased activity, the regional concentration of deoxygenated blood increases proportionally, with a concomitant decrease of MRI signal. This basic principle has revolutionized the study of the relationship between brain activity and common behaviors by using functional MRI (fMRI) (17). Although the specific neural correlates of this effect remain under investigation and a matter of a scientific debate (18), the underlying assumption that it is possible to localize some aspects of brain activity using BOLD has mostly remained unchallenged. Concerns have, nevertheless, been raised that regions of activity are often referred to as, for instance, “the region” of happiness (19). This has led some to argue that care must be taken not to commit the same misleading pseudo-scientific interpretations of localized brain activity than those associated with phrenology (20–22). It is important to acknowledge that brain areas are interconnected and multiple regions may be involved in one particular behavior. Regions of activity on fMRI scans merely indicate what area is mostly correlate during a specific paradigm.

8

Modo and Bulte

Fig. 1.1. A chronological comparison of neuroimaging publications. a The three main imaging modalities revolutionized brain imaging during the 1980s. However, only a limited number of publications exist for each year (~20). At the start of the 1990s, during the NIH-proclaimed “decade of the brain,” the advent of neuroimaging exhibited a dramatic tenfold increase. Although the use of PET and CT remained fairly consistent from thereon, there is a linear increase in papers using MRI to study the brain. b By comparing the different MRI techniques, it becomes obvious that the increase in MRI since the mid-1990s is predominantly due to functional MRI (fMRI). In 2009, fMRI accounted for over 60% of all papers using MRI to study the brain. Other techniques, such as structural MRI (sMRI), diffusion tensor imaging (DTI), cellular/molecular MRI, and magnetic resonance spectroscopy (MRS), also steadily increased, albeit at a slower rate than fMRI. In contrast, reports using magnetic resonance angiography (MRA) have not increased over almost 20 years. The diversity of techniques and ease of use of MRI compared to CT and PET increasingly advocate MRI as the modality of choice for scientific and clinical studies.

From Molecules to Man: The Dawn of a Vitreous Man

9

To uncover multiple regions of activity involved in a given behavior (e.g., speech), a systematic modification of task paradigms and their effect on brain activity need to be conducted. Apart of brain activity that is sited in the gray matter of the brain, it is also important to uncover the connections, i.e., white matter fiber tracts, between these sites of activity. Taking advantage of the anisotropic movement of water in white matter, diffusion tensor images can be generated to indicate the presence and direction of fiber tracts (23). This brain hodology (i.e., the study of pathways) allows scientists and clinicians now to go beyond finding one particular area that is involved in a given behavior but afford the construction of brain networks that are interconnected in solving a particular behavioral task. For instance, it is now possible to revisit the early conclusions by Paul Broca to uncover a network of brain areas that are involved in the understanding and generation of speech, with damage to each subcomponent leading to a specific neuropsychological deficit (24). Advances in MRI are no longer just dependent on technological advances in scanner hardware or acquisition, but also increasingly depend on sophisticated image analysis. Early studies merely inspected images to find particular hallmarks that were apparent (e.g., hyperintensity on T2-weighted images of patients with stroke), but ever more subtle differences were revealed by steadily refining these methods to measure regions of interests (ROIs) and to statistically compare individual voxels. For instance, functional MRI often detects signal changes by 2–5% difference using dedicated statistical image analysis. Similar subtle differences can also be found in structural images that are taken over a long time span. For instance, in patients with Alzheimer’s disease, statistical comparisons of serial images (i.e., voxel-based morphometry) can reveal which brain regions shrink or enlarge (25). The degree of change in tissue can be compared using deformation-based morphometry (DBM) that can be further integrated with a subject’s behavioral performance in so-called voxel lesion-symptom mapping (VLSM) (26) analyses to indicate which structural changes are actually associated with a given change in behavior.

3. Molecular and Cellular Imaging: Back to the Future?

Although conventional human brain imaging studies can unravel the regional connectivity and functional contributions to behavior, such investigations do not uncover the molecular and cellular underpinnings of these. Although magnetic resonance spectroscopy (MRS) is frequently used for studying brain tumors, its requirement for acquiring signal from large voxels of interest

10

Modo and Bulte

has limited its application to link metabolic changes to behavior. Nevertheless, in some instances, MRS can provide longitudinal data on the molecular/cellular composition of a region of interest. For instance, neuronal loss or cell death may be assessed using MRS (27). Some groups have even suggested that it can be applied to assess neurogenesis due to specific lipid profiles of neural stem cells (28), although this has been questioned (29). To reduce sampling bias, chemical shift imaging (CSI) is anticipated to provide detailed spatial information about metabolite changes in the living brain. However, MRS and CSI can only detect a limited range of molecules, and consequently additional approaches are required to visualize the true spatial distribution of particular molecules and cells. Analogous to PET and SPECT imaging, for MRI this may be achieved by applying contrast agents that selectively recognize specific (targeted) molecules. To achieve this, contrast agents act by changing the relaxation of nearby protons. One way of targeting an MR contrast agent to depict a molecule of interest is conjugating it to a monoclonal antibody (30). “Smart” contrast agents can be engineered to only change relaxivity in the presence of a particular molecule (31). By applying specific radiofrequency pulses, it is also possible to employ a variety of nonmetallic contrast agents using their chemical exchange saturation transfer (CEST) properties (32). This approach potentially allows “multicolor” imaging (33) that could visualize the presence of more than one targeted molecule within the same region, akin to multiple fluorescent dyes in microscopy. It is hoped that more than one type of cell could be tracked simultaneously in vivo. Tracking of cells using MRI has been of growing interest to determine the infiltration of immune cells (30), the cellular contribution to organogenesis (34), and more recently the localization of cell transplants (35). One drawback of this approach is that contrast agents also produce contrast if cells die. Developments of MRI reporters that only produce contrast if cells are alive can overcome this problem (36), but effects on cellular functions need to be further investigated. One issue with modifying proton contrast is that many pathological measurements depend on this contrast and hence cells that modify this contrast can interfere with the detection of pathology. MRI can also detect other nuclei, for instance, 19 F, and 19 F-MRI has been used to detect cells (37) and the presence of fluorinated compounds (38). However, sensitivity remains a major issue with cellular and molecular MRI, given that only a few molecules or cells of interest may be located within the brain. To overcome this, micron-sized particles of iron oxide (MPIOs) (39) or magnetic particle imaging (MPI) (40) might be alternative approaches to explore. To overcome practical limitations of the different imaging modalities, multimodal imaging is gradually working its way into

From Molecules to Man: The Dawn of a Vitreous Man

11

clinical implementation. For instance, in cancer PET/CT imaging, FDG-PET images of tumor metabolism are superimposed on high-resolution anatomical CT images in order to improve tumor resection (41). As patients only require a single scanning session, information flow can be optimized and provide a substantial benefit as compared to using two independent machines. Similar benefits can be expected from PET/MRI (42), especially in areas such as acute stroke. For instance, PET can provide information regarding penumbral tissue metabolism, whereas anatomical MRI and MR angiography can be used to visualize tissue and blood vessel injury. As pharmacological therapies to limit stroke are expected to have a short time window, obtaining this information quickly within a single scanning session will improve the delivery of acute treatments that can have a maximum impact on the degree of damage caused to the brain. Hence, a wide variety of brain abnormalities can now be studied within a single scanning session. Eventually, this will generate a comprehensive picture of the brain that will increasingly form the basis for a differential diagnosis determining the most appropriate action in treating brain disease.

4. Conclusion: The Dawn of a Vitreous Man

With its humble beginnings in “pseudoscience,” the study of the living brain – which controls our behavior – has rapidly developed into a systematic experimental science, where novel frontiers are mostly driven by technological and methodological innovations. Novel and improved technology protocols will be the fundamental catalyst in the interdisciplinary field of brain imaging. To support the implementation and innovation of these approaches, this book provides a collection of methods and protocols that are state of the art in MR neuroimaging. The last 20 years in MR neuroimaging have already allowed us to visualize in ever-greater detail the structure, function, connectivity, and molecular composition of the brain. Further integration of these methods will eventually allow us to unravel ever-greater mysteries of the brain. Indeed, the dawn of a vitreous man is likely to be upon us soon.

References 1. Gall, F. J. Cranologie, ou decouvertes nouvelles concernant le cerveau, le crâne et les organes. Paris: F: H. Nicolle; 1807. 2. Spurzheim, J. G. Manuel de phrenologie. Paris: F: Porthann; 1832.

3. Magendie, F. An Elementary Treatise on Human Physiology. New York, NY: Harper; 1843. 4. Gould, S. J. The Mismeasure of Man. London: Penguin Books; 1981.

12

Modo and Bulte

5. Harlow, J. M. Passage of an iron rod through the head. Boston Med Surg J 1848;39: 389–393. 6. Broca, P. Perte de la parole, ramolissement chroniqe et destruction partielle du lobe anterieure du cerveau. Bulletin De La Societe D’anthropologie 1861;2:235–238. 7. Dandy, W. E. Ventriculography following injection of air into the cerebral ventricles. Ann Surg 1918;68:5–11. 8. Dandy, W. E. Rontgenography of the brain after injection of air into the spinal canal. Ann Surg 1919;70:397–403. 9. Grossman, G. Tomography I. RöFO – Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren. 1935;51:61–80. 10. Grossman, G. Tomography II. RöFO – Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren. 1935;51:191–209. 11. Moniz, E. L’encephalographie arterielle, son importance dans la localization des tumeurs cerebrales. Rev Neurol 1927;34:72–90. 12. Cormack, A. M. Reconstruction of densities from their projections with applications in radiological physics. Phys Med Biol 1973;18:195–207. 13. Ambrose, J., Hounsfield, G. Computerized transverse axial scanning (tomography). Br J Radiogr 1973;46:1016–1022. 14. Wagner, H. N., Jr., Burns, H. D., Dannals, R. F. et al. Imaging dopamine receptors in the human brain by positron tomography. Science 1983;221:1264–1266. 15. Friston, K. J., Passingham, R. E., Nutt, J. G., Heather, J. D., Sawle, G. V., Frackowiak, R. S. Localisation in PET images: Direct fitting of the intercommissural (ACPC) line. J Cereb Blood Flow Metab 1989;9: 690–695. 16. Ogawa, S., Lee, T. M., Kay, A. R., Tank, D. W. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990;7: 9868–9872. 17. Ogawa, S., Tank, D. W., Menon, R. et al. Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA 1992;89:5951–5955. 18. Mangia, S., Giove, F., Tkac, I. et al. Metabolic and hemodynamic events after changes in neuronal activity: Current hypotheses, theoretical predictions and in vivo NMR experimental findings. J Cereb Blood Flow Metab 2009;29:441–463. 19. Breiter, H. C., Etcoff, N. L., Whalen, P. J. et al. Response and habituation of the human

20.

21. 22. 23. 24. 25.

26. 27.

28.

29.

30.

31.

32. 33.

amygdala during visual processing of facial expression. Neuron 1996;17:875–887. Vul, E., Harris, C., Winkielman, P., Pashler, H. Puzzling high correlations in fMRI studies of emotion, personality, and social cognition. Perspect Psychol Sci 2009;4:274–290. Kennedy, D. Neuroimaging: Revolutionary research tool or a post-modern phrenology? Am J Bioeth 2005;5:19, discussion W3–4. Raichle, M. E. Modern phrenology: Maps of human cortical function. Ann N Y Acad Sci 1999;882:107–118, discussion 28–34. Basser, P. J., Mattiello, J., LeBihan, D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259–267. Catani, M., ffytche, D. H. The rises and falls of disconnection syndromes. Brain 2005;128:2224–2239. Good, C. D., Scahill, R. I., Fox, N. C. et al. Automatic differentiation of anatomical patterns in the human brain: Validation with studies of degenerative dementias. Neuroimage 2002;17:29–46. Bates, E., Wilson, S. M., Saygin, A. P. et al. Voxel-based lesion-symptom mapping. Nat Neurosci 2003;6:448–450. Griffith, H. R., Stewart, C. C., den Hollander, J. A. Proton magnetic resonance spectroscopy in dementias and mild cognitive impairment. Int Rev Neurobiol 2009;84:105–131. Manganas, L. N., Zhang, X., Li, Y. et al. Magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain. Science 2007;318:980–985. Jansen, J. F., Gearhart, J. D., Bulte, J. W. Comment on “magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain”. Science 2008;321:640. Kornguth, S. E., Turski, P. A., Perman, W. H. et al. Magnetic resonance imaging of gadolinium-labeled monoclonal antibody polymers directed at human T lymphocytes implanted in canine brain. J Neurosurg 1987;66:898–906. Louie, A. Y., Huber, M. M., Ahrens, E. T. et al. In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 2000;18:321–325. Wolff, S. D., Balaban, R. S. Nmr imaging of labile proton-exchange. J Magn Reson 1990;86:164–169. McMahon, M. T., Gilad, A. A., DeLiso, M. A., Berman, S. M., Bulte, J. W., van Zijl, P. C. New “multicolor” polypeptide diamagnetic chemical exchange saturation transfer (DIACEST) contrast agents for MRI. Magn Reson Med 2008;60:803–812.

From Molecules to Man: The Dawn of a Vitreous Man 34. Jacobs, R. E., Fraser, S. E. Magnetic resonance microscopy of embryonic cell lineages and movements. Science 1994;263:681–684. 35. Bulte, J. W., Zhang, S., van Gelderen, P. et al. Neurotransplantation of magnetically labeled oligodendrocyte progenitors: Magnetic resonance tracking of cell migration and myelination. Proc Natl Acad Sci USA 1999;96:15256–15261. 36. Koretsky, A. P., Traxler, B. A. The B isozyme of creatine kinase is active as a fusion protein in escherichia coli: In vivo detection by 31p NMR. FEBS Lett 1989;243:8–12. 37. Ratner, A. V., Hurd, R., Muller, H. H. et al. 19f magnetic resonance imaging of the reticuloendothelial system. Magn Reson Med 1987;5:548–554. 38. Burt, C. T., Moore, R. R., Roberts, M. F., Brady, T. J. The fluorinated anesthetic halothane as a potential NMR biologic probe. Biochim Biophys Acta 1984;805:375–381. 39. Hinds, K. A., Hill, J. M., Shapiro, E. M. et al. Highly efficient endosomal labeling of progenitor and stem cells with large magnetic particles allows magnetic resonance imaging of single cells. Blood 2003;102:867–872. 40. Gleich, B., Weizenecker, J. Tomographic imaging using the nonlinear response of magnetic particles. Nature 2005;435:1214–1217. 41. Yang, S., Zhang, C., Zhu, T. et al. Resection of gliomas using positron emission tomography/computed tomography neuronavigation. Neurol Med Chir (Tokyo) 2007;47:397–401, discussion 2. 42. Garlick, P. B., Marsden, P. K., Cave, A. C. et al. PET and NMR dual acquisition (PANDA): Applications to isolated, perfused rat hearts. NMR Biomed 1997;10:138–142. 43. Rontgen, W. C. 1895. Eine neure Art von Strahlen. Sitzungsberichte der Physikalishmedizinischen Gesellschaft Zu Wurzburg 1895. 44. Bachem, C., Gunther, H. Z. Bariumsulfat also schattenbioldendes kontrastmittel bei rontgenuntersuchungen. Zeitschrift Fur Rontgenkunde Und Radiumforschung 1910;12:369–376. 45. Christiansen, I. A., Hevesy, G., Lomholt, S. Chimie physiologique. Recherches, par une methode radiochimique, sur la circulation du bismuth dans l’organism. Compte Rendu De L’academie Des Sciences 1924;178:1324–1326. 46. Gorter, C. J. Paramagnetic relaxation. Physica 1936;3:503–514. 47. Rabi, I. I., Zacharias, J. R., Millman, S., Kusch, P. A new method of measur-

48. 49. 50. 51.

52. 53.

54. 55.

56. 57.

58.

59.

60.

61.

62.

63.

13

ing nuclear magnetic moment. Phys Rev 1938;53:526–535. Bloch, F., Rabi, I. I. Atoms in variable magnetic fields. Rev Mod Phys 1945;17:237–244. Purcell, E. M., Torrey, H. C., Pound, R. V. Resonance absorption by nuclear magnetic moments in a solid. Phys Rev 1946;69:37–38. Brownell, G. L., Sweet, W. H. Localization of brain tumors with positron emitters. Nucleonics 1953;11:40–45. Kuhl, D. E., Chamber, R. H., Hale, J., Gorson, R. O. A high-contrast photographic recorder for scintillation counter scanning. Radiology 1956;66:730–739. Anger, H. O. Scintillation camera. Rev Sci Instrum 1958;29:27–33. Rankowitz, S., Robertson, J. S., Higinbotham, W. A., Rosenblum, M. J. Positron scanner for locating brain tumors. Proc Inst Radio Eng Int Conv Rec 1962;9:49–56. Kuhl, D. E., Edwards, R. Q. Image separation radioisotoe scanning. Radiology 1963;80:653–662. Harper, P. V., Lathrop, K. A., Jiminez, F., Fink, R., Gottschalk, A. Technetium 99m as a scanning agent. Radiology 1965;85:101–109. Damadian, R. Tumor detection by nuclear magnetic resonance. Science 1971;171:1151–1153. Garroway, A., Grannell, P. K., Mansfield, P. Image formation in NMR by a selective irradiative process. J Phys Part C Solid State Phys 1974;7:L457–L462. Lauterbu, P. C. Image formation by induced local interactions – Examples employing nuclear magnetic-resonance. Nature 1973;242:190–191. Budinger, T. F., Gullberg, G. T. Letter: Three-dimensional reconstruction of isotope distributions. Phys Med Biol 1974;19:387–389. Hoult, D. I., Busby, S. J., Gadian, D. G., Radda, G. K., Richards, R. E., Seeley, P. J. Observation of tissue metabolites using 31p nuclear magnetic resonance. Nature 1974;252:285–287. Ter-Pogossian, M. M., Phelps, M. E., Hoffman, E. J., Mullani, N. A. A positron-emission transaxial tomograph for nuclear imaging (PETT). Radiology 1975;114:89–98. Phelps, M. E., Hoffman, E. J., Mullani, N. A., Ter-Pogossian, M. M. Application of annihilation coincidence detection to transaxial reconstruction tomography. J Nucl Med 1975;16:210–224. Kuhl, D. E., Reivich, M., Alavi, A., Nyary, I., Staum, M. M. Local cerebral blood vol-

14

64. 65.

66.

67.

68. 69.

70.

71.

Modo and Bulte ume determined by three-dimensional reconstruction of radionuclide scan data. Circ Res 1975;36:610–619. Kumar, A., Welti, D., Ernst, R. R. NMR fourier zeugmatography. J Magn Reson 1975;18:69–83. Jaszczak, R. J., Murphy, P. H., Huard, D., Burdine, J. A. Radionuclide emission computed tomography of the head with 99mCc and a scintillation camera. J Nucl Med 1977;18:373–380. Damadian, R., Goldsmith, M., Minkoff, L. NMR in cancer: XVI. FONAR image of the live human body. Physiol Chem Phys 1977;9:97–100. Reivich, M., Kuhl, D., Wolf, A. et al. Measurement of local cerebral glucose metabolism in man with 18f-2-fluoro-2deoxy-d-glucose. Acta Neurol Scand Suppl 1977;64:190–191. Mansfield, P., Maudsley, A. A. Medical imaging by NMR. Br J Radiol 1977;50:188–194. Edelstein, W. A., Hutchison, J. M., Johnson, G., Redpath, T. Spin warp NMR imaging and applications to human whole-body imaging. Phys Med Biol 1980;25:751–756. Young, I. R., Clarke, G. J., Bailes, D. R., Pennock, J. M., Doyle, F. H., Bydder, G. M. Enhancement of relaxation rate with paramagnetic contrast agents in NMR imaging. J Comput Tomogr 1981;5:543–547. Weinmann, H. J., Brasch, R. C., Press, W. R., Wesbey, G. E. Characteristics of gadoliniumDTPA complex: A potential NMR contrast agent. AJR Am J Roentgenol 1984;142: 619–624.

72. Nishimura, D. G., Macovski, A., Pauly, J. M. Magnetic resonance angiography. IEEE Trans Med Imaging 1986;5:140–151. 73. Kalender, W. A., Seissler, W., Klotz, E., Vock, P. Spiral volumetric CT with singlebreath-hold technique, continuous transport, and continuous scanner rotation. Radiology 1990;176:181–183. 74. Weissleder, R., Elizondo, G., Wittenberg, J., Rabito, C. A., Bengele, H. H., Josephson, L. Ultrasmall superparamagnetic iron oxide: Characterization of a new class of contrast agents for MR imaging. Radiology 1990;175:489–493. 75. Tjuvajev, J. G., Stockhammer, G., Desai, R. et al. Imaging the expression of transfected genes in vivo. Cancer Res 1995;55:6126–6132. 76. Wright, I. C., McGuire, P. K., Poline, J. B. et al. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage 1995;2:244–252. 77. Kinahan, P. E., Townsend, D. W., Beyer, T., Sashin, D. Attenuation correction for a combined 3d PET/CT scanner. Med Phys 1998;25:2046–2053. 78. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. 79. Ward, K. M., Aletras, A. H., Balaban, R. S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST). J Magn Reson 2000;143:79–87.

Section II General Techniques

Chapter 2 Magnetic Resonance Safety Andrew Simmons and Kristina Hakansson Abstract The safe operation of both clinical and pre-clinical MR systems is critical. There are a wide range of potential MR hazards. This chapter covers both the theoretical background to issues of MR safety and the guidance on more practical issues. The main sources of information on national and international MR safety guidance and advice are discussed, as well as local safety policies which are required for all MR installations. The projectile effect and other MR safety issues due to static and time-varying magnetic fields are considered, such as peripheral nerve stimulation, tissue heating and RF burns. Finally, contrast agents, auditory effects and medical implants and devices are discussed, as well as the less thought about issue of biological safety of clinical and pre-clinical MR systems. Key words: MR safety, MRI safety, MR safe, projectile effect, quench, SAR.

1. Legislation, Guidance and Best Practice

Legislation, guidance and best practice are all continually evolving within the field of MR safety, and there are often multiple sets of documents applicable within a single country (Table 2.1). This section cannot, therefore, be an exhaustive guide, and the reader is advised to consult local and national experts for the latest information. Good sources of up-to-date information include the International Society of Magnetic Resonance in Medicine and European Society of Magnetic Resonance in Medicine and Biology web sites (www.ismrm.org/mr_sites.htm#Spotlight and www.esmrmb.org) and www.mrisafety.com maintained by Professor Shellock. Regular MR safety updates are given at the ISMRM and ESMRMB annual conferences amongst others.

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_2, © Springer Science+Business Media, LLC 2011

17

18

Simmons and Hakansson

Table 2.1 Main hazards of MR imaging MRI hazards Static magnetic field

• Projectiles • Medical devices • Rotational forces • Lenz effect • Cryogenic liquids (magnet quenches)

Field gradients

• Peripheral nerve and muscle stimulation • Acoustic noise

Radiofrequency pulses

• Thermal heating • Contact burns • Induced current burns

Contrast agents

• Gadolinium side effects

Biological hazards

• Transmission of bacteria and viruses

National and international radiology organisations also provide guidelines and advice. At the time of writing, current international advice on exposure limits for patients and volunteers is given in the ICNIRP (International Commission on NonIonizing Radiation Protection) publications on MR procedures (1–5), as well as the IEC (International Electrotechnical Commission) International Standard 60601-2-33 Edition 2.1, published in 2006 (6). At a national level, the UK Health Protection Agency’s (HPA) published advice on “Protection of Patients and Volunteers Undergoing MRI Procedures” in 2008 (7), for example, is based on ICNIRP’s recommendations and focuses on their application in the UK. MHRA (Medicines and Healthcare products Regulatory Agency) has also produced guidelines based on UK and international MR safety advice (8). In the USA, the American College of Radiology produces guidance documents, the most recent one in 2007 (9). In addition, the FDA (Food and Drug Administration) publication entitled “Criteria for Significant Risk Investigations of Magnetic Resonance Diagnostic Devices” contains advice on exposure limits (10). Within the European Union EU Directive 2004/40/EC (11) contains a set of limits and action values affecting different frequencies of electromagnetic fields which initially stated that all EU member states should bring into force laws and regulations to comply with the directive by 30 April 2008. However, in an amendment to the directive (Directive 2008/46/EC (12)) this date was changed to 30 April 2012.

Magnetic Resonance Safety

19

2. Local Safety Policies All MR installations must have detailed local safety policies which are regularly reviewed and updated to reflect changes in working practices, updated safety guidance and national legislation. The requirements may differ slightly depending on whether the unit has been established for health care, human research or preclinical imaging, but the approaches to safety and the risks are the same. Local safety policies need to be written to cover the work of departmental staff, occasional visitors, cleaners, maintenance and security staff as well as the emergency services. Training should be provided to each group of staff who are likely to need access to the MR scanning unit, tailored to their specific needs. 2.1. Controlled Area

Access to the MR scanning suite is typically defined with respect to controlled areas. Two levels of controlled areas are used to avoid accidents involving pacemakers and projectiles. Both of these areas are three-dimensional in nature, since the magnetic field in the vertical direction needs to be considered to include floors above and below the MRI suite. • The MR controlled area includes all accessible areas where the magnetic field is above 5 G (0.5 mT). All entrances to the controlled area must have warning signs, and access should be restricted to authorised staff and screened patients (and their escorts). Other screened staff and screened visitors may also enter the controlled area if accompanied by an authorised member of staff. Persons with pacemakers or other medical implants must not enter this area. • The inner MR controlled area is defined by the 30-G (3 mT) field contour, which is inside the scanner room. This is used to reduce the risk of projectiles. No ferromagnetic objects should be brought into this area.

2.2. Classification of Persons

The overall responsibility for the safety of all staff and members of the public lies with the employer. In an MRI department, this responsibility is delegated to the responsible person. The responsible person updates the operational and safety policies, ensures adequate training and is responsible for the maintenance of safety facilities. The employer typically also appoints an MR safety adviser, who provides specialist advice on the scientific and technical issues relating to MR safety. Authorised persons are members of staff who have completed an MR safety induction and are authorised to enter the controlled

20

Simmons and Hakansson

area. The authorised person who is in control of the MR system at a given time is called the MR operator. 2.3. Local Rules

All MR departments should have a written set of local rules, which contain information about controlled areas, contingency plans, patient and equipment management and the names of the department’s responsible and authorised persons. The local rules are issued by the MR responsible person after full consultation with the MR safety advisor and representatives of all MR authorised personnel and should be reviewed and updated at regular intervals.

2.4. MR Safety Classification of Equipment

Items used in or near MR environments can be classified as MR safe, MR unsafe or MR conditional. The symbols and definitions of these terms are shown in Fig. 2.1.

Fig. 2.1. MR safe, MR unsafe and MR conditional markings. The MR environment is defined as the volume within the 5-G field contour.

Marking items with the relevant symbol is an efficient way of reducing the risk of accidents. For example, a defibrillator may be marked as MR unsafe and an aluminium patient trolley with no ferromagnetic parts as MR safe. 2.5. Safety in Practice

Classification of areas, persons and equipment makes it easier to control access to MR areas and prevents accidents from happening. The controlled area for one MR unit is shown in Fig. 2.2. All rooms inside the bold outer line are part of the controlled area.

3. Safety Screening All MR centres must have in place clear policies and procedures for screening anyone who may enter the MR environment.

Magnetic Resonance Safety

21

Fig. 2.2. Plan of MRI suite including control room, scanner room and equipment room with field lines.

Figure 2.3 gives an example of a form used to screen patients, research volunteers or staff to be scanned in a whole-body MR scanner. The form includes both questions designed to elucidate any safety queries and several questions designed to investigate the subject’s likely tolerance for the MR scan. In most instances, the form will be explained to the subject who will then fill in the form. An experienced member of staff will then go through the form with the subject, following up on any areas of concern and erring on the side of caution when deciding whether the subject can either enter the controlled areas or be scanned as appropriate.

4. MagnetRelated Safety Issues 4.1. Static Magnetic Field

The strength of the static magnetic field of a scanner is expressed in Tesla (T). One Tesla equals 10,000 G, and 1 G equals 0.1 mT. The earth’s magnetic field is approximately 0.05 mT (0.5 G). There are three principal types of MR magnets in use today. • The most widely used are superconducting magnets which rely on liquid helium to cool the specially constructed coil windings to extremely low temperatures close to absolute zero.

22

Simmons and Hakansson

Fig. 2.3. Example MR screening form.

• Permanent magnets are similar in concept to bar magnets in that they do not require cooling or an electrical power source to operate. • Resistive electromagnets require a permanent electrical source to operate. The magnetic field will cease once the power is turned off.

Magnetic Resonance Safety

23

Typically most MR magnets (superconducting and permanent magnets) are permanently on, even in the event of a power failure. Permanent and resistive magnets are much less common than superconducting magnets. One of the major MR safety issues relating to static magnetic fields is the projectile effect which happens when ferromagnetic materials are attracted by the main magnetic field. Figure 2.2 shows the fringe magnetic field for a whole-body MR system. The magnetic field increases rapidly close to the magnet in a nonlinear manner dependent on the magnet design. Ferromagnetic materials should therefore not be brought into the MR scanner room. Magnetophosphenes are flashes of light experienced by people in high magnetic fields which are thought to represent direct stimulation of the optic nerve and/or retina. They are generated by time-varying magnetic fields caused by movements of the head in a high magnetic field. They are not reported below 2.0 T but are experienced frequently at 4.0 T and above. Another effect experienced in high magnetic fields is the generation of a metallic taste which again has been reported to be generated by movement in a magnetic field. To date, there has been no conclusive evidence for irreversible or hazardous bioeffects of static magnetic fields, so the projectile effect for ferromagnetic materials remains the main safety concern for static magnetic fields. 4.2. Magnet Quenches

Although some low-field MR systems utilise permanent or resistive magnets, most high-field MR systems (1.5 T and above) use superconducting magnets. The magnet windings are cooled using liquid nitrogen or liquid helium in order to reach the low temperatures needed for superconductivity. Over time the amount of cryogens will gradually decrease due to low levels of boil off. If the amount of cryogens drops too low or the magnet begins to heat then an uncontrollable release of freezing gases termed a quench can occur. Modern MR systems should be fitted with an extraction system for these gases via an external piping system. The MRI suite should be fitted with a number of oxygen monitors at critical locations in order to detect any increase in nitrogen or helium caused by a quench. The volume of gas given off by an uncontrolled quench can be extremely large, and there is the potential for a patient to be lying on the scanner bed when a quench starts. Staff must be familiar with the procedures for evacuating a patient from the MR room full of freezing gases and should practise evacuation periodically. It is possible for high pressures to build up in MR rooms making it impossible to open the scanner door, so it may be necessary to break the viewing window to release the pressure. It is important to remember that the only remaining sign of a quench occurring overnight or at weekends may be an oxygen

24

Simmons and Hakansson

alarm ringing when staff start work the next day, indicating low levels of oxygen in the MR suite. MR scanners are typically provided with an emergency quench button in order to turn off the magnetic field in the case of an emergency. Quenching the magnet in this way should only be considered in extreme cases, such as a person trapped against the side of the magnet by a ferromagnetic item.

5. Time-Varying Gradient Fields Time-varying gradient fields are a key and integral part of MRI. Gradients are turned on and off rapidly, for varying durations and with varying maximum strength, depending on the pulse sequence and scanner under consideration. This has two main effects: peripheral nerve and muscle stimulation and acoustic noise. The induced current density J in a circular loop of radius r placed in a time-varying field B is given by J =σ×

dB r × 2 dt

where σ is the conductivity of the material, in this case the type of tissue carrying the current. Faraday’s law of induction means that a time-varying magnetic field induces a voltage in a conductor and the induced voltage can lead to an induced current. The rate of change of the gradient field is termed dB/dt, and this can vary greatly depending on the scanner, pulse sequence and application. The time-varying gradient fields have negligible thermal effects, but strong time-varying gradient fields could potentially lead to seizures, magnetophosphenes, changes in nerve conduction, peripheral nerve stimulation, cardiac arrhythmias or cardiac arrest. Peripheral nerve stimulation (PNS) and muscle stimulation can occur when a time-varying magnetic field induces electric currents in nerve and muscle cells. Peripheral nerve stimulation occurs at up to 5 kHz. At frequencies of about 10–100 Hz, cardiac muscle stimulation may lead to ventricular fibrillation. The threshold current density for this is about 1.2 A/m2 , so it can be avoided by keeping the current densities below 0.4 A/m2 . The maximum gradient strength has increased substantially over the last two decades with improvements in engineering and applications such as echo planar imaging and diffusion imaging which both rely on strong rapidly switching gradients. The

Magnetic Resonance Safety

25

threshold for nerve excitation in the human body varies greatly with the lowest threshold being for retinal neurones, then largediameter peripheral nerves, small-diameter peripheral nerves and finally cardiac muscle. The strongest concern focuses around cardiac excitation in the impaired patient. Clinical scanners are designed with restrictions to ensure that only peripheral nerve excitation, if at all, is possible. With the patient’s nose at isocentre, large-diameter peripheral nerve excitation tends to occur in the lower back, while with the patient’s naval at isocentre, excitation is mostly likely to occur at the shoulder. The y-gradient tends to be most effective in producing excitation, and the shape of the radiofrequency pulses used is important (13).

6. Radiofrequency Effects

The radiofrequency (RF) fields used to manipulate the magnetisation lead to induced currents in the body, which in turn causes power dissipation, i.e. heating. The amount of heating that is acceptable for any particular organ depends on its blood flow, since blood carries the heat away and spreads it through the body. In general, human tissues can tolerate a rise of about 1◦ C. Concern here is most focused on compromised patients and on organs without thermoregulation, such as the eyes, or those that are particularly heat sensitive, such as the reproductive organs. The quantity used to measure RF exposure is the specific absorption rate (SAR), defined as SAR =

σ × E2 2ρ

where σ is the conductivity, E is the induced electric field and ρ is the density of the tissue. The factor of 1/2 comes from averaging over time for an alternating field. SAR is measured in units of W/kg. There are limits on whole-body SAR for different operating modes of the scanner, which are calculated from limits on temperature rises (see Table 2.2). For patients with metallic implants, heating is more of a concern because the metal will absorb more energy (since they have higher conductivity than tissue), and this heat will spread to surrounding tissues. Another heating effect of RF pulses is RF burns. These burns are caused by highly concentrated absorption of RF energy at a single point, resulting in local increases in temperature and (if

26

Simmons and Hakansson

Table 2.2 SAR values averaged over 6 min Operating mode

Limit on core temperature Whole-body SAR rise (◦ C) limit (W/kg)

Normal

0.7

2

First-level controlled

1

4

Second-level controlled

>1

>4

the temperature is high enough) tissue burning. RF burns occur when there is a conductive loop. The risk can be minimised by avoiding loops of conductor, e.g. keeping ECG leads from forming a loop. Loops in the patient’s body should also be avoided, such as clasped hands. Padding can be used to position the patient correctly.

7. Use-Related Safety Concerns 7.1. Contrast Agents

Contrast agents are often used for clinical MRI imaging and are also used less frequently for some research applications, such as dynamic susceptibility contrast MRI to measure perfusion and contrast-enhanced MR angiography. For example, gadoliniumbased contrast agents are widely used for brain and spine imaging, as well as for contrast-enhanced MR angiography. There are also a variety of organ-specific contrast agents, such as liver contrast agents. Side effects from MR contrast agents are generally low, for example often showing no significant difference between an injection of a gadolinium-based contrast agent and a placebo injection of saline. Side effects in small numbers of patients may include nausea, headache, vomiting and hives. Anaphylactic reactions may occur in 1 in 500,000 subjects, and the MR unit must be prepared for the possibility of this. One major side effect of some gadolinium-based contrast agents is nephrogenic systemic fibrosis (NSF) which can affect patients with kidney disease. This leads to swelling and tightening of the skin with large areas of hardened skin. Glomerular filtration rate should be measured for all patients with kidney disease prior to deciding on the use of gadolinium contrast agents.

7.2. Auditory Effects

Sound is produced by Lorentz forces acting on the MR gradient coils. Noise levels as high as 135 dB have been measured in

Magnetic Resonance Safety

27

MR systems. Fast pulse sequences and higher field MR systems in particular can lead to higher levels of noise in MRI systems. Sound levels are a safety issue for both patients and staff. Wearing ear plugs typically reduces noise by 10–20 dB and MRcompatible headphones by more than this. Staff should always wear headphones when in the MR scanner room while the scanner is operating, and patients should be provided with effective ear protection matched to the application in hand. 7.3. Implants and Devices

Some patients or staff will have an implant or device, such as an aneurysm clip, cardiac pacemaker or metal screw, used to fix a broken bone. A particularly good resource is a book by Frank G. Shellock entitled Reference Manual for MR Safety, Implants and Devices which is updated annually, the latest version at the time of writing being the 2009 edition (14). Over 2,300 objects have been tested in the MR environment and are reported in the book, typically at a field strength of 1.5 T. Approximately 900 of these have additionally been tested at 3 T. Manufacturers of implants will often produce many types of an implant, such as an aneurysm clip over time, some of which may be MR unsafe, while others may be MR safe. MR conditional implants may be safe with a particular combination of field strength, maximum spatial gradient and RF coil position but unsafe with other combinations. The static magnetic field can exert a rotational force on nonspherical ferromagnetic objects. Some implanted clips, such as aneurysm clips, can twist inside the patient causing injury or death. The static magnetic field can also interact with implanted medical devices, such as pacemakers and defibrillators. Even at low fields of about 1 mT, the magnetic field can alter the operating mode of some pacemakers.

7.4. Biological Safety

There is often a strong emphasis on MR-specific safety issues in relation to MR scanners and peripheral equipment, such as MRcompatible patient monitors. It is important, however, to ensure that biological safety is also considered for MR installations. Like any other area of a hospital or pre-clinical facility, there is the potential for the transmission of bacteria, viruses and other biological hazards. The scanner bore, control panels, MR door handles and the scanner room must be cleaned regularly according to facility guidelines, using appropriate cleaning techniques and MR safe cleaning equipment.

7.5. Physiological Monitoring

Physiological monitoring is important for three key reasons in MRI. First, respiratory and cardiac/peripheral monitoring is a requirement for some MR applications. Second, physiological monitoring is necessary for some research applications. Lastly, physiological monitoring is required for some patients,

28

Simmons and Hakansson

particularly those who are impaired in some way or are under general anaesthetic. Specific MR-compatible physiological monitoring equipment may be required for these applications.

References 1. ICNIRP. Statement on medical magnetic resonance (MR) procedures: Protection of patients. Health Phys 2004;87(2): 197–216. 2. Amendment to the ICNIRP. Statement on medical magnetic resonance (MR) procedures: Protection of patients. Health Phys 2009;97(3):259–261. 3. ICNIRP. Guidelines on limits of exposure to static magnetic fields. Health Phys 2009;96(4):504–514. 4. ICNIRP. Guidelines for limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz). Health Phys 1998;74(4):494–522. 5. ICNIRP. Statement on the “guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to 300 GHz)”. Health Phys 2009;97(3): 257–259. 6. IEC 60601-2-33 Consol. ed2.1 (incl. am1): Medical electrical equipment – Part 2: Particular requirements for the safety of magnetic

7. 8. 9. 10.

11. 12. 13. 14.

resonance equipment for medical diagnosis, 2006. HPA Protection of patients and volunteers undergoing MRI procedures: Advice from the Health Protection Agency, 2008. MHRA Device Bulletin: Safety Guidelines for Magnetic Resonance Imaging Equipment in Clinical Use (DB2007(03)), 2007. ACR. Guidance document for MR safe practices. AJR 2007;188:1–27. FDA Guidance for Industry and FDA Staff: Criteria for Significant Risk Investigations of Magnetic Resonance Diagnostic Devices, 2003. EU Directive 2004/40/EC EU Directive 2008/46/EC Abart, J. et al. Peripheral nerve stimulation by time-varying magnetic fields. J Comput Assist Tomogr 1997;21(4):532–538. Shellock, F. G. Reference Manual for Magnetic Resonance Safety, Implants, and Devices 2009. Los Angeles, CA: Biomedical Research Publishing Company; 2009.

Chapter 3 Measuring the Absolute Water Content of the Brain Using Quantitative MRI Nadim Joni Shah, Veronika Ermer, and Ana-Maria Oros-Peusquens Abstract Methods for quantitative imaging of the brain are presented and compared. Highly precise and accurate mapping of the absolute water content and distribution, as presented here, requires a significant number of corrections and also involves mapping of other MR parameters. Here, either T1 and T2 ∗ or T2 is mapped, and several corrections involving the measurement of temperature, transmit and receive B1 inhomogeneities and signal extrapolation to zero TE are applied. Information about the water content of the whole brain can be acquired in clinically acceptable measurement times (10 or 20 min). Since water content is highly regulated in the healthy brain, pathological changes can be easily identified and their evolution or correlation with other manifestations of the disease investigated. In addition to voxelbased total water content, information about the different environments of water can be gleaned from qMRI. The myelin water fraction can be extracted from the fit of very high-SNR multiple-echo T2 decay curves with a superposition of a large number of exponentials. Diseases involving de- or dysmyelination can be investigated and lead to novel observations regarding the water compartmentalisation in tissue, despite the limited spatial coverage. In conclusion, quantitative MRI is emerging as an unparalleled tool for the study of the normal and diseased brain, replacing the customary time–space environment of the sequential mixed-contrast MRI with a multi-NMR-parametric space in which tissue microscopy is increasingly revealed. Key words: Quantitative imaging, water mapping, T1 mapping, T2 ∗ mapping, T2 mapping, brain imaging.

1. Introduction In the last few years, the novel approach of quantitative MRI (qMRI) has gained importance. In comparison to conventional qualitative MRI, qMRI is an unbiased measurement method and M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_3, © Springer Science+Business Media, LLC 2011

29

30

Shah, Ermer, and Oros-Peusquens

allows for more straightforward statistical modelling. Especially for human brain imaging, it provides an attractive method to study changes in the brain caused by disease. Moreover, qMRI facilitates large national or international multi-centre studies since the data can be more directly compared, as well as longitudinal experiments for precise disease evaluation. Degenerative diseases, such as multiple sclerosis (MS), can be more accurately monitored and investigated in order to better understand, for instance, the correlation of lesion burden and physical disabilities, and especially their long-term behaviour. Quantitative measures of relaxation parameters, such as T1 and/or T2 , provide a good basis for the detection of cerebral abnormalities in many diseases. The detection of disease can potentially occur at an earlier stage than allowed by standard imaging methods, as significant confounding effects present in conventional approaches are explicitly corrected in qMRI. A number of methods for mapping of tissue relaxation parameters and water content have been published (1–10). We will address the methods developed in our group in more detail below and present a brief overview of the few other existing methods. Recently, Warntjes et al. (11) presented a novel method for the simultaneous quantification of T1 , T2∗ , proton density and B1 field. Using a multi-echo acquisition of a saturation recovery with a turbo spin-echo readout approach, the authors were able to acquire full brain water content maps with a resolution of 0.8 × 0.8 × 5 mm3 at 1.5 T in 5 min. The published water maps, however, show a rather inhomogeneous distribution reminiscent of B1 effects. Preibisch et al. (12) suggested the use of exponential excitation pulses for a more accurate estimation of S0 in 2D acquisitions, which forms the basis for water content mapping in a similar manner to that presented later in this chapter. The water content of the brain is a very sensitive measure for a variety of pathologies, such as MS, hepatic encephalopathy (HE) or brain tumours. The non-invasive determination of the local or global increase or decrease of brain water content enables an exact differentiation of, for instance, tumour and oedema. This differentiation may avoid a possible lethal side effect for the patient arising from elevated pressure in the skull cavity. The human body consists of up to 90% water distributed either within (intracellular) or outside the cells (extracellular). To actively participate in biological processes, water interacts with biological macromolecules to form a hydration layer (“bound water”), which is involved in processes such as metabolism or biosynthesis, or acts as a solvent and/or transport medium throughout the biological system (“free water”). In the first place, the direct influence of the bound and intracellular water pools on relaxation properties enables the differentiation of tissues by

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

31

NMR/MRI (13), since the free and extracellular water might be expected to have more homogeneous properties across tissue. There are several characteristics of hydration water which play a crucial role in this context, such as restricted motion and partial orientation of water molecules, distribution of correlation times, multiple frequencies of motion, rotation and anisotropic diffusion. These properties of water near the surface of macromolecules result in time-dependent, local magnetic field fluctuations which influence the relaxation mechanisms and therefore provide valuable information about the molecular states and dynamic structure of water at macroscopic and microscopic levels (13). Numerous biological processes, both normal and pathological, can modify the state of water in vivo. Possible reasons for a change in the relaxation times, as well as in the water content, are different phases of cell cycle and cell growth, changes in the physiological pH value, age and maturation or metabolism (13). For instance, it has been found that faster growing cells demonstrate an increased T1 relaxation time. Such fast-growing cells are possible indicators for tumours (13). Physical conditions, such as the state of nutrition, alcohol or drug abuse, or stress, might influence the water balance and therefore also the relaxation times. Three major aspects should be investigated to address the water characteristics in vivo: the total water content, the microscopic (intra/extracellular distribution) and macroscopic (anatomy-related) distribution of water and macromolecular– water interactions. Changing one or more of these factors can alter the tissue-specific water balance in vivo. Abnormal values of water content and distribution, of T1 and T2 , will follow and can be characterised by MRI. The NMR relaxation times are field dependent, and relaxation time-based comparisons between diseased and normal states should use data acquired at the same field strength. The relaxation times are also temperature dependent and can change, e.g. due to fever or hyperthermia treatment. These aspects should be carefully considered in a proper experimental design (7, 13). The MRI-measured water content, on the other hand, is in principle field independent and could be used for comparison of data acquired at different field strengths. However, a careful assessment of possible changes in the MR-visible water compartment with field must be performed and is not available at present. Over the past few years, a fast, precise and accurate method for quantitative mapping of the absolute water content of the human brain in vivo has been developed in Juelich. Accurate information on cerebral water content is of high relevance for many diseases and has been specifically used for the study of hepatic encephalopathy (14).

32

Shah, Ermer, and Oros-Peusquens

The first part of this chapter describes two different methods to quantitatively map the free water compartment in the brain in vivo. Both strategies have been developed at the authors’ Institute and are based on the combination of fast multi-slice and multitime point MR sequences. The common basis of both approaches is the QUTE (quantitative T2∗ image) sequence1 which is suitable for rapidly mapping the T2∗ relaxation time (15, 16) and is used to extrapolate the signal to echo time, TE = 0. The major difference between the two procedures is the way in which the longitudinal relaxation time is mapped. That is, the first approach uses an additional MR sequence, called TAPIR (T1 mapping with partial inversion recovery), to obtain a series of T1 -weighted images and produce T1 maps (1, 2, 17). This method, which is based on the Look–Locker (18) approach with FLASH readout and an innovative interleaving of slices and time points, provides highresolution T1 maps in clinically applicable times. The strength of TAPIR is that it enables the acquisition of multiple time points and thereby enables accurate fitting of the resulting data. In addition, multi-component T1 relaxation can be investigated. The most significant disadvantage of FLASH-based Look–Locker methods, the long acquisition time, can easily be ameliorated through the use of saturation instead of inversion recovery and the acquisition of more than one phase-encoded echo (FLASHEPI readout) for a given slice and inversion time. The use of sliceselective inversion pulses (19) can further speed up the method. The TAPIR-QUTE method for water content mapping has the significant advantage of delivering accurate T1 and T2∗ maps, providing a three-quantitative-parameter description of the brain. An alternative solution to determine the longitudinal relaxation time is to acquire a second set of images with the QUTE sequence with a different TR and/or different flip angle (20, 21). The method can provide a speed-up of the data acquisition but usually involves a decrease of the accuracy of T1 mapping. Further details are described in Section 2.3. The different combinations of QUTE and TAPIR have led to two established methods which are both capable of quantitatively mapping the absolute water content in vivo and result in both accurate and precise maps of the free water content of the brain. The second part of this chapter gives a current literature review of techniques for myelin water mapping and its applications.

1

This sequence has recently become available on Siemens scanners as the multiecho variant of the standard multi-slice gradient echo. Other manufacturers also have similar product sequences.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

33

An important question related to the mapping of the brain water content with MRI is whether all mobile water protons have been detected. If not, then the degree of mobility and whether other protons have contributed to the signal are also of importance. In addition to the water protons, MR signals can be generated by protons bound in molecules, such as lipids, proteins and nucleic acids – non-aqueous protons. However, this signal decays to zero in less than 100 μs, due to dipolar interactions of the nonaqueous protons with the other nuclei bound in the molecule. In contrast, the signal from water protons in tissue has T2 times of longer than 10 ms (22–24). Conventional MRI therefore detects the signal from brain water with no contamination from the fast decaying non-aqueous tissue. The relaxation properties of water protons are sensitive to the microscopic water environment and thus are expected to vary over dimensions much smaller than the voxel size. It has been recognised over the years that the MRI signal from water in the healthy human brain is generated from three different components: a long T2 component (~2 s), an intermediate component (~80 ms) and a short T2 component (~20 ms) (22, 24, 25). The long component can be easily attributed to CSF and the intermediate one to intra/extracellular water. The short component is thought to be due to myelin water. Myelin is a fatty sheath that surrounds neurons and allows for faster conduction of electrical impulses. It consists of numerous windings (lamellae) with a spacing of the order of 150 Å. According to Brownstein and Tarr (26, 27), the observed T2 of water in a confined compartment can be described as a function of the bulk T2 (T2,bulk ), the radii of the compartment along the x, y and z direction (Rx,y,z ) and the rate of wall relaxation or surface sink strength density (H): 1/T2,obs = H (1/Rx + 1/Ry + 1/Rz ) + 1/T2,bulk. The shortest T2 component can therefore be attributed to water in the most restricted environment. However, observation of the fast relaxing species is hindered by the relatively small contribution this compartment makes to the overall measured signal. For a reliable fit of the fast relaxing component, the short TE (<50 ms) portion of the recovery curve needs to be sampled with a large number of points. Even when the decay curve is well characterised, the ability to separate the signal from a number of distinct compartments depends on the condition that T2 is much less than the mean residence time tm of the water protons in each physical environment. Here tm represents the average time a water molecule spends in a given compartment before exchanging to another one. Under this condition, there is insufficient time for the magnetisation to “sample” both water environments before

34

Shah, Ermer, and Oros-Peusquens

the phase coherence between the spins is irreversibly lost. Consequently, the two compartments may be considered essentially isolated, and the spin-echo signal is a linear summation of the signal from each T2 species. The loss of the myelin sheath insulating the nerves (demyelination) is characteristic of some neurodegenerative autoimmune diseases, including multiple sclerosis, acute disseminated encephalomyelitis, transverse myelitis, and chronic inflammatory demyelinating polyneuropathy. The defective structure and function of the myelin sheath is called dysmyelination; unlike demyelination, it does not produce lesions. Such defective sheaths often arise from genetic mutations affecting the formation of myelin. The shiverer mouse represents one animal model of dysmyelination. Human diseases where dysmyelination has been implicated include leukodystrophies (Pelizaeus–Merzbacher disease, Canavan disease, phenylketonuria) and schizophrenia. A significant number of people are affected by diseases that either break down the myelin (MS alone affects about 2.5 million people worldwide) or impair its initial growth. For the study of such diseases, and to gain a better understanding of the normal development of the brain, it is important to have an in vivo technique able to measure myelin water. The short T2 component, attributed to myelin water, was first observed in vivo in the pioneering work by MacKay et al. (25). The ratio of the short T2 signal (myelin water) to the total signal (total water content) gives the myelin water fraction (MWF) (28). This quantity has been shown to correlate strongly with histological staining for myelin (29–32). The multi-component T2 relaxation has been characterised in the healthy brain and spinal cord as well as in several diseases such as multiple sclerosis, schizophrenia, phenylketonuria and Alzheimer’s disease (25, 33–39). A multi-component T2 mapping method based on rapid steady-state (SSFP) imaging has been very recently reported (10, 40). The so-called driven-equilibrium single-pulse observation of T1 (DESPOT1) and T2 (DESPOT2) (8, 9) has been extended in order to be able to include multi-component relaxation (mcDESPOT) of both relaxation components. The quantification of multi-compartment T1 and T2 values of the whole brain is obtained from data acquired in measurement times of between 16 and 30 min, followed by lengthy computer processing. The exchange of protons between two compartments is taken into account in simulations of the signal delivered by the SSFP sequence. The relevant parameters of each compartment, such as relaxation times, occupancy and residence time, can be adjusted to best reproduce the data (SSFP scans acquired for 15 different flip angles). While this approach offers the indisputable advantage of fast 3D whole-brain mapping, the analysis is cumbersome, difficult to implement and potentially more

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

35

prone to inaccuracies than the non-negative least squares (NNLS) method. Furthermore, banding artefacts in SSFP measurements, resulting from imperfect shimming or chemical shift effects, are notorious.

2. Free Water Mapping In the brain, most protons (that is, hydrogen nuclei) belong to water molecules and therefore the proton density is often directly referred to the water content of the tissue. Mapping of the water content of the brain with MRI is based on mapping the proton density (PD). The concentration of protons in the tissue can be directly measured by MRI and expressed in percentage units, pu, as the percentage of the proton concentration in tissue with respect to pure water (7). Thus cerebrospinal fluid (CSF) has about 100 pu water and white matter (WM) of the brain about 70 pu. In principle, the signal measured with an MR sequence with ultra-short echo time, TE, and long repetition time, TR ~ 5 T1 , directly reflects the proton density distribution. Given that the dominant contribution to the MR signal intensity is from protons in water molecules, the proportionality can be used to quantify the absolute water content in tissue if imperfections, such as radiofrequency (RF) non-uniformity, have been properly corrected (7). Using QUTE, a series of T2∗ -weighted images at different TEs is acquired and is then extrapolated back to zero echo time by exponentially fitting the measured signal for each voxel to determine the tissue equilibrium magnetisation, S0,T2 ∗ (tissue). Since S0,T2 ∗ (tissue) is directly proportional to the proton density, the absolute water content can be derived from this parameter (3, 4, 6). In order to be able to transfer S0,T2 ∗ (tissue) into a quantitative measure of the water content, a reference probe filled with 100% water needs to be placed in the field of view (FoV) during the whole measurement. The measured equilibrium magnetisation of the reference probe, S0,T2 ∗ (reference probe), can then be correlated to S0,T2 ∗ (tissue) (3, 4, 6). Usually, the signal decay arising from T2∗ effects is described by an exponentially decaying function and is caused by spin–spin interactions, magnetic field inhomogeneities and susceptibility effects. Depending on the structure of the sources of inhomogeneity, the MR signal decay can well deviate from being exponential (41). The changes of the magnetic field across an imaging voxel can be described in first order as linear. If the voxel has

36

Shah, Ermer, and Oros-Peusquens

constant density and perfect shape, the integrated effect of this inhomogeneity will give rise to a product of three sinc modulations (one for each dimension) of the signal decay. A simple way to correct the effect of the presence of an intra-voxel gradient in slice selection direction on the signal decay is offered in (12). In general, the signal decay depends on the combined effect of the local magnetic field distribution, the slice profile which in turn depends on the RF pulse and flip angle, as well as the distribution of the proton density inside the voxel. Analytical models for the MR signal in the presence of static inhomogeneity sources are based on assumptions made about the distribution of field-disturbing sources, such as a randomly oriented network of cylinders (41). Thus, the resulting signal strongly depends on the strength of the intra-voxel field inhomogeneity, and the aspect must be properly treated for the determination of S0,T2 ∗ . This is a very important issue for water mapping and can give the largest correction factor, as well as the one whose accuracy is the most difficult to control. In order to accurately quantify the absolute water content in vivo, the measured signal intensity, influenced by several sources of error, needs to be corrected. In the following, the four correction factors taken into account for the final water content map are described in general. The two water mapping approaches developed in our group use different solutions to calculate most of the individual correction factors, and thus each of these is described further below. Because both the water density and the net magnetisation created by proton orientation in an applied magnetic field depend on temperature by a relation which can be considered linear over a restricted temperature range (42), the correction of temperature differences between brain tissue (assuming a body temperature of 37◦ C) and the reference probe is required. The temperature in the reference probe is therefore recorded throughout the measurement, resulting in a correction factor CT . For instance, at a field strength of 1.5 T, the net magnetisation depends on temperature with a rate of –0.32%/◦ C. Consequently, the in vivo MR signal needs to be scaled according to the ratio of the temperature of the reference probe and the temperature of the brain (body). Even at field strengths below 3 T, RF non-uniformities are a serious problem in qMRI and can influence the measurements of MR parameters such as T1 and tissue water content tremendously. RF non-uniformities, resulting from a spatially varying flip angle (FA), lead to a spatially varying signal intensity throughout the volume to be imaged. By calculating the exact flip angle in each voxel and incorporating this flip angle map into the final water content map, spatial variations of the measured signal intensity can be easily corrected (with a correction factor, CB1 ).

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

37

Using the QUTE sequence, the effective TR becomes relatively long, but even so the longitudinal relaxation does not fully recover within one cycle. This effect leads to signal saturation which propagates into the final water content map and therefore needs to be corrected (with a correction factor, CT1 ). Finally, one needs to correct for imperfections originating from the coil used for signal reception, resulting in a correction factor, CRC . The temperature correction is rather straightforward to perform and results in a global shift of the water distribution. It involves the (well-verified) assumption that the temperature distribution over the healthy brain is homogeneous. The other correction factors can be divided into two main categories: B1 related (transmit and/or receive) and relaxation time related (T1 and T2∗ ). These correction factors, addressed below, can have a large variation over the brain and influence the details of water distribution as well as its global properties. 2.1. Relaxation Time Mapping

In order to map the T1 and T2∗ relaxation times, two MR sequences are used and they are described in the following. Both TAPIR and QUTE are fast multi-slice and multi-time point sequences, and in combination they lead to the two approaches for the localised and quantitative measurement of the absolute water content of the brain in vivo. By merging the essentials of two fast imaging sequences, the echo planar imaging (EPI (43)) sequence and the fast low angle shot (FLASH (44)) sequence, QUTE (quantitative T2∗ image) is suitable for rapidly mapping the T2∗ relaxation time (15, 16). Figure 3.1 shows the pulse diagram of the multi-slice and multitime point QUTE sequence, combining a multi-echo FLASH-like acquisition and an oscillating EPI-like readout train. Following application of a slice-selective pulse, the signal is phase encoded

Fig. 3.1. QUTE MR pulse sequence diagram (1–4). Following a slice-selective pulse, the signal is phase encoded for the highest line of k-space and two gradient echoes are read out. This loop is repeated until the required number of echoes have been acquired.

38

Shah, Ermer, and Oros-Peusquens

(highest frequency line) and a set of two gradient echoes is read. The gradients have opposite polarity and could, in principle, be separated by a time delay, which can be freely configured. The loop consisting of the two gradients is then repeated until the desired number of echoes has been read. This scheme is repeated for the remaining slices, and the highest line of k-space before the whole procedure is repeated for the next line of k-space. The measurement is complete when all lines of k-space have been acquired. This implementation allows for the acquisition of a large number of points on the T2∗ relaxation curve. Using this interleave looping structure increases the effective TReff to TR times number of slices (3, 4, 15) and therefore enables employment of higher flip angles in order to maximise the signal-to-noise ratio (SNR). Using TAPIR, high-resolution, multi-slice T1 relaxation time mapping can be performed to acquire a series of T1 -weighted images to produce T1 maps (1, 2). Figure 3.2 represents the sequence diagram of TAPIR with EPI readout factor of 3 (arbitrarily chosen); the separation of the 90–180◦ pulses is not shown to scale in this diagram. Following the application of a non-selective 90◦ pulse, the transverse magnetisation thus created completely dephases during the long delay period, τ (~2 s). After this delay time, a non-selective 180◦ pulse inverts all recovered magnetisation and any residual transverse magnetisation is immediately spoiled by means of a large crusher gradient. Directly thereafter, the inverted magnetisation is sampled in the following way: the most peripheral line in k-space of each band is acquired by the application of a single, slice-selective α pulse. The α pulse excitation module is repeated for the next slice, but again for the same three lines in k-space. Following the acquisition of n slices, the whole procedure, for the same three

Fig. 3.2. Pulse sequence diagram of the TAPIR sequence (1). The spin system is prepared by a sequence of two non-selective pulses with flip angles 90◦ and 180◦ , respectively, separated by a delay, τ . The delay, τ , is sufficiently large to allow full dephasing of transverse magnetisation created by the 90◦ pulse. Any transverse magnetisation left after the inversion pulse is removed by means of spoiler gradients in all three directions. Following the inversion pulse, the spin system is repeatedly sampled in a Look–Lockertype acquisition scheme by the application of slice-selective a-pulses, interleaving readout of different slices and time points.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

39

k-space lines, is repeated starting at slice 1; this loop ensures the acquisition of multiple time points. Following acquisition of the required number of time points and slices, a new 90–180◦ pulse combination is applied and the next three lines, one from each k-space band, are acquired in an identical manner. By judicious application of blipped phase-encoding gradients it is ensured that the echoes belong to different bands of k-space. Prior to image reconstruction, echoes labelled “ε 2 ” are time reversed, centred, and phase corrected. The principle of interleaving slices and time points allows extremely efficient data acquisition and enables the recovery curve to be sampled with a much higher apparent temporal resolution than other competing methods. Here, the term apparent is used to denote the fact that the sampling of the magnetisation recovery is achieved through multiple shots; this has to be distinguished from the real temporal resolution achievable in single-shot methods. The overall speed of the sequence is enhanced not only by the rapid repetition of the sampling radiofrequency pulses but also by minimising the dead time needed for the longitudinal magnetisation recovery. Additionally, TAPIR uses readout gradient reversals to acquire multiple lines of k-space after a single excitation pulse and subsequently reorders the acquired data to fill k-space segments. The number of k-space segments is freely determinable within the constraints of the experiment. More segments lead to a much faster acquisition, but care has to be taken to avoid potential image artefacts. The relatively long recovery time for in vivo applications is used to facilitate multi-slice acquisition. This has the great advantage that whole-brain coverage can be achieved. However, it is noted that the total acquisition time for a given measurement is a compromise between the number of slices, the number of k-space segments and the number of time points required for fitting. 2.2. Water Mapping: TAPIR and QUTE 2.2.1. Methods

In the method presented in this section the calculation of the proton density is based on the extrapolation to zero echo time of the T2∗ relaxation curve as extracted from QUTE (15, 16) with subsequent application of the different corrections to result in water maps. This approach enables the acquisition of minimally distorted maps, and differences in T2∗ are implicitly considered by the extrapolation. Fitting the T2∗ relaxation curve results in the determination of the parameter S0,T2 ∗ (tissue) which is the tissue equilibrium magnetisation determined from the T2∗ map. S0,T2 ∗ (tissue) can be translated into a quantitative measure of absolute localised water content by reporting it to S0,T2 ∗ (reference), the reference equilibrium magnetisation of the water reference probe.

40

Shah, Ermer, and Oros-Peusquens

A number of corrections are then applied. The correction accounting for the difference between the temperature of the probe and that of the tissue under investigation gives a factor CT . The transmit B1 field inhomogeneities cause a local variation of S0,T2 ∗ . In order to account for them, the measurement of the effective flip angle, α eff , at each point within the object is required, resulting in a correction factor CB1 . The receiver coil inhomogeneity needs to be considered as a correction factor CRC because it directly influences the amplitude of the measured signal. Separate corrections for receive and transmit inhomogeneities are necessary, and they are different whenever the transmit and the receive coil are not identical. Differences in T1 saturation need to be incorporated into the final analysis. In addition to T1 , the corresponding correction factor, CT1 , explicitly depends on the effective flip angle and therefore CB1 Because maps acquired with TAPIR and QUTE are essentially calculated from gradient echo images, they can be simply overlaid without additional registration procedures in the absence of volunteer motion. This makes them favourable for H2 O content mapping as T1 , and S0,T2 ∗ information has to be perfectly coregistered for the reconstruction of high-precision water content maps. After inclusion of all the described correction factors, the localised quantitative water content, WMR , is then given by WMR =

S0,T 2 · (tissue) · CB1 (tissue) · CRC (tissue) · CT 1 (tissue) . S0,T 2 (reference) · CT CB1 (reference) · CRC (reference) · CT 1 (reference)

[1]

This relation accounts for the number of protons in a voxel and relates this to the number of protons that are contained in an ROI of 100% H2 O. This measure of the water content is, therefore, a dimensionless number that represents the relative amount of water in an image voxel. Determination of each quantity appearing in Eq. [1] is described below. As mentioned earlier, the signal behaviour can differ from being exponential and depends on the details of frequency and density distribution inside the voxel. It is, in principle, possible to rely on a reasonable model to fit the measured data points and extract S0,T2 ∗ from the fit. On the other hand, for a high-precision measurement of water content in vivo, it is crucial to determine S0,T2 ∗ with a minimum systematic bias as this directly results in a systematic error in WMR . Any prior decision about the use of specific models for fitting would result in a bias (at least for some regions of interest) that is not tolerable if high-precision results are desired. Thus, a different approach was chosen here. Instead

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

41

of relying on a specific analytical signal model, the first point on the signal decay curve was acquired with an echo time, TE, as short as possible and a polynomial of third order was used to extrapolate the signal back to TE = 0. In this approach, special care was taken to prevent artificial extrapolation values that result from the bad “predictive power” of the polynomial function by a proper variation of the number of points used for fitting. Details of the fitting method are given in Neeb et al. (4), where it is shown that precisely in regions where the exponential fit performs poorly, the polynomial fit enables recovery of reasonable values of water content and, indeed, even enables recovery of anatomical detail. Spatial variations in the B1 excitation field are an important source of systematic error in the study of absolute water content. As such, the local flip angle should be determined for each subject individually. This can be performed without additional measurements by using the fact that S0,T2 ∗ and S0,T1 are acquired with different nominal flip angles of α QUTE = 90◦ and α TAPIR = 25◦ , respectively. The effective flip angle, α eff , for a nominal QUTE excitation pulse at each voxel can be extracted from the ratio of extrapolated QUTE and TAPIR signal intensities S0,T2 ∗ and S0,T1 , respectively, because it depends on the effective flip angle, α eff  CT 1(αeff ) sin kM ) (α 0 eff S0,T 2∗   = [2] αTAPIR S0,T 1 CT 2∗Decay . kM0 CT sin αQUTE αeff Here, the numerator is given by the analytical expression for the extrapolated signal intensity of the T2∗ measurement, S0,T2 ∗ , while the denominator represents the corresponding expression for T1 , S0,T1 . Both values depend on the equilibrium magnetisation, M0 , and a constant, k. The latter includes factors, such as the receiver gain or a scaling factor from the Fourier transform which depends on the matrix size. Importantly, the constant k has to be the same for both measurements in order to cancel out, that is, matrix size and receiver gain have to be identical. Otherwise an appropriate correction must be performed. CT2∗ Decay and CT1 (α eff ) correct for decay and saturation effects of S0,T1 and S0,T2 ∗ , respectively. The term CT in the above equation corrects for the signal difference caused by temperature changes in the water probe between the acquisition of the central kspace lines of TAPIR and QUTE. In practice this effect has been observed to be small. It is important to consider the explicit α eff dependence of CT1 : CT 1 (αeff ) =

1 − cos(αeff )e−TReff IT1 1 − e−TReffIT1

,

[3]

42

Shah, Ermer, and Oros-Peusquens

since disregarding this dependence results in a systematic error of the flip angle measurement. The effective repetition time, TReff , of the QUTE sequence is defined as TReff = TR × NSlices , given that the readout interleaves slices and lines of k-space. Here, TR is the time between two consecutive RF pulses, and NSlices is the number of slices. Equation [2] can be solved numerically to obtain the effective flip angle, α eff , at each point within the subject. The corresponding correction factor for the B1 field inhomogeneity is given by CB1 = 1/sin(γ ·α eff ) with γ = 1 for QUTE and γ = α TAPIR /α QUTE for the TAPIR sequence. For measurements performed at 1.5 T and using the body coil for transmit and a circularly polarised birdcage head coil (standard Siemens head coil) for signal reception, the dependence of the B1+ and B1− fields on volunteer characteristics was found to be reduced. The calibration of the two fields can be, therefore, performed on phantoms with great precision. This is not a stringent requirement of the method but more a bonus of using an RF regime which is not strongly influenced by dielectric effects. The inhomogeneity in the receiver profile of, for example, the head coil used for signal reception can be corrected using a large cylindrical phantom with a diameter of 26 cm and a length of 32.6 cm. The phantom should be filled with vegetable oil to prevent significant standing wave effects which otherwise would distort the measurement of absolute signal intensity. Repeating the QUTE measurement with two different flip angles is necessary to prevent mixing of transmitter and receiver inhomogeneities as the measured signal intensity S0,T2 ∗ is determined by both receiver profile inhomogeneity and flip angle miscalibration. The signal decay can be measured twice with flip angles α = 45◦ and α = 90◦ , respectively, using QUTE with, for example, the following measurement parameters: TR = 10 s; TE = 4 ms; α = 90◦ /45◦ ; 45 slices; 64 time points; D = 2 ms; matrix size = 2562 ; FOV = 256 mm; slice thickness = 5 mm; 10 averages. Due to the long repetition time, T1 saturation effects can be neglected. S0,T2 ∗ (α = 45◦ ) and S0,T2 ∗ (α = 90◦ ) can be extracted from the signal time course by the procedure described above. The effective flip angle, α eff (assuming a nominal flip angle of α QUTE = 90◦ ), at each position within the phantom is then given by   S0,T 2∗ (αQUTE = 90◦ ) , [4] αeff = 2 arccos 2S0,T 2∗ (αQUTE = 45◦ ) where S0,T2 ∗ (α = 90◦ ) = k × sin(α eff ) and S0,T2 ∗ (α = 45◦ ) = k × sin(α eff /2) with a constant k which is the same for both acquisitions and therefore cancels out. Equation [4] results in a correction factor for the receiver coil inhomogeneity of CRC = sin(α eff )/S0,T2 ∗ (α = 90◦ ).

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

43

The following parameters have been used in our group for several studies. For TAPIR: TR = 15 ms; TE1 :TE2 :TE3 = 2.8:5.1:7.4 ms; α = 90◦ ; NS = 17 slices; matrix size = 2562 ; FOV = 220 mm; slice thickness = 5 mm; sequential excitation; delay time τ = 2 s. Twenty time points with a spacing of 270 ms (TR ∗NS) tS = {tS1 , tS1 + 270 ms, tS1 + 540 ms, . . . , tS1 + 5,130 ms} sampled on the T1 relaxation curve of slice s. Here, tS1 is the first time point after inversion sampled on the relaxation curve where t11 = 10 ms and tS1 = 10 ms + (s – 1) ∗ 270 ms. The measured data points s(t) can be fitted to the signal intensity of a Look–Locker acquisition scheme s(t) = s(M0 , T1 , α, t) to determine the three unknown parameters (M0 , T1 , α) using a Levenberg–Marquard minimisation. The analytical expression for s(t) and further details about the fitting procedure can be found in Zaitsev et al. (17). In addition, inefficiencies of the inversion pulse can be determined and corrected using the procedure described by Zaitsev et al. (17), with an additional calibration scan called TAPIR-IE (inversion efficiency). TAPIR-IE has the same FOV, slices, TR, TE and BW as TAPIR, but reduced matrix size (64 × 64) and one time point only. For QUTE: TR = 138 ms; TE = 4 ms; α = 90◦ ; 17 slices; 64 time points; D = 2 ms; matrix size = 2562 ; FOV = 220 mm; slice thickness = 5 mm; RF spoiling employed; 10 preparation scans. For both in vivo and phantom measurements, a 10-cm-long cylindrical reference probe with a diameter of 1.5 cm containing Gd-doped water (for example, 60 μl Gd-DTPA for a 50 ml EPI filled with distilled water) is placed in the FOV during image acquisition. The total measurement time for TAPIR, QUTE and the inversion efficiency measurement is approximately 21 min. The temperature of the reference probe is continuously monitored during the whole measurement by a sensor optically linked to a recording device located outside the magnet room (Fotemp; OPTOcon GmbH, Dresden, Germany). Experience shows that the temperature changes observed in the in vivo applications (5) vary considerably between experimental conditions and volunteers, ranging from a virtually constant T(reference) throughout the whole experiment to situations where the temperature increased by up to 4.5–5◦ C. The large variations are usually observed for the first volunteer of the day when the probe is at room temperature. In a series of scans with less than 30 min time between volunteers, the temperature of the probe remains practically constant. 2.2.2. Representative Results

In vivo experiments have been reported on patients with different grades of hepatic encephalopathy (HE) (14) and on several young and healthy volunteers using the method described (4, 5). These authors studied the average white and grey matter water

44

Shah, Ermer, and Oros-Peusquens

Fig. 3.3. Water map of the multi-compartment phantom consisting of tubes with different H2 O/D2 O mixing ratios (3).

contents to investigate the typical physiological changes of cerebral water content, as well as the reproducibility and stability of the presented method in vivo. Results from quantitative water content mapping are presented in Fig. 3.3. This shows the reconstructed quantitative water map of the multi-compartment H2 O/D2 O phantom located at the scanner isocentre. A clear contrast between the tubes is visible and is determined only by differences in H2 O content, as opposed to total water (H2 O + D2 O) content. The comparison between the known H2 O content and the MRmeasured H2 O content is shown in Fig. 3.4. The values reconstructed with the method described here are consistent within one standard deviation with the known quantities of H2 O in the compartments. The mean systematic deviation from the theoretical values is calculated as σsys = 0.5%. The statistical measurement error ranges between 1.1 and 2.1% with a mean of σstat = 1.8%. This results in a total within-slice measurement error of 2 + σ 2 ) = 1.87% σinslice = sqrt(σstat sys The results for WMR if corrections for flip angle miscalibration CB1 and receiver profile inhomogeneity CRC are ignored are shown in Fig. 3.4. Ignoring B1 field changes results in an absolute systematic error between –0.3 and +4.3% with a mean of +1.7%. The errors are worse if the receiver profile is not considered properly. The systematic overestimation of the absolute water content in this case ranges between +2.1 and +8.3% with a mean error of +5.2%. This can be explained by the fact that the reference probe was placed above the phantom in a region where the head coil has a low reception sensitivity. The low sensitivity translates directly

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

45

Fig. 3.4. Known phantom water content versus WMR with all corrections applied (red rectangles), no correction for receiver coil inhomogeneity (green line) and no correction for flip angle miscalibration (orange line). The solid red line represents the 45◦ line of identity where the measurement equals the known water content (3).

into an apparently lower reference signal intensity and therefore an overestimation of the water content. Changes of the water content depending on the slice position are shown in Fig. 3.5. The mean deviation of the absolute water content from its value measured for the slice at isocentre is σ slice = 0.17%. This systematic error can be neglected in comparison to the much larger statistical error component of 1.8%.

Fig. 3.5. Dependence of water content measurement with respect to slice position from the magnet isocentre (3).

46

Shah, Ermer, and Oros-Peusquens

Fig. 3.6. Single slice of the quantitative water content map of a healthy volunteer. The values of the water content of result in WMR (WM) = (70.7 ± 2.1)% for white and WMR (GM) = (80.3 ± 2.9)% for grey matter, respectively (4).

The distribution of the absolute water content in a representative slice through the brain of a healthy 23-year-old male volunteer is shown in Fig. 3.6. The quantification of the difference between both tissues results in an estimate of WMR (WM) = (70.7 ± 2.1)% for white and WMR (GM) = (80.3 ± 2.9)% for grey matter water content, respectively. A quantitative water content map from a randomly chosen patient with grade HE-II is shown in Fig. 3.7. HE-II denotes the highest grade of disease severity investigated in the study (14). The contrast between white and grey matter, as well as CSF, is clearly visible as a result of the differences in water content. Compared to a healthy volunteer, the water content distribution is remarkably different in this patient as exemplified in Fig. 3.7. The average water content in white matter in this patient is + 2.1% higher; this increase is best visible within the basal ganglia and in white matter. In general, a significant increase of water content with increasing HE severity grade can be reported. 2.3. Water Mapping: The 2-Point Method

In order to establish water mapping as a clinical tool that can be widely and independently used, another imaging protocol was developed as described in this section. With the new method, it is possible to map the water content of the whole brain with a resolution of 1 × 1 × 2 mm3 (slice gap 50%) within 10 min. This acceleration makes it more useable for a longitudinal disease monitoring. Furthermore, the protocol uses MR sequences which are

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

47

Fig. 3.7. Quantitative water content maps from a randomly chosen patient with grade HE-II. The colour bar shown on the right side of each map relates the figure colour to the corresponding water content in each image voxel (14).

provided by most manufacturers and can be implemented without any programming efforts. 2.3.1. Methods

The 2-point method uses the QUTE sequence to acquire a series of T2∗ -weighted images at different TEs. This series of images is then extrapolated back to zero echo time by exponentially fitting the measured signal for each voxel to determine the tissue equilibrium magnetisation, S0,T2 ∗ (tissue) (see Section 2.2.1). As already mentioned, the measured signal intensity is influenced by several factors, such as saturation effects or B1 field inhomogeneities and needs to be corrected in order to allow for an accurate quantification of the absolute water content. Temperature differences between brain tissue and the reference probe are corrected for this method in the same way as described before (Section 2.2.1). In order to accelerate the acquisition, a different way of determining the rest of the correction factors has been chosen for the 2-point method. Figure 3.8 depicts the exact processing chain of quantitative water mapping using the 2-point method. To accurately measure MR parameters, such as T1 and tissue water content, it is necessary to exactly map the B1 field, as inhomogeneities in the B1 field result in a spatial variation of the measured signal intensity. Thus, it is necessary to determine the exact FA for each voxel. This is performed by acquiring two EPI images with different nominal flip angles (EPIHEAD (30◦ ) and EPIHEAD (90◦ ), respectively) where the relative signal intensity of both measurements is a function of the effective flip angle (21). However, especially at higher magnetic field strength, it becomes necessary to correct for geometric distortions in the underlying

48

Shah, Ermer, and Oros-Peusquens

Fig. 3.8. The processing chain of quantitative water mapping. Five MR measurements are performed (QUTE-S0,T2 , QUTE-T1 , EPI(30◦ ), EPIHEAD (90◦ ), EPIBODY (90◦ )) in order to determine S0,T2 ∗ and all the required correction factors such as CT1 , CB1 , CRC and CT (adapted from Neeb 2008).

EPI images; this can be performed using the approach proposed by Jezzard et al. (45). Receiver coil inhomogeneities can be calculated by acquiring a third EPI image (EPIBODY (90◦ )) using the body coil for transmission and reception and relating this measurement to the corresponding EPI dataset (EPIHEAD (90◦ )) acquired with the head coil for signal reception (7). This correction is performed for each subject individually (6). Even though the effective TR is relatively long in QUTE, the longitudinal relaxation does not fully recover within one TR cycle. Therefore, the resulting saturation effects need to be corrected as they alter the measured signal intensity. One possible solution for T1 saturation is the acquisition of two spoiled gradient echo images with different TRs and/or different flip angles; this allows the determination of T1 given that the effective flip angle is known (20, 21). By simply performing a second FLASH measurement with a different TR and a different nominal flip angle with respect to the first QUTE measurement (which was used to determine the equilibrium magnetisation), T1 can be mapped as the effective excitation flip angle is known (correction factor CB1 ). Furthermore, instead of a simple FLASH measurement, QUTE with its capability of acquiring multiple time points can be used in order to reconstruct multiple T1 maps, one for each echo time. This approach allows for a reduction of the random noise component in the resulting T1 maps which would otherwise propagate into the water content measurement. The water mapping protocol consists of two measurements with QUTE, as well as EPI acquisitions for the determination

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

49

of the excitation flip angle, and the receiver coil sensitivity was acquired first. The T2∗ decay data for the extrapolation of S0,T2∗ are acquired with the QUTE sequence (QUTE-S0,T2 ∗ ) with the following parameters: TR = 42 ms; α = 40◦ ; 10 time points with a spacing of 3.61 ms. A second QUTE acquisition (QUTE-T1 ) required for T1 mapping uses the following parameters: TR = 14 ms; α = 70◦ ; two time points with a spacing of 3.61 ms. In both cases 50 slices with 2 mm thickness and 1 mm gap, an FoV of 192 × 256 mm2 and a matrix size of 192 × 256 are acquired. In addition, the following parameters should be the same for QUTE-S0,T2 ∗ and QUTE-T1 : TE = 4.8 ms; BW = 310 Hz/Px; RF spoiling employed and 10 preparation scans to reach a steady state. The data required for the flip angle reconstruction can be obtained by the acquisition of two single-shot EPI datasets with nominal excitation flip angle of 90◦ (EPIHEAD (90◦ )) and 30◦ (EPIHEAD (30◦ )), respectively. In both cases, the same slices with the same field of view as for the two QUTE measurements are acquired. The matrix size is 48 × 64, the BW = 2,895 Hz/Px and TE is 20 ms. All four datasets are acquired with the body coil used for signal transmission and the four-channel-phasedarray head coil for signal reception. Finally, the EPIHEAD (90◦ ) measurement is repeated with the same parameters, but using the body coil for signal transmission and reception, resulting in the acquisition of the EPIBODY (90◦ ) dataset. Special care needs to be taken in all acquisitions to choose the same FFT scale factors and the same receiver gain. The total measurement time for the whole protocol described above is 9 min 22 s. 2.3.2. Representative Results

Following phantom validation, the method was applied in a group of 10 healthy young volunteers (6 male and 4 female subjects; mean age 30.8 years; range 24–37 years) to obtain quantitative water content maps in vivo. The measured water content was then interactively determined for regions of interest in, i.e. the splenium of the corpus callosum and the head of the caudate nucleus. The average water content from all volunteers as well as the corresponding average standard deviations were determined for each region. Additionally, the water content of one patient suffering from relapsing remitting multiple sclerosis (RRMS) was measured in order to show the indication of the usefulness of the method for the investigation of brain pathology. The reconstructed water content matches well the known water content of the individual tubes as demonstrated in Fig. 3.9. The average systematic deviation of WMR from its real value is 0.64%. In addition, the water content can be reliably measured over the whole size of the phantom as demonstrated by the horizontal profile through the centre of the phantom shown in Fig. 3.10.

50

Shah, Ermer, and Oros-Peusquens

Fig. 3.9. Comparison of the MR-measured water content with the known water content in tubes filled with different H2 O/D2 O mixing ratios between 50 and 100%. The line shows the value where the measured water content perfectly agrees with the known phantom H2 O content (6).

Fig. 3.10. Slice dependence of the measured water content of the eight tubes filled with different H2 O/D2 O mixing ratios (Neeb 2008).

Figure 3.11 shows a quantitative water content map for one representative transverse slice through the brain of a healthy volunteer. The contrast between grey and white matter, as well as CSF, is clearly visible in the map. This is also obvious from the results of the regional analysis of cerebral water content: the water content ranges between 66.2% in the splenium of the corpus callosum and 85.8% in the head of the caudate nucleus (6). In addition, the standard deviation of the measured water content in vivo is consistent predictions and ranges between 1.7 and 2.6%.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

51

Fig. 3.11. Representative slice of a quantitative water content map of a healthy volunteer. The colour bar shown on the right side of each map relates the figure colour to the corresponding water content in each image voxel (Neeb 2008).

Figure 3.12 shows a slice of the water content map of an MS patient. Lesions in WM are clearly visible (see red arrows for four selected lesions). The slice chosen here is representative for four large lesions in the upper WM. Compared to the water content

Fig. 3.12. One slice of the water content map of a patient suffering from MS is presented. As shown by the red arrows, lesions prominent in WM are easily visible due to the higher water content compared to the surrounding WM.

52

Shah, Ermer, and Oros-Peusquens

in WM of healthy volunteers, the water content in the lesions is between 11 and 18% higher.

3. Myelin Water We describe in the following the method used by the group at the University of British Columbia of Alex MacKay and collaborators. 3.1. Methods

The first step involves the acquisition of T2 decay curves. Accurate quantification of multi-component relaxation requires data with high SNR (~500) and spanning a wide range (at least 32) of echo times from well less than the shortest T2 time of the system to greater than five times the longest anticipated T2 (46). The most common approach is to collect multiple echoes in a single MR sequence. Many investigators have used the Poon–Henkelman multiple spin-echo imaging sequence (47). Key issues for accurate in vivo T2 decay measurement are maintaining perfect 180◦ pulses in the presence of inhomogeneous B1 and B0 fields, and elimination of all contributions from stimulated echoes accruing from signal excited outside the selected slice. For quantitative analysis, T2 decay curves must have high signal-to-noise ratios with the minimum acceptable noise standard deviation being about 1% of the signal strength at the shortest echo time (46). The echo spacing should be as short as feasible, and the echo train length should be such that the last echoes report only noise. For in vivo human brain studies, the echo spacing should be 10 ms or less, and the echo train should exceed 1 s to measure the shortest T2 components and to be sensitive to T2 times on the 400 ms timescale, respectively. Unfortunately, the number of echoes acquired is often limited by considerations of power deposition and MR scanner pulse programmer restrictions. In the presence of microscopic magnetic field gradients, such as those caused by ferromagnetic or paramagnetic constituents, the shape of the T2 decay curve depends upon the echo spacing (48). Most of the published in vivo T2 decay curve results have been acquired from 32 echoes with a 10 ms echo spacing (25, 28, 31). More recently, the protocol involved collecting 32 echoes at 10 ms spacing plus a further 16 echoes with an echo spacing of 50 ms (49). This protocol was used in more recent studies where an intermediate T2 component has been investigated (50, 51) and consists of a 48-echo modified multi-echo sequence with variable TR. A 90◦ slice-selective pulse is followed by 48 rectangular composite 180◦ pulses flanked by slice-selective crusher gradient pulses for elimination of signal from outside the slice (TR 2,120–3,800 ms,

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

53

echo spacing for first 32 echoes is 10 ms, and for the last 16 echoes is 50 ms, 5 mm thick, four averages, matrix size 128 × 128 or 256 ×128). The TR was 3,800 ms for the k = 0 line and was decreased linearly to 2,120 ms at the highest positive and negative k lines. For the results described below, from a multi-echo T2 relaxation measurement, a single axial slice was acquired in two patients (PKU and MS) and for healthy subjects. The acquisition time was 25.4 min for the sequence with variable TR (51). This is an important time saving from the 32.4 min acquisition time for the 48-echo constant TR = 3,800 ms. While this sequence is proven to yield robust T2 decay curves, it suffers from the disadvantage of being a single slice technique. A more practical approach to collecting T2 decay curves was recently applied by Oh et al. (34), who employed a novel spiral acquisition technique (52) and (46) which collected images at 12 echo times (TE) for 16 slices in 10 min. Further, Mädler et al. (53) developed a 3D multi-echo pulse sequence which is capable of collecting 32 echoes from seven slices in less than 20 min. 3.2. Data Analysis

While there are several different methods available to analyse in vivo multi-exponential decay data (54–57) the method most frequently applied in the literature is the NNLS algorithm (30, 46, 58–61). The NNLS approach, which uses a χ 2 minimisation algorithm to fit the decay curve with a T2 distribution (i.e. a plot of amplitude versus T2 time), assumes a large number of T2 times and solves for the corresponding zero or non-zero amplitudes; it thereby does not involve a priori assumptions as to the number of contributing exponentials. From the T2 distribution, the MWF can be determined as the fractional signal with T2 between 15 and 40 ms (61). Different white and grey matter structures decay at dissimilar rates. Furthermore, different brain pathologies have been observed to give rise to unique decay curves. The semilogarithmic T2 plots are not well fit by a single straight line indicating that the volumes of interest contain more than one distinguishable water reservoir, each with its own T2 time. Fitting such curves to extract meaningful information in the form of decay constants and corresponding signal amplitudes becomes increasingly difficult in the presence of multiple compartments and noise. Since the number of distinguishable water environments is generally not known a priori, the fitting algorithm must be capable of estimating the number of exponentials. Furthermore, brain T2 decay curves might be best characterised not by a fixed number of discrete T2 components but rather by a continuous T2 distribution (62). The regularised non-negative least squares algorithm produces robust analyses of multi-component T2 decay curves in terms of a smooth T2 distribution (61, 63). Non-negativity

54

Shah, Ermer, and Oros-Peusquens

is a very reasonable physical constraint since there cannot be a negative number of protons contributing to the MR signal. The following equation is used to describe the T2 decay curve, y(ti ): y(ti ) =

M 

s(T2j )e−ti /T2j + εi , i = 1, 2, · · · , N ,

j=1

where ti are the N measurement times, T2j are M logarithmically spaced T2 times constituting the T2 partition, s(T2j ) is the T2 distribution and εi is the noise contributing to the ith decay curve point. The regularised NNLS algorithm fits a function defined by Eq. 5 to the measured data by minimising , where is given by 2

=χ +μ

M 

s(T2j )2 , μ ≥ 0.

j=1

The first term in , χ 2 , is chi-squared, and the second term is the regulariser which, in this case, is the energy of the T2 distribution. The larger the μ parameter, the more the algorithm smoothes the T2 distribution at the cost of χ 2 misfit: the case of μ = 0 provides the χ 2 min result (63) for which s(T2j ) consists of discrete T2 spikes. One can regularise the fitting process by placing a constraint on the χ 2 value in accordance to χ 2 min . Regularisation results in more robust fits in the presence of noise (63) and provides smooth solutions that better represent the distribution of relaxation times expected from tissue microstructure (62). Using NNLS, several methods can be employed to determine MWF for a defined region in an image. A common approach (the “ROI” method) is to outline a region of interest (ROI) on an image from one echo of the T2 relaxation data and then calculate a single decay curve using the average of the signal intensity from all the voxels within this ROI for each echo. This average ROI decay curve is then inverted to a single T2 distribution using NNLS. A second approach is to carry out the NNLS inversion using voxel-based analysis (VBA). In this method (labelled the “VBA” method), the MWF is first determined for every voxel in the image in order to create a “myelin water map”. ROIs are then applied to the myelin water map, and the mean or median MWF of contributing voxels within the ROI is calculated. If NNLS was a linear inversion technique, MWF values from the ROI and VBA analysis approaches defined above would be identical for the same region; however, NNLS is a complex algorithm that produces T2 distributions which are sensitive to noise. Since the two approaches sample the noise differently, the resulting MWF values are expected to differ slightly.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

55

3.3. Representative Results 3.3.1. Normal Subjects

T2 decay curve measurements demonstrate that individual brain structures have markedly different MWFs and different geometric mean T2 s for the intra/extracellular T2 peak (28). In the central nervous system tissue, T2 behaviour most definitely cannot be described by a single time. Average IE water geometric mean T2 (mean on a logarithmic scale), water content and MWF for different white and grey matter structures were investigated in normal subjects for the first time more than a decade ago (28). The geometric mean T2 times for the intra/extracellular water pool varied between 70 and 90 ms with no separation between grey and white matter structures. Water content is found to be higher for grey matter than for white matter, in agreement with the previous literature (64). Myelin water fraction varies by more than a factor of 2 between different brain white matter structures and is quite low in all grey matter areas. It is believed that the variations in these three parameters between different brain regions arise from differing tissue microstructures. More recently (50), a long T2 component (200– 800 ms) was observed in patients affected by demyelination diseases, but not observed in any healthy volunteers except for the small focal regions involving the posterior internal capsules.

3.3.2. Pathological Changes (MS)

Multiple sclerosis is an autoimmune disease characterised by oedema, inflammation, demyelination and axonal loss within the central nervous system (65). Over the past two decades, magnetic resonance imaging has become a powerful diagnostic tool for MS. Clinical MR images of MS brain and spinal cord can easily visualise lesions produced by the disease. Lesions can appear and disappear, although they typically increase in number and volume as the disease progresses. Unfortunately, little correlation has been found between total lesion volume, which can easily be characterised by MRI, and measures of clinical disability (66). This dissociation between total lesion volume and clinical disability suggests that there is ongoing active disease. Specifically, that is, in areas that appear either unchanged from month to month, such as visually stable lesions, or normal, such as normal-appearing white matter (NAWM), that is not seen on conventional MRI, but is contributing to disease progression. Work examining 33 subjects with MS compared to 18 controls has shown that, on average, MS lesions had 7.7% higher total water content and 50% less MWF than NAWM (67). Wide heterogeneity in lesion MWF was observed, including some lesions that exhibited a total loss of myelin water, suggesting complete destruction of myelin (67). It has also been shown that NAWM

56

Shah, Ermer, and Oros-Peusquens

total water content was, on average, 1.8% greater than that of NWM and that the MWF was 15% lower in NAWM than in NWM (29). It has been suggested, therefore, that MS affects more of the brain than just the bright lesions visible on conventional images. A simple model for water in brain demonstrates that oedema alone cannot account for the observed changes (67), suggesting that NAWM has undergone some loss of myelin, a finding which is supported by pathological studies (68, 69). Recently, a component with a markedly prolonged T2 (200–800 ms) was identified in lesions and NAWM in 12 out of 15 subjects with phenylketonuria (PKU) and in 10 out of 20 subjects with multiple sclerosis (MS) (50). The properties of the long T2 component differ among pathologies, as well as between pathologies and the normal brain. Its origin is not yet understood.

4. Discussion and Conclusions Two fast and reliable strategies to quantitatively measure the absolute water content of the brain have been described. To allow for an accurate quantification of the water content several factors that influence the measured signal intensity need to be corrected. The results presented for the first method, called “TAPIRQUTE” demonstrate that the high-precision measurement of water content using MR is feasible as long as all relevant corrections are considered. From comparison with phantom data, the error of the H2 O measurement is <2% for a single voxel including statistical and systematic variations. As the signal from the reference probe filled with 100% water was determined for each slice individually, the resulting absolute systematic error arising from changing slice positions is σ slice = 0.17% and can therefore be neglected. The precise determination of T1 and of temperature differences between reference probe and tissue, the correction for B1 field inhomogeneities and the measurement of receiver coil imperfections are vital for a reduction of the systematic error component. The methods presented correct for all those effects. Due to the non-linearity of the underlying equations, the final water content estimate is relatively immune to changes in S0,T2 ∗ /S0,T1 which also explains the very good agreement between known and measured water content in the phantom experiment even though slice profile imperfections were not corrected in the work presented by Neeb et al. (6) Several different approaches for the quantitative mapping of water content based on MRI have been published in the literature. A surrogate measurement of water content is based on the measurement of a parameter which appears well correlated to the

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

57

water content, such as T1 (70–73). This, however, is not really a quantitative measure of water content per se, and the precise correlation between T1 and [H2 O] seems, from our own results, to be quite pronouncedly region dependent. This fact is not only important in pathological conditions but also for the study of healthy brain tissue. On the other hand, several authors have published methods to determine M0 directly and calculate the water content by relating M0 in tissue to the corresponding value in a reference phantom (28, 42, 74–76). The first approach presented here provides not only an accurate measure of the absolute localised quantitative water content in vivo but also highly precise T1 and T2∗ information. In principle, it is possible to omit the T1 measurement with TAPIR by increasing TR and/or reducing the flip angle of the QUTE sequence in a separate measurement. However, the acquisition of three quantitative maps – for WMR , T1 and T2∗ – with high precision makes the TAPIR-QUTE method favourable for image segmentation tasks. Finally, no additional measurement of the effective local flip angle is necessary because this quantity can be extracted from the ratio of S0,T2 ∗ and the extrapolation of the T1 relaxation curve to zero inversion time. No separation between the signals arising from water and fat protons was made for the in vivo measurement presented here. However, protons from both compartments contribute to the measured value of S0,T2 ∗ . The contribution from protons bound in proteins or macromolecules can be neglected, however, due to their shortT2∗ . Disregarding the contribution of fat protons to S0,T2 ∗ results in a systematic overestimation of water content, especially in regions with a significant fat content. Due to the high resolution of the method and reasonable band width (no chemical shift artefacts), contamination to brain tissue from fat signal surrounding the skull can be excluded. With a total measurement time of approximately 22 min including temperature determination and a precision of >98%, the results presented clearly demonstrate that quantitative measurement of water content in vivo is feasible, but needs to be improved and accelerated in order to establish it as a routine clinical application. Faster imaging variants of the TAPIR and QUTE sequences are currently being developed and evaluated. One possible speed-up of the method is offered by the use of parallel imaging. The SNR loss must, however, be compensated by higher signal, as obtained, for example, at a higher field strength; otherwise the accuracy and precision of the method will suffer. First results in HE patients have been demonstrated and emphasise the high potential of the proposed method for careful and longitudinal monitoring of neurological diseases accompanied by a rather small change in the localised water content in the brain.

58

Shah, Ermer, and Oros-Peusquens

The results presented in Section 2.3 for the 2-point method demonstrate that a precise and fast measurement of the cerebral water content in clinically relevant measurement times is possible even on a 1.5 T clinical system. Again, this requires that all error sources which affect the measured MR signal intensity are properly considered and corrected in order to obtain an accurate measure of tissue water content. The individual sampling of the reference signal intensity for each slice is more important to constrain systematic errors resulting from a potential bias in the flip angle measurement. While the quantitative water content is extremely sensitive to errors of the flip angle, the flip angle bias effect largely cancels out if all voxels in a given slice are biased by the same amount. Such a global, slicedependent bias may result from different slice profiles of the two underlying EPI acquisitions with nominal flip angles of 30◦ –90◦ as well as other non-linear effects. As the measured effective flip angle depends on the ratio of the two EPI measurements, such imperfections therefore result in a global bias of the flip angle determination (6). The measurement precision, i.e. the component of the random fluctuating error, has been optimised here in order to achieve the best average precision for human brain imaging with full brain coverage in an acquisition time below 10 min. The random error for all brain tissues (grey and white matter, CSF) is virtually constant (approximately 2–3%) as part of the increased error due to the longer T1 in CSF is recovered by the accompanying decrease in the error associated with the increased T2 and T2∗ in CSF. In general, the highest precision can be achieved for compartments with short T1 and long T2∗ consistent with the requirement T2∗ ≤ T1 . However, T2∗ relaxation times at 1.5 T in the human brain are typically larger than 20–40 ms and stay above that range for most of the voxels imaged. If one is primarily interested in the study of water content in regions with reduced T2∗ (e.g. regions with a significant deposition of endo- or exogenous paramagnetic contrast agents), the measurement protocol can easily be adapted to also enable precise measurement in this regime. In general, the precision was specifically tailored here to perform imaging of a healthy human brain. However, the analysis is straightforward and can easily be repeated using different constraints and boundary conditions. Moreover, the simulated precision matches well with the values determined experimentally, both in phantom and in vivo experiments (6). In summary, both the TAPIR-QUTE and the 2-point methods enable a precise and accurate mapping of tissue water content. Systematic error sources were carefully evaluated, and the precision was optimised for human brain imaging at 1.5 T, although the described framework offers sufficient flexibility to optimise the experimental protocol for different applications or different

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

59

constraints. In addition to generating highly accurate water maps, the TAPIR T1 mapping of the first method offers an excellent time resolution on the inversion recovery curve, which can be used for stunning visualisation of anatomical details, identification of voxels with partial volume effects or multi-exponential fitting. Highly accurate T1 maps with high temporal resolution on the inversion recovery curves are an essential ingredient of multiplefield comparisons on the same volunteers (77, 78). The second water mapping method is characterised by a measurement time of below 10 min, precise and accurate water content maps with a resolution of 1 × 1 × 2 mm2 and full brain coverage (with 50% slice gaps). Both methods can be used at 1.5 T without using parallel imaging techniques. The use of quantitative, whole-brain water maps produced using the methodology described herein will facilitate the development of quantitative voxel-based morphometry (qVBM). Future work will concentrate on the production of relevant templates to allow the interrogation of these quantitative water maps on a voxel-by-voxel basis using the apparatus of VBM. Here, the development and use of the combined parameter atlases of H2 O, T1 andT2∗ could lead to a better knowledge of the different correlations in a healthy population and could finally lead to a more precise diagnosis of different diseases related to a pathological change in one or more of these parameters. T2 relaxation has the potential to quantitatively define the role of myelin-specific pathology and further the understanding of demyelinating disease pathogenesis. Histological validation is, however, crucial: if myelin water is indeed a marker for myelin it must correlate quantitatively with histopathological measures of myelin. It has been shown in studies on guinea pig (60, 79) and rat (30) models that white matter T2 distributions contain multiple components and that the short T2 component was correlated with histological measures of myelin. The T2 distribution from formalin-fixed human brain is, fortuitously, shaped similar to that from brain in vivo, albeit with T2 shifted to lower times. This is perhaps not surprising since water T2 times are strongly influenced by cellular structure (80) which is relatively unchanged by fixation. The qualitative correspondence between MWF and the anatomical distribution of myelin, as demarked by the luxol fast blue stain, has been shown in histopathological studies (29). A strong correlation was found in (31) between MWF and blue optical density in fixed brain slices stained by the luxol fast blue stain for myelin. These studies provide strong evidence that the MWF is, indeed, a marker for myelin in white matter. In addition to the three components assigned to myelin water, intra/extracellular water and CSF, a long T2 signal in the centre of the posterior internal capsules of healthy volunteers was

60

Shah, Ermer, and Oros-Peusquens

reported (28, 50). It was possible to separate this component from the one due to CSF by lengthening the sampling of the T2 decay curve by a factor of 3.5. The region exhibiting a long T2 signal corresponds to the centre of the corticospinal tract, and it is hypothesised that the long T2 signal arises from compartmentalisation of intra- and extracellular water due to thicker myelin sheaths in this area, which therefore reduces exchange between intra- and extracellular water (81). The situation is more complicated in lesions (especially in MS) where the main tissue water peak is pushed to higher T2 due to pathology. Then NNLS cannot distinguish between two water pools in exchange giving rise to a single T2 peak and two water environments with similar, but not identical, T2 times. Potential sources for the long T2 component detected in MS and PKU patients include vacuolation or increases in the amount of extracellular water or reduced exchange between intra- and extracellular water. Once the T2 distribution in the normal and diseased populations is fully characterised, a very appealing possibility of imaging of the different water components arises, which requires a much reduced measurement time and/or full brain coverage. Imaging of short T2 spins can be conceptualised as a filter design problem. A linear combination of magnitude MR images taken at different echo times is a function of T2 . By properly choosing the weight of each image in the linear combination, one can filter out specific T2 species without specifically estimating their T2 . This method was applied to imaging of phantoms with three different T2 values more than two decades ago by Brosnan et al. (82). The linear combination weights could be chosen such that all phantoms with a given T2 were filtered out (82). Jones et al. (83) used a linear combination filter to find the fraction of myelin water in vivo. The SNR efficiency of linear combination filters and NNLS were compared, and the performance was found to be similar, but the linear combination filters required a fraction of the computation and measurement time (83). While the CPMG sequence used by Jones et al. (83) acquires all 32 echoes in a single repetition and single slice in a measurement time of the order of 30 min, a 6-slice, 3-echo linear combination filter to image myelin was presented by Vidarsson et al. (37). By optimising the echo times for maximum SNR efficiency, the required data can be acquired using a spin-echo sequence in 5 min of scan time. SNR efficiency of the 3-echo filter was found to be roughly the same as 4-, 5-, and 32echo filters. The 3-echo filter was validated using phantoms with known T1 and T2 and in vivo in both a healthy volunteer and an MS patient (37). In conclusion, a method for the investigation of myelin water has been developed and refined over the course of 11/2 decades. This period has allowed for several aspects in the data acquisition and processing to be investigated and optimised and con-

Measuring the Absolute Water Content of the Brain Using Quantitative MRI

61

structive criticism from the community to be incorporated, resulting in a mature method which can be applied to large groups of volunteers and patients. The current understanding of the distribution of myelin water in the normal brain is based on a large number of data points and facilitates the detection of pathology through comparison. Several patient groups have been investigated (MS, PKU, schizophrenia and Alzheimer’s). Applications to various other diseases and normal aging are expected to emerge, where the changes not only in the myelin water peak but probably also in the long T2 component need to be understood. On the downside, the method suffers from extremely limited brain coverage (in most cases one single slice has been acquired), and thus regional changes in the water components are still to be investigated. Further effort needs to be dedicated to the development of fast acquisition methods with minimal impact on the precision, accuracy and reproducibility of the method. A measurement time of nearly 30 min is difficult for elderly or severely impaired subjects. A few faster methods are emerging, which in addition provide more complete brain coverage (34, 38, 40). The latter two referenced methods, however, involve either an extensive modelling of the water compartments (40) or a different MR parameter than T2 , i.e. T2∗ (38). Extensive validation is therefore required before the methods can be employed in large cohorts. References 1. Shah, N. J., Zaitsev, M., Steinhoff, S., Zilles, K. A new method for fast multislice T1 mapping. Neuroimage 2001;14(5):1175–1185. 2. Steinhoff, S., Zaitsev, M., Zilles, K., Shah, N. J. Fast T(1) mapping with volume coverage. Magn Reson Med 2001 Jul;46(1):131– 140. 3. Neeb, H., Shah, N. J. Enhancing the precision of quantitative water content mapping by optimizing sequence parameters. Magn Reson Med 2006a;56(1):224–229. 4. Neeb, H., Zilles, K., Shah, N. J. A new method for fast quantitative mapping of absolute water content in vivo. Neuroimage 2006b;31(3):1156–1168. 5. Neeb, H., Zilles, K., Shah, N. J. Fullyautomated detection of cerebral water content changes: Study of age- and genderrelated H2O patterns with quantitative MRI. Neuroimage 2006c;29(3):910–922. 6. Neeb, H., Ermer, V., Stocker, T., Shah, N. J. Fast quantitative mapping of absolute water content with full brain coverage. Neuroimage 2008;42(3):1094–1109.

7. Tofts, P. Quantitative MRI of the Brain: Measuring Changes Caused by Disease. London: Wiley; 2003. 8. Deoni, S. C. L., Rutt, B. K., Peters, T. M. Rapid combined T1 and T2 mapping using gradient recalled acquisition in the steady state. Magn Reson Med 2003;49(3): 515–526. 9. Deoni, S. C. L., Peters, T. M., Rutt, B. K. High-resolution T1 and T2 mapping of the brain in a clinically acceptable time with DESPOT1 and DESPOT2. Magn Reson Med 2005;53(1): 237–241. 10. Deoni, S. C. L., Rutt, B. K., Arun, T., Pierpaoli, C., Jones, D. K. Gleaning multicomponent T1 and T2 information from steady-state imaging data. Magn Reson Med 2008;60(6):1372–1387. 11. Warntjes, J. B. M., Leinhard, O. D., West, J., Lundberg, P. Rapid magnetic resonance quantification on the brain: Optimization for clinical usage. Magn Reson Med 2008;60(2):320–329.

62

Shah, Ermer, and Oros-Peusquens

12. Preibisch, C., Volz, S., Anti, S., Deichmann, R. Exponential excitation pulses for improved water content mapping in the presence of background gradients. Magn Reson Med 2008;60(4):908–916. 13. Mathur-DeVre, R. Biomedical implications of the relaxation behaviour of water related to NMR imaging. Br J Radiol 1984;57(683):955–976. 14. Shah, N. J., Neeb, H., Kircheis, G., Engels, P., Haeussinger, D., Zilles, K. Quantitative cerebral water content mapping in hepatic encephalopathy. Neuroimage 2008;41(3):706–717. 15. Shah, N. J., Zaitsev, M., Steinhoff, S., Wiese, S., Zilles, K. (2000). Development of Sequences for fMRI: Keyhole Imaging and Relaxation Time Mapping. eenc.unileipzig.de/Shah.pdf. 16. Dierkes, T., Neeb, H., Shah, N. (2004). Distortion correction in echo-planar imaging and quantitative T2 ∗ mapping. Proceedings of the international workshop on quantitation in biomedical imaging with PET and MRI. 17. Zaitsev, M., Steinhoff, S., Shah, N. J. Error reduction and parameter optimization of the TAPIR method for fast T1 mapping. Magn Reson Med 2003;49(6):1121–1132. 18. Look, D. C., Locker, D. R. Time saving in measurement of NMR and EPR relaxation times. Rev Sci Instrum 1970;41: 250–251. 19. Deichmann, R. Fast high-resolution T1 mapping of the human brain. Magn Reson Med 2005;54(1):20–27. 20. Tong, C. Y., Prato, F. S. A novel fast T1mapping method. J Magn Reson Imaging 1994;4(5):701–708. 21. Mihara, H., Sekino, M., Iriguchi, N., Ueno, S. (2005). A method for an accurate T1 relaxation-time measurement compensating B1 field inhomogeneity in magneticresonance imaging. 49th Annual Conference on Magnetism and Magnetic Materials, Jacksonville, Florida (USA), AIP. 22. Menon, R. S., Allen, P. S. Application of continuous relaxation time distributions to the fitting of data from model systems and excised tissue. Magn Reson Med 1991;20(2):214–227. 23. Fischer, H. W., Rinck, P. A., Van Haverbeke, Y., Muller, R. N. Nuclear relaxation of human brain gray and white matter: Analysis of field dependence and implications for MRI. Magn Reson Med 1990;16(2):317–334. 24. Stewart, W. A., MacKay, A. L., Whittall, K. P., Moore, G. R., Paty, D. W. Spin-

25.

26. 27.

28.

29.

30.

31.

32.

33.

34.

35.

spin relaxation in experimental allergic encephalomyelitis. Analysis of CPMG data using a non-linear least squares method and linear inverse theory. Magn Reson Med 1993;29(6):767–775. MacKay, A., Whittall, K., Adler, J., Li, D., Paty, D., Graeb, D. In vivo visualization of myelin water in brain by magnetic resonance. Magn Reson Med 1994;31(6):673–677. Brownstein, K. R., Tarr, C. E. Spin-lattice relaxation in a system governed by diffusion. J Magn Reson 1977;26:17–24. Brownstein, K. R., Tarr, C. E. Importance of classical diffusion in NMR studies of water in biological cells. Phys Rev A 1979;19:2446–2453. Whittall, K. P., MacKay, A. L., Graeb, D. A., Nugent, R. A., Li, D. K., Paty, D. W. In vivo measurement of T2 distributions and water contents in normal human brain. Magn Reson Med 1997;37(1):34–43. Moore, G. R. W., Leung, E., MacKay, A. L., Vavasour, I. M., Whittall, K. P., Cover, K. S. et al. A pathology-MRI study of the short-T2 component in formalinfixed multiple sclerosis brain. Neurology 2000;55(10):1506–1510. Webb, S., Munro, C. A., Midha, R., Stanisz, G. J. Is multicomponent T2 a good measure of myelin content in peripheral nerve? Magn Reson Med 2003;49(4):638–645. Laule, C., Leung, E., Lis, D. K., Traboulsee, A. L., Paty, D. W., MacKay, A. L. et al. Myelin water imaging in multiple sclerosis: Quantitative correlations with histopathology. Mult Scler 2006;12(6):747–753. Laule, C., Kozlowski, P., Leung, E., Li, D. K., Mackay, A. L., Moore, G. R. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. Neuroimage 2008;40:1575–1580. Flynn, S. W., Lang, D. J., Mackay, A. L., Goghari, V., Vavasour, I. M., Whittall, K. P. et al. Abnormalities of myelination in schizophrenia detected in vivo with MRI, and post-mortem with analysis of oligodendrocyte proteins. Mol Psychiatry 2003;8(9):811–820. Oh, J., Han, E. T., Pelletier, D., Nelson, S. J. Measurement of in vivo multi-component T2 relaxation times for brain tissue using multislice T2 prep at 1.5 And 3 T. Magn Reson Imaging 2006;24(1):33–43. Sirrs, S. M., Laule, C., Maedler, B., Brief, E. E., Tahir, S. A., Bishop, C. et al. Normal appearing white matter in subjects with phenylketonuria: Water content, myelin water fraction, and metabolite concentrations. Radiology 2007;242(1):236–243.

Measuring the Absolute Water Content of the Brain Using Quantitative MRI 36. Tozer, D. J., Davies, G. R., Altmann, D. R., Miller, D. H., Tofts, P. S. Correlation of apparent myelin measures obtained in multiple sclerosis patients and controls from magnetisation transfer and multicompartmental T2 analysis. Magn Reson Med 2005;53(6):1415–1422. 37. Vidarsson, L., Conolly, S. M., Lim, K. O., Gold, G. E., Pauly, J. M. Echo time optimization for linear combination myelin imaging. Magn Reson Med 2005;53(2):398–407. 38. Wu, Y., Alexander, A. L., Fleming, J. O., Duncan, I. D., Field, A. S. Myelin water fraction in human cervical spinal cord in vivo. J Comput Assist Tomogr 2006;30(2): 304–306. 39. Deoni, S. C., Rutt, B. K., Jones, D. K. Investigating exchange and multicomponent relaxation in fully-balanced steady state free precession imaging. J Magn Reson Imaging 2008;27:1421–1429. 40. Du, Y. P., Chu, R., Hwang, D., Brown, M. S., Kleinschmidt-DeMasters, B. K., Singel, D. et al. Fast multislice mapping of the myelin water fraction using multicompartment analysis of T2∗ decay at 3t: A preliminary postmortem study. Magn Reson Med 2007;58(5):865–870. 41. Yablonskiy, D. A., Haacke, E. M. Theory of NMR signal behavior in magnetically inhomogeneous tissues: The static dephasing regime. Magn Reson Med 1994;32(6):749– 763. 42. Lin, W., Venkatesan, R., Gurleyik, K., He, Y. Y., Powers, W. J., Hsu, C. Y. An absolute measurement of brain water content using magnetic resonance imaging in two focal cerebral ischemic rat models. J Cereb Blood Flow Metab 2000;20(1):37–44. 43. Mansfield, P. Multi-planar image formation using NMR spin echoes. J Phys C Solid State Phys 1977;10(3):L55–L58. 44. Haase, A., Frahm, J., Matthaei, D., Hanicke, W., Merboldt, K. D. FLASH imaging. Rapid NMR imaging using low flip-angle pulses. J Magn Reson 1986;67(2):258–266 (1969). 45. Jezzard, P., Balaban, R. S. Correction for geometric distortion in echo planar images from B0 field variations. Magn Reson Med 1995;34(1):65–73. 46. Graham, S. J., Stanchev, P. L., Bronskill, M. J. Criteria for analysis of multicomponent tissue T2 relaxation data. Magn Reson Med 1996;35:370–378. 47. Poon, C. S., Henkelman, R. M. Practical T2 quantitation for clinical applications. J Magn Reson Imaging 1992;2(5):541–553. 48. Ye, F. Q., Martin, W. R., Allen, P. S. Estimation of brain iron in vivo by means of

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59. 60.

63

the interecho time dependence of image contrast. Magn Reson Med 1996;36(1):153–158. Skinner, M. G., Kolind, S. H., Mackay, A. L. The effect of varying echo spacing within a multiecho acquisition: Better characterization of long T(2) components. Magn Reson Imaging 2007;25(6):834–839. Laule, C., Vavasour, I. M., Mädler, B., Kolind, S. H., Sirrs, S. M., Brief, E. E., Traboulsee, A. L., Moore, G. R., Li, D. K., MacKay, A. L. MR evidence of long T2 water in pathological white matter. J Magn Reson Imaging 2007;26(4):1117–1121. Laule, C., Kolind, S. H., Bjarnason, T. A., Li, D. K., Mackay, A. L. In vivo multiecho T(2) relaxation measurements using variable TR to decrease scan time. Magn Reson Imaging 2007;25:834–839. Foltz, W. D., Al-Kwifi, O., Sussman, M. S., Stainsby, J. A., Wright, G. A. Optimized spiral imaging for measurement of myocardial T2 relaxation. Magn Reson Med 2003;49(6):1089–1097. Mädler, B., MacKay, A. L. 2006, In-vivo 3D Multi-component T2-Relaxation Measurements for Quantitative Myelin Imaging at 3T (2006), Proceedings of the 14-th ISMRM Scientific Meeting and Exhibition 2006, Seattle. USA, p. 2112. Armspach, J. P., Gounot, D., Rumbach, L., Chambron, J. In vivo determination of multiexponential T2 relaxation in the brain of patients with multiple sclerosis. Magn Reson Imaging 1991;9(1):107–113. Papanikolaou, N., Maniatis, V., Pappas, J., Roussakis, A., Efthimiadou, R., Andreou, J. Biexponential T2 relaxation time analysis of the brain: Correlation with magnetisation transfer ratio. Invest Radiol 2002;37(7):363–367. Stanisz, J., Henkelman, R. M. Diffusional anisotropy of T2 components in bovine optic nerve. Magn Reson Med 1998;40(3):405–410. Moody, J. B., Xia, Y. Analysis of multiexponential relaxation data with very short components using linear regularization. J Magn Reson 2004;167(1):36–41. Beaulieu, C., Fenrich, F. R., Allen, P. S. Multicomponent water proton transverse relaxation and T2-discriminated water diffusion in myelinated and nonmyelinated nerve. Magn Reson Imaging 1998;16(10):1201–1210. Fenrich, F. R., Beaulieu, C., Allen, P. S. Relaxation times and microstructures. NMR Biomed 2001;14(2):133–139. Gareau, P. J., Rutt, B. K., Karlik, S. J., Mitchell, J. R. Magnetisation transfer and multicomponent T2 relaxation measure-

64

61. 62.

63. 64. 65. 66.

67.

68.

69.

70.

71.

72.

Shah, Ermer, and Oros-Peusquens ments with histopathologic correlation in an experimental model of MS. J Magn Reson Imaging 2000;11(6):586–595. Whittall, K. P., MacKay, A. L. Quantitative interpretation of NMR relaxation data. J Magn Reson 1989;84:134–152. Kroeker, R. M., Henkelman, R. M. Analysis of biological NMR relaxation data with continuous distributions of relaxation times. J Magn Reson 1986;69:218–235. Lawson, C. L., Hanson, R. J. Solving Least Squares Problems. Englewood Cliffs, NJ: Prentice-Hall; 1974. Tofts, P. S. Quantitative Magnetic Resonance Methods. London: Elsevier; 2004. Keegan, B. M., Noseworthy, J. H. Multiple sclerosis. Annu Rev Med 2002;53: 285–302. Paty, D. W., Li, D. K. Interferon beta-1b is effective in relapsing-remitting multiple sclerosis: II. MRI analysis results of a multicenter, randomized, double-blind, placebocontrolled trial. UBC MS/MRI study group and the IFNB multiple sclerosis study group. Neurology 1993;43(4):662–667. Laule, C., Vavasour, I. M., Moore, G. R. W., Oger, J., Li, D. K. B., Paty, D. W. et al. Water content and myelin water fraction in multiple sclerosis: A T2 relaxation study. J Neurol 2004;251(3):284–293. Allen, I. V., Glover, G., Anderson, R. Abnormalities in the macroscopically normal white matter in cases of mild or spinal multiple sclerosis (MS). Acta Neuropathol Suppl 1981;7:176–178. Itoyama, Y., Sternberger, N. H., Webster, H. D., Quarles, R. H., Cohen, S. R., Richardson, E. P., Jr. Immunocytochemical observations on the distribution of myelinassociated glycoprotein and myelin basic protein in multiple sclerosis lesions. Ann Neurol 1980;7(2):167–177. Estilaei, M., MacKay, A., Whittall, K., Mayo, J. In vitro measurements of water content and T2 relaxation times in lung using a clinical MRI scanner. J Magn Reson Imaging 1999;9(5):699–703. Gideon, P., Rosenbaum, S., Sperling, B., Palle, P. MR-visible brain water content in human acute stroke. Magn Reson Imaging 1999;17(2):301–304. Lüsse, S., Claassen, H., Gehrke, T., Hassenpflug, J., Schünke, M., Heller, M., Glüer, C. C. Evaluation of water content by spatially resolved transverse relaxation times

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

of human articular cartilage. Magn Reson Imaging 2000;18(4):423–430. Andersen, C. In vivo estimation of water content in cerebral white matter of brain tumour patients and normal individuals: Towards a quantitative brain edema definition. Acta Neurochir 1997;139:249–256. Fernandez-Seara, M. A., Song, H. K., Wehrli, F. W. Trabecular bone volume fraction mapping by low-resolution MRI. Magn Reson Med 2001;46(1):103–113. Farace, P., Pontalti, R., Cristoforetti, L., Antolini, R., Scarpa, M. An automated method for mapping human tissue permittivities by MRI in hyperthermia treatment planning. Phys Med Biol 1997;42(11): 2159–2174. Lin, W., Paczynski, R. P., Venkatesan, R., He, Y. Y., Powers, W. J., Hsu, C. Y., Haacke, E. M. Quantitative regional brain water measurement with magnetic resonance imaging in a focal ischemia model. Magn Reson Med 1997;38(2):303–310. Rooney, W. D., Johnson, G., Li, X., Cohen, E. R., Kim, S. G., Ugurbil, K., Springer, C. S., Jr. Magnetic field and tissue dependencies of human brain longitudinal 1H2O relaxation in vivo. Magn Reson Med 2007 Feb;57(2):308–318. Oros-Peusquens, A.-M., Laurila, M., Shah, N. J. Magnetic field dependence of the distribution of NMR relaxation times in the living human brain. Magma 2008 Mar;21(1–2):131–147. Gareau, P. J., Rutt, B. K., Bowen, C. V., Karlik, S. J., Mitchell, J. R. In vivo measurements of multi-component T2 relaxation behaviour in guinea pig brain. Magn Reson Imaging 1999;17(9):1319–1325. Araujo, C., MacKay, A. L., Whittall, K. P., Hailey, J. R. T. A diffusion model for spin– spin relaxation of compartmentalized water in wood. J Magn Reson 1993;101:248–261. Yagishita, A., Nakano, I., Oda, M., Hirano, A. Location of the corticospinal tract in the internal capsule at MR imaging. Radiology 1994;191:455–460. Brosnan, T., Wright, G., Nishimura, D., Cao, Q., Macovski, A., Sommer, F. G. Noise reduction in magnetic resonance imaging. Magn Reson Med 1988;8:394–409. Jones, C. K., Xiang, Q. S., Whittall, K. P., MacKay, A. L. Linear combination of multiecho data: Short T2 component selection. Magn Reson Med 2004;51:495–502.

Chapter 4 Magnetic Resonance Relaxation and Quantitative Measurement in the Brain Sean C.L. Deoni Abstract Underlying the exquisite soft tissue contrast provided by magnetic resonance imaging are the inherent biophysical processes of relaxation. Through the intricate relationships between tissue microstructure and biochemistry and the longitudinal and transverse relaxation rates, quantitative measurement of these relaxation parameters is informative of tissue change associated with disease, neural plasticity, and other biological processes. Quantitative imaging studies can further facilitate more detailed characterizations of tissue, providing a more direct link between modern MR imaging and classic histochemical and histological studies. In this chapter, we briefly review the biophysical basis of relaxation, introducing and focusing specifically on the T1 , T2 , and T2 ∗ relaxation times and detail some of the more widely used and clinically feasible techniques for their in vivo measurement. Methods for analyzing relaxation data are covered, and a summary of significant results from reported neuroimaging studies is provided. Finally, the combination of relaxation time data with other quantitative imaging data, including diffusion tensor and magnetization transfer, is examined, with the aim of providing more thorough characterization of brain tissue. Key words: Longitudinal relaxation, transverse relaxation, relaxometry, quantitative imaging.

1. Introduction The acquisition of T1 and T2 -weighted images, and the relative tissue contrast provided by each, is familiar to the majority of clinicians and imaging scientists. Yet, the underlying biophysical phenomena that ultimately give rise to this contrast, and their relationships to tissue biochemistry and structure, remain poorly understood. In this chapter, the biophysical origins of relaxation and its utility in clinical, diagnostic, and research imaging will M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_4, © Springer Science+Business Media, LLC 2011

65

66

Deoni

be discussed. Following an overview of commonly employed measurement techniques, a survey of significant findings from quantitative relaxation neuroimaging studies will be presented. The chapter will conclude with a discussion of how relaxation time data may be used in complement with other imaging data, for example, diffusion tensor imaging, to provide a more thorough description of tissue. Throughout this chapter, a classical physics description of NMR will be used, as this provides a more intuitive picture of the magnetic resonance and relaxation phenomena. In conventional magnetic resonance imaging, the observed signal is generated principally by the hydrogen nuclei (a.k.a. protons and referred to as spins in MR speak) within the water molecules that comprise more than 75% of brain tissue by weight (1). Though a simplification of the quantum mechanical phenomenon, it is useful to think of proton spins as behaving like tiny bar magnets (Fig. 4.1a). When placed in a large magnetic field, denoted by B0 (such as the MRI scanner), proton spins align parallel (lower energy state) or anti-parallel (higher energy state) to the direction of the external field (Fig. 4.1c). At equilibrium, slightly more protons are aligned in the lower energy parallel direction, resulting in a small but measurable net magnetic vector (moment in MR speak) and denoted by M0 . If tilted away from the direction of B0 , this moment precesses about the external field at a specific frequency (the Larmor frequency, 0 ) equal to the external magnetic field strength (B0 ) multiplied by the proton gyromagnetic ratio (©). At 1.5 T, the Larmor frequency for protons is 1.5 T × 42.58 MHz/T ≈ 64 MHz, just below the bottom end of the FM radio spectrum. In practice, a radio frequency (RF) pulse applied at the Larmor frequency tilts the magnetic moment away from the direction of B0 into the transverse plane (Fig. 4.2). In addition to tilting the longitudinal moment away from B0 , the RF pulse also

Fig. 4.1. a A proton behaves like a tiny bar magnetic, spinning with a defined precessional frequency (the Larmor frequency, ω0 ) and producing a small magnetic moment. In a collection of protons, these tiny moments are randomly aligned (b), however, when placed in a large magnetic field (B0 ), they orient with the field, with slightly more in the parallel direction, producing the net magnetic moment (M0 ) that is the basis for MRI signal (c) and, which, precesses about the direction of B0 at the Larmor frequency (d).

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

67

Fig. 4.2. Illustration of T1 recovery along the longitudinal direction (a). An RF pulse (shown here is an inversion 180◦ pulse) ‘flips’ the parallel-aligned protons to the anti-parallel direction, inverting the net magnetization vector. When removed, the protons return to their equilibrium orientation with rate R1 (= 1/T1 ) and the net magnetization recovers.

causes the individual proton moments to align in orientation (or to become phase coherent), producing a net magnetic moment in the transverse plane. This precessing transverse magnetization is the source of the signal measured in an MR experiment. When the RF pulse is removed, the magnetization recovers back to equilibrium, with the individual protons returning to their original parallel or anti-parallel direction and the individual moments dephasing in the transverse plane. The rate of return of the longitudinal magnetization and the rate of loss of the transverse magnetization are characterized by the T1 , T2 , and T2 ∗ relaxation times. T1 governs the re-growth of the longitudinal magnetization, and T2 and T2 ∗ describe the loss of phase coherence of the transverse magnetization. Intrinsically, the processes of relaxation are driven by molecular motion, interaction, and energy exchange. T1 relaxation involves an exchange of energy between water protons and protons attached to other macromolecules (collectively referred to as the lattice). Hence, T1 is also referred to as the spin–lattice relaxation time. T2 relaxation, in contrast, involves an exchange of energy between the water protons themselves and, accordingly, is termed the spin–spin relaxation time. An important distinction between T1 and T2 relaxation is that while T1 relaxation is an energy-loss process (with energy exchanged to the bulk tissue), T2 relaxation is an energy-conserving process (Fig. 4.3). As T1 and T2 arise from random molecular motion and proton–proton interactions, they are directly influenced by the local biophysical and biochemical environments and, therefore, contain information reflective of these environments, including tissue density (i.e., water content and mobility); macromolecule,

68

Deoni

Fig. 4.3. Illustration of T2 decay in the transverse plane. Following an RF pulse (a saturation 90◦ pulse is used here), the individual proton moments are aligned (in phase) in the transverse plane. Over time, this phase coherence is lost and the net transverse magnetization decays with rate R2 (= 1/T2 ).

protein, and lipid composition; paramagnetic atom (i.e., iron) concentration; as well as other pathologically related characteristics. Thus, T1 and T2 differ for tissues with different composition, and changes in T1 and T2 are indicative of tissue changes associated with disease or other biological processes, including neurodevelopment, learning, and neuroplasticity, as well as aging and neurodegeneration. Despite this seemingly obvious statement, it was not until 1971, some 33 years after the discovery of the MR phenomenon (2), that different tissues where shown to have different relaxation characteristics (3) and later still that T1 and T2 were shown to change in pathology. To exploit these differences to create tissue contrast in conventional T1 - or T2 -weighted images, the acquired signal is made sensitive to tissue differences through the choice of pulse sequence and manipulation of acquisition parameters. For example, the tissue contrast of a spoiled gradient recalled echo (SPGR) image can be made more or less sensitive to T1 differences by varying the repetition time (TR), echo time (TE), flip angle (α), or adding preparation pulses (Fig. 4.4). However, in addition to T1 differences, the T1 -weighted SPGR signal is also influenced by differences in proton spin density (ρ) and T ∗ 2 , as well as factors extraneous to tissue, including field strength, transmit flip angle and receiver coil sensitivity variations and signal amplifier gains. This nonlinear blend of signal sources, coupled with inconsistent sources of hardware corruption, makes the physical interpretation of signal changes challenging and preclude direct comparisons of signal intensity values across subjects, time-points, or imaging centers. The interpretation of imaging data can be simplified by removing or separating these independent signal sources. This

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

69

Fig. 4.4. Tissue contrast in T1 - or T2 -weighted images can be varied through pulse sequence choice, adjusting acquisition timings and parameters, such as the flip angle, or adding preparation pulses, such as an inversion or saturation pulse. Here we demonstrate the affect of changing flip angle and adding a preparatory inversion pulse (IR-) on the T1 -weighted spoiled gradient recalled echo (SPGR) image contrast.

Fig. 4.5. The T1- weighted imaging signal is related to not only intrinsic tissue properties, such as T1 and proton density (ρ), but also extrinsic hardware effects, such as variations in flip angle. Quantitative relaxation time mapping cleanly separates these effects, providing a map of each parameter.

uncoupling can be achieved by calculating the relaxation times of each voxel in the image, generating ‘maps’ of T1 , T2 , etc. as shown in Fig. 4.5. A consequence of separating these complex signal interdependencies, quantitative T1 , T2 , and T2 ∗ imaging can facilitate improved characterizations of tissue change, enhance imaging contrast, and provide a more direct link between observed signal changes and the micro-anatomical alterations distinguished via histochemistry and histology.

2. The Biophysical Origins of Relaxation and Relationship to Tissue Structure

As briefly introduced above, T1 and T2 relaxation are phenomenon driven by fluctuations in the local magnetic fields as a result of proton motion and their physical interactions (4). As the individual water molecules and attached protons undergo translational and rotational motion, they produce and experience oscillating magnetic fields (Fig. 4.6). The effect of these time-varying fields is perhaps most easily understood in the context of T2 decay.

70

Deoni

Fig. 4.6. A precessing proton creates its own magnetic field. As a proton randomly moves, it creates and experiences an oscillating magnetic field caused by neighboring protons. This basic principal lies at the heart of the relaxation processes.

As the precessional frequency of a proton is governed by the magnitude of the magnetic field, variations in this field will alter the precessional frequency of the proton. In a collection of randomly moving protons, with their magnetic moments initially aligned (in phase) with each other, each proton will experience slightly different field oscillations as they move about and collide. The precessional frequency of each proton is, therefore, time varying and will differ for different protons. This variation between protons causes the individual moments to become out of phase with each other (de-phase), and the net magnetic moment (and, therefore, signal) is reduced. At equilibrium, the magnetic moments are randomly aligned in the transverse plane (Fig. 4.2) and produce no net transverse magnetization. Bloembergen, Purcell, and Pound presented the theoretical relationship between T1 and T2 and magnetic proton–proton interactions resulting from molecular motion in their seminal 1948 paper (4) and is referred to as the BPP theory of relaxation. Although derived for protons in idealized, homogeneous media (such as aqueous solutions), these expressions offer useful insight even in non-ideal heterogeneous media and are summarized here:

1 =K T1



τc 1 + ω02 τ 2c

+

4τc

 [1]

1 + 4ω 20 τ 2c

and 1 =K T2



3 5/2τc τc τc + + 2 2 2 1 + ω0 τc 1 + 4ω02 τc2

 [2]

Equations (1) and (2) relate relaxation time, Larmor frequency, molecular rotational motion (denoted by |C or the correlation time), and the strength of proton interactions (K). The correlation time is most easily considered a measure of molecular motion and reorientation (also referred to as tumbling), with short cor-

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

71

relation times expressing rapid motion and long correlation times corresponding to slow molecular motions. Graphical evaluations of these expressions (Fig. 4.7) demonstrate the behavior of T1 and T2 in different biophysical environments.

Fig. 4.7. The BPP theory predicts differential behavioural for T1 and T2 . At slow molecular motions, such as in rigid materials, T1 is long while T2 is extremely short. For soft tissue, with intermediary motion, T2 < T1 . In unrestricted aqueous solutions, such as cerebral spinal fluid, T2 ≈ T1 .

Though both T1 and T2 are sensitive to motion at the Larmor frequency, the additional |C term in the T2 expression accurately describes the sensitivity of T2 to slow motions, such as in rigid structures including bone and teeth that are difficult to image due to their rapid signal decay. In unrestricted aqueous solutions, such as cerebral spinal fluid, T2 is approximately equal to T1 . Representative T1 and T2 values for gray matter, white matter, spinal cord and optic nerve at 1.5, 3, and 7 T are presented in Table 4.1 (5) (a more extensive review of relaxation times in various tissues and across field strengths is provided by Bottonley et al. (6)). The enhanced sensitivity of T1 to motion at the Larmor frequency explains the prolongation of T1 at higher field strengths. In addition to the motion-related magnetic field fluctuations, variations in the local magnetic field are also caused by macroscopic inhomogeneity of the external magnetic field and the presence of large paramagnetic molecules with different magnetic susceptibilities from tissue. These larger scale variations accelerate the de-phasing and signal decay beyond T2 and can be characterized by an additional term, T  2 . The signal decay caused by the combination of molecular motion and macroscopic inho-

72

Deoni

Table 4.1 Representative T1 and T2 times for an assortment of neurological tissues at 1.5, 3, and 7 T 1.5 T

3.0 T

7.0 T

Tissue

T1 (ms)

T2 (ms)

T1 (ms)

T2 (ms)

T1 (ms)

White matter

778 (±84)

79 (±8)

1,084 (±45)

56 (±8)

1,220 (±36) 2,132 (±103)

Gray matter

1,086 (±228)

95 (±8)

1,820 (±114)

71 (±10)

Spinal cord

815 (±30)

77 (±9)

993 (±47)

78 (±2)

Optic nerve

745 (±37)

74 (±6)

1,083 (±39)

78 (±5)

T2 (ms)

mogeneity is denoted by T2 ∗ and given by 1 1 1 = + . ∗ T2 T2 T2

[3]

Though the BPP theory provides general insight into the relationships between T1 and T2 and gross tissue structure and describes the physical mechanism underpinning relaxation, it is insufficient to fully characterize complex environments, such as human brain tissue. Proton–proton interactions between water protons and larger macromolecules, proteins and lipids, the presence of paramagnetic atoms and molecules (i.e., iron, hemoglobin), magnetic shielding by macromolecules, and the multiple aqueous environments contained within a single imaging voxel, all influence the local relaxation times and are beyond the scope of Eqs. [1] and [2]. Nevertheless, while it is not currently possible to theoretically predict the relaxation times for different tissues, the BPP expressions can provide an appreciation of expected tissue differences. The lipid-rich myelin sheath and associated proteins, cholesterol, iron-containing oligodendrocytes and glial cells, combined with reduced free water content, is primarily responsible for the shorter T1 and T2 of white matter compared to gray matter (7). Similarly, factors such as differing concentrations of iron, principally in the form of ferritin, give rise to the T1 , T2 , and T2 ∗ variations observed between deep gray matter structures (8, 9). Developmental changes, such as myelination (as well as de-myelination and dys-myelination), axonal growth, and gyrification, as well as pathological processes including edema, inflammation, tumor infiltration, iron accumulation, and necrosis, alter the local tissue structure and biochemistry and, consequently, can lead to substantial changes in T1 and T2 (2, 7, 10–13).

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

73

3. Methods 3.1. Measurement of T1 and T2 Relaxation 3.1.1. Conventional Techniques 3.1.1.1. T1 Measurement

Despite their generally lengthy acquisition times, the most common techniques for measuring T1 and T2 remain multiple inversion-time inversion recovery (IR) and multiple echo-time spin-echo (T2 -CPMG), respectively. The classic inversion recovery sequence comprises a 180◦ inversion pulse, an inversion time (TI) delay during which the magnetization recovers, and a conventional spin-echo (90–180◦ combination) to sample the recovered magnetization. Before the next inversion pulse, the spin system is usually allowed to return to thermal equilibrium. Following this sequence step-wise (illustrated in Fig. 4.8), the inversion pulse inverts the net magnetic moment from +Mo to −Mo . During TI, individual spins return to their equilibrium orientation and the magnetic vector slowly recovers to +Mo . The recovered magnetization vector can be

Fig. 4.8. Graphical illustration of the magnetization dynamics during an inversion recovery sequence.

74

Deoni

separated into two components: a longitudinal component, ML , and a transverse component, MT . The 90◦ pulse rotates the recovered longitudinal magnetization ML into the transverse plane where it can be measured. As will be discussed in the following section, the 180◦ pulse refocuses the transverse magnetization and is measured at time TE. The spin system is then allowed to recover to time TR (the repetition time). To reconstruct the T1 recovery curve, the sequence is repeated with a range of inversion times and T1 is estimated by fitting the function S(T1 ) = ρ[1 − βe + e −TR/T 1 ]

[4]

R R to the data.  is included to account for imperfor T1 , and  fect inversion pulses and is often referred to as the ‘inversion efficiency.’ In the case of full magnetization recovery (TR >> T1 ), the exp(−TR/T1 ) term is omitted from Eq. [4]. Robust estimation of these three parameters requires fitting the data to more than three signal measurements and, in general, at least 7–8 points along the T1 recovery curve are usually sampled (Fig. 4.9). Example multiple TI inversion recovery brain data is shown in Fig. 4.10, along with corresponding calculated T1 , ρ, and β maps.

Fig. 4.9. Measurement of T1 with an IR experiment. Following an inversion pulse and delay (TI), the longitudinal magnetization is sampled (black circles) and then allowed to fully recover to equilibrium. This process is repeated with varied TI to characterize the recovery curve. Theoretically, the magnetization recovers along the dashed curve. However, in practice, we measure the magnitude of the signal (solid curve). T1 is calculated by fitting the IR signal expressions to the measured data.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

75

Fig. 4.10. Example IR T1 mapping experiment data comprising images acquired with different inversion times and the corresponding calculated T1 , ρ, and inversion efficiency maps.

Sample Protocol (3.0 T): Axial oriented 25 × 25 cm field of view (FOV); 256 × 128 matrix; 5 mm slice thickness; TE = 10 ms; TR = 6,000 ms; TI = {50, 100, 150, 200, 400, 800, 1,600, 3,200} ms; Flip angle = 90◦ ; Receiver bandwidth (BW) = 488 Hz/voxel; Acquisition time = ~13 min per TI image. 3.1.1.2. T2 Measurement

T2 is customarily measured using a Carr–Purcell–Meiboom–Gill (CPMG) spin-echo sequence (14, 15) consisting of a 90◦ pulse followed by a series of equally spaced 180◦ refocusing pulses (separated by echo time TE) with the magnetization measured at the mid-point between 180◦ pulses. The 180◦ pulses eliminate the effect of macroscopic field inhomogeneities and other non-motion-related sources of T2 dephasing (15). If we consider a pair of neighboring stationary protons with aligned magnetic moments, but each experiencing slightly different magnetic fields due to a macroscopic inhomogeneity (B0 and B0 + B). After time TE/2, the individual magnetic moments will be separated by angle B × TE/2. The 180◦ pulse ‘flips’ the spin system, inverting this angle from B × TE/2

76

Deoni

to – B × TE/2. After a second TE/2 interval, the total angle difference between the two protons is – B × TE/2 + B × TE/2 = 0. This is graphically illustrated in Fig. 4.11 and is also commonly explained using the running race example, where each proton is a runner in a race. At the half-way point (TE/2), the runners turn around and run back to the start line and are once again realigned in spite of the different speeds.

Fig. 4.11. The basics of a spin echo. Following an RF pulse (time = 0), all the proton moments are aligned in phase. Over time, these moments disperse or fan out. At time TE/2 a 180◦ pulse is applied which ‘flips’ the protons and the moments begin to re-phase, forming an echo at time TE.

However, it is important to note that random magnetic field variations caused by proton thermal motion will not be re-phased. Thus, over a series of spin echoes, the magnetization measured at each echo decays according to T2 (Fig. 4.12). To reconstruct the T2 decay curve (Fig. 4.13), 7–8 spinechoes are acquired. T2 is estimated either by a non-linear or linear (after log transforming the data) fit of SCPMG = ρe −TE/T2

[5]

to the data for T2 and ρ. While the linear property of Eq. [5] allows estimation of T2 from as few as two points, within heterogeneous tissues such as brain white and gray matter, two point measures are seldom enough to accurately characterize the relaxation curve and the estimates will be sensitive to the chosen TE values (16). An example spin echo data set is shown in Fig. 4.14, along with corresponding calculated T2 and ρ maps. Sample Protocol (3.0 T): Axial oriented 25 × 25 cm field of view (FOV);

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

77

Fig. 4.12. Magnetization envelope over a series of spin echoes. During the free induction decay (FID) the signal decays by T2 ∗ . The 180◦ spin echo at time TE/2 causes an echo to form at time TE. However, the overall amplitude of the echoes decreases by T2 .

Fig. 4.13. Measurement of T2 with a CPMG experiment. Following a 90◦ saturation pulse, an echo is formed at time TE, the transverse magnetization sampled, and then allowed to recover back to equilibrium. This process is repeated for different TE to characterize the decay curve and T2 is calculated by fitting the SE signal expressions to the measured data.

256 × 128 matrix; 5 mm slice thickness; TR = 10,000 ms; TE = {10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160} ms; Flip angle = 90◦ ;

78

Deoni

Fig. 4.14. An example of spin echo T2 mapping experiment consisting of eight spin echo images, along with corresponding calculated T2 and ρ maps.

Receiver bandwidth (BW) = 488 Hz/voxel; Acquisition time = ∼21 min per slice. 3.1.1.3. T2 ∗ Measurement

To measure T2 ∗ , the 180◦ (spin-echo) pulses are replaced with a pair of balanced, but opposing magnetic field gradients (creating a gradient echo). Considering our neighboring stationary protons again, if a field gradient is applied across them, then after time TE/2, their individual magnetic moments will be separated by B × TE/2 + δB. Here B is the macroscopic field inhomogeneity and δB is the difference in the field due to the applied gradient. At time TE/2, the applied gradient is reversed, so that over a further TE/2 interval, the angle between the protons is B × TE/2 – δB. Summing these results gives the overall angle at time TE of B × TE. Unlike the spin echo case, where the 180◦ pulse perfectly corrected for B, in the gradient echo case this factor is not corrected and the signal is a function of T2 ∗ rather than T2 . Measurement of T2 ∗ is, therefore, performed in a similar manner as T2 but using a gradient recalled echo sequence (GRE) (Fig. 4.15), with the signal is modeled as ∗

SGRE = ρe −TE/T2 .

[6]

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

79

Fig. 4.15. Magnetization envelope over a series of gradient echoes. Application of a gradient causes the magnetization to de-phase. Reversing the gradient, re-phases the spins to form an echo. However, macroscopic magnetic field inhomogenieties are not re-phased like in a spin echo.

3.1.2. Equation Model Fitting

Derivation of T1 , T2 , or T2 ∗ estimates from the acquired data requires accurate and precise fitting of the governing signal expressions (Eqs. [4], [5], and [6], respectively). Almost exclusively, this is performed through a least squares minimization approach, in which the sum-of-squares residuals between the model and acquired data are minimized (Fig. 4.16). Accepted methods for performing this minimization differ in their computational complexity, speed, and sensitivity to local minima. For the well-behaved functions described by Eqs. [4], [5], and [6], usual fitting approaches include Powell’s method (17), the simplex approach of Nelder and Mead (18), and gradient descent techniques, such as that of Levenburg and Marquardt (19). Each of these approaches begins from some initial estimation (guess) of T1 , T2 , T2 ∗ , ρ, and β and iterates to a solution.

Fig. 4.16. Least squares minimization curve fitting. a Relaxation time estimates are derived by minimizing the sum of squared residuals between the model and the acquired data. b Fitting algorithms iterate on a solution to provide the best fit of the model to the data.

80

Deoni

While these fitting routines afford rapid convergence times, they can suffer sensitivity to local minima and depend upon the initial estimate. More global search techniques, such as genetic algorithms (20), simulated annealing (21), swarming (22), or region contraction (23), can alleviate these local minima concerns, but at the expense of computation time. 3.1.3. Rapid Techniques for Measuring T1 and T2

Although inversion recovery and CPMG approaches are considered the ‘gold standards’ for T1 and T2 estimation, they suffer lengthy acquisition times making them unsuitable for most clinical applications. Consequently, a number of accelerated techniques have been proposed. Of these, we will briefly describe the more commonly employed Look–Locker and variable flip angle spoiled gradient (DESPOT1) method for measuring T1, the variable flip angle steady-state-free precession (DESPOT2) technique for measuring T2, and the inversion-prepared steady-state-free precession technique for simultaneous measurement of T1 and T2 . Many of the alternative techniques proposed may be considered as variants of these or the IR and CPMG techniques. For example, echo planar imaging (EPI) or spiral readout trains in combination with the IR approach can significantly shorten the acquisition time associated with IR.

3.1.3.1. T1 Measurement with the Look–Locker Method

Originally proposed by Look and Locker in 1970 (24), and later evolving into the TOMROP (T One by Multiple ReadOut Pulses) (25), this technique offers a subtle but important distinction to the conventional IR approach. With IR, the recovering longitudinal magnetization is sampled by tipping it back into the transverse plane (90◦ pulse) and then forming a spin echo. The 90◦ pulse saturates the magnetization, substantially disrupting the longitudinal recovery process and necessitating a lengthy delay to allow the spin system to recover fully. However, if the 90◦ pulse is replaced by a much smaller RF pulse and the spin-echo measurement by a gradient-echo measurement, the longitudinal recovery is only moderately disrupted and can be sampled continuously without requiring full recovery to equilibrium. As a result, only a single inversion pulse is needed, following by a train of small angle α pulses (Fig. 4.17) to measure the T1 relaxation curve. While conceptually simple, this technique can be confounded by corruption of the T1 signal by residual transverse magnetization and stimulated echoes. The small portion of the magnetization that is tipped into the transverse plane by each α pulse decays with rate R2 . If the time between successive α pulses is less than T2 , a portion of this magnetization will still be present when the next α pulse is applied. The magnetization in the transverse plane will, therefore, contain the residual magnetization from the preceding pulse, plus the amount tipped by the current pulse,

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

81

Fig. 4.17. In the method of Look and Locker, the longitudinal magnetization recovery is sampled with small angle RF pulses following a single inversion pulse. This eliminates the need to apply multiple inversion pulses and wait for the magnetization to fully recovery, as required by the IR approach.

corrupting our measurement. It is essential, therefore, that any magnetization in the transverse plane must be eliminated between each pulse. This can either be accomplished by spacing the α pulses more than 5 × T2 apart (i.e., long enough that less than 1% of the magnetization remains) or by spoiling the transverse magnetization. By applying a magnetic field gradient across the voxel, we can de-phase the protons, spreading their individual magnetic moments about 360◦ (or some other increment of PI) and eliminating the residual coherent transverse magnetization within the voxel (Fig. 4.18). Such a process is termed gradient spoiling and is a common feature of most rapid T1 -weighted imaging sequences, including SPGR and spoiled FLASH. In the above description, we assumed that the magnetization recovery is not disturbed by the train of small angle α pulses. In practice, however, even a very small α pulse (less than 5◦ ) will sufficiently disturb the longitudinal magnetization recovery depending on the T1 of the sample. The effect of this continued perturbation is to drive the recovery to equilibrium via an effective T1 , T1 ∗ , related to T1 and equal to T1∗ =

T1  . 1 − T1 TRln(cos α)

[7]

In practical application, a T1 map image is created by fitting for T1 ∗ from the acquired data and then converting to T1 . Despite offering substantive time savings over the conventional IR approach, for large three-dimensional volumes the

82

Deoni

Fig. 4.18. Gradient spoiling of the transverse magnetization. Applying a gradient across each voxel causes the proton moments to de-phase, eliminating, or ‘spoiling’ the transverse magnetization.

Look–Locker technique can still require several hours since the inversion pulse – α pulse train is repeated for each phase-encode step. Accelerated three-dimensional variants have, therefore, been presented requiring only minutes for reasonable spatial resolution (26, 27), or EPI readouts are used to reduce the number of phase encodes (28). 3.1.3.2. The Method of Variable Flip Angles: Driven Equilibrium Single Pulse Observation of T1 (DESPOT)

The principal rate-limiting factors in inversion recovery are (1) the inversion pulse and subsequent inversion time delay, which increases for each measurement; (2) the 90–180 spin-echo measurement, which perturbs the longitudinal magnetization recovery; and (3) the lengthy delay required to allow the spin system to return to equilibrium before repeating the process. The Look– Locker approach addressed the latter two points, replacing the spin-echo measurement with a train of small tip angle pulses and gradient-echo measurements and eliminating the length recovery time. What if the inversion pulse and delay were also eliminated, leaving only a train of small angle, swiftly spaced, RF pulses? Curiously, if the transverse magnetization is adequately spoiled between pulses, this rapid application of α pulses drives the longitudinal magnetization to a dynamic equilibrium that is governed by the sample ρ, T1 , T2 ∗ , and the magnitude of the RF pulse. The described pulse sequence, referred to as SPGR (SPoiled Gradient Recalled echo) or spoiled FLASH (Fast Low Angle Single sHot), is a commonly used clinical sequence on account of its rapid acquisition time and high T1 -weighted signal, described by the expression SSPGR = ρ

1 − e −TR/T1 sin α

1 − e −TR/T1

cos α

∗

e −TE/T2 .

[8]

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

83

Provided TE is kept constant, or short relative to T2 ∗ ; TR kept constant; and ρ remains unchanged, SSPGR becomes a function of only T1 and α. By measuring SSPGR over a range of flip angles, an SSPGR vs. α curve like that shown in Fig. 4.19 can be obtained (the peak of which occurs at the Ernst angle,  αERNST = α cos e −TR/T1

[9]

Fig. 4.19. The SPGR signal curve. T1 can be calculated either by fitting the signal function to SPGR data acquired over a range of flip angles or by calculating where the peak of the signal curve occurs (Ernst angle).

The use of this signal curve to calculate T1 was first described by Christensen in 1974 in the context of NMR spectroscopy (29). Using the Ernst angle definition, Christensen showed that T1 could be calculated directly if the peak of the SSPGR curve is known. This concept was further refined by Homer and Beevers (30), who reasoned that only a single measure of SSPGR was necessary to calculate T1 (provided it was collected at the Ernst angle), and termed the method DESPOT – Driven Equilibrium Single Pulse Observation of T1 . However, this is a bit of a misnomer. To calculate T1 from a single point requires a priori knowledge of the Ernst angle, which is only known if T1 is also known a priori. In practice, a number of signal measurements are required in order to determine the curve peak and, subsequently, T1 . If several measures of SSPGR are acquired, however, it is more appropriate to fit Eq. [7] for T1 and ρe−TE/T 2 ∗ (31, 32) by exploiting the linearization property of the signal model. Rewriting the expression in the linear Y = slope × X + intercept form as

84

Deoni

SSPGR SSPGR ∗ = e −TR/T1 + ρe −TE/T2 tan α sin α

[10]

allows T1 to be calculated directly from the slope of the line. Further, since calculation of the slope of a line requires only two points, T1 may be calculated from just two SSPGR measurements. Despite the use of more than one point, this multi-point approach is still commonly referred to as DESPOT. Since the original description of this approach, a number of refinements and optimizations have been presented (33–35), making DESPOT one of the most common accelerated T1 measurement methods in use. 3.1.3.3. Driven Equilibrium Single Pulse Observation of T2 (DESPOT2)

In both Look–Locker and DESPOT techniques, magnetization spoiling was required to eliminate T2 effects from the T1 measurements. Without spoiling, the freely evolving signal depends upon ρ, T1 , T2 , T2 ∗ , RF pulse number, and flip angle in a complex manner. This freely evolving signal can, however, be modified by recycling (or refocusing) the transverse magnetization between RF pulses. In a typical gradient echo, the transverse magnetization is dephased, re-phased, and de-phased over the course of the applied gradient (Fig. 4.20). To recycle the magnetization, an additional re-phasing gradient lobe is added. The described sequence, termed SSFP (Steady-State Free Precession), balanced SSFP, or true-FISP (Fast Imaging with Steady Precession), is one of the oldest sequences in NMR (dating back to the 1950s (36)) and was the basis of the one of the very first NMR imaging methods (37). Under ideal conditions, the resultant SSFP signal is well described by the expression SSSFP



1 − e−TR/T1 e−TE/T2 sin α

. =ρ 1 − e−TR/T1 e−TR/T2 + e−TR/T1 − e−TR/T2 cos α [11]

If TR is kept constant and ρ remains unchanged; SSFP becomes a function of T1 , T2 , and flip angle. If T1 is known a priori, SSFP may be used to calculate T2 in an analogous fashion as to how T1 is calculated using SPGR. Similar to Eq. [7], the SSFP signal expression can be rewritten in the Y = slope × X + intercept form as −TR/T



1 − e−TR/T2 S e ρ 1 − e−TR/T1 e−TE/T2 SSSFP SSFP = + tan α 1 − e−TR/T1 e−TR/T2 sin α 1 − e−TR/T1 e−TR/T2 [12] with T2 calculated from the slope of the line. This linear property allows T2 to be calculated from as few as two SSFP measurements.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

85

Fig. 4.20. Gradient refocusing in SSFP. Balancing the gradients around the echo refocuses the transverse magnetization just before the next RF pulse.

The similarity of this technique with the DESPOT approach led to its being dubbed DESPOT2 – Driven Equilibrium Single Pulse Observation of T2 . The combination of DESPOT and DESPOT2 allows combined calculation of T1 and T2 from as few as four rapidly acquired images. 3.1.3.4. Simultaneous T1 and T2 Measurement with IR-SSFP

Given the mixed contribution of T1 and T2 to the SSFP signal, is it necessary to calculate T1 separately to T2 ? Indeed, cannot both be calculated directly from the SSFP signal? Scheffler and Hennig (38) first demonstrated the ability to quantitatively measure T1 using an inversion prepared (IR-) SSFP sequence. The inversion pulse was added to increase the T1 dependence of the signal (97). This technique is similar to the Look–Locker approach described above, but with an SSFP readout used in place of the spoiled FLASH readout, and like the Look–Locker signal, the longitudinal magnetization in IR-SSFP is driven to equilibrium via an effective T1 , which depends on the flip angle and sample T1 and T2 , as T1∗

 =





1 1 2 α 2 α cos /2 + sin /2 . T1 T2

[13]

86

Deoni

Performing a more thorough examination of the IR-SSPP signal, Schmitt et al. (40) used the effective T1 recovery rate, with combined T1 and T2 influence, to simultaneously measure both relaxation times. The rapid acquisition times associated with this technique make it a useful alternative to the combined DESPOT and DESPOT2 techniques. 3.1.4. Errors in T1 and T2 Measurements

Though straightforward in concept, accurate and precise measurement of T1 and T2 using the described techniques requires careful consideration of the potential sources of error. Over the following section, we outline some of the more egregious pitfalls and describe techniques for avoiding or correcting for them.

3.1.4.1. Flip Angle Inhomogeneity

The accelerated measurement techniques described above rely on small tip angle RF pulses to sample the magnetization and incorporate these flip angles values into the T1 calculation itself. Thus, accurate knowledge of the applied flip angle is essential to correct T1 and T2 estimations. Deviations of the transmitted flip angle from the intended or ‘prescribed’ value arise from two main sources: RF pulse profile errors and RF attenuation and tissue dielectric effects. Ideally, the excitation profile of the RF pulse has a square shape, providing the desired flip angle value across the excited volume or slice and zero elsewhere (Fig. 4.21). Unfortunately, this is rarely the case. Instead, the pulse profile is more apt to be Gaussian shaped with the transmitted flip angle varying across the volume or slice. For three-dimensional (3D) volumetric acquisitions, this profile affect can be tolerated if the anatomy of interest is within the center portion of the excited slab, where the flip angle is approximately uniform and of the desired value. For single and multiple 2D slice applications, this profile effect will yield variation through the image slice, with the measured signal becoming an integrated function of flip angle (41).

Fig. 4.21. Imperfections in the RF pulse profile can lead to significant variations in signal across the image volume.

RF coil inhomogeneities and RF attenuation and dielectric resonance effects can also produce significant deviations from the intended flip angle throughout the image volume. Asymmetric RF coils have non-uniform RF power profiles that cause the transmitted flip angle to vary with distance from the coil. Dielectric

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

87

resonance, or RF penetration, effects are perhaps best understood through consideration of the RF wavelength relative to object size. When the wavelength is large relative to the object, the transmitted power is approximately equal throughout the volume. However, when the RF wavelength and object size are of the same order, the RF power varies throughout the object, resulting in variations in the transmitted flip angle. These effects become particularly pronounced at high magnetic field strengths where the RF wavelength decreases. Minimization of RF pulse profile and attenuation and dielectric effects can be achieved through improved RF pulse design (42), use of B1 insensitive pulses (43, 44), numerical modeling (41), or calibration of the transmitted flip angle (45). Through optimized RF pulse design, such as SLR pulses (42), the RF pulse shape can be made as box like as possible, minimizing the fall-off regions at the edges of the slab. In single-slice applications, were pulse design is much less of a factor, Parker et al. (41) showed how the flip angle could be numerically modeled and accounted for in T1 measurements. Composite or fast passage adiabatic pulses, which are less sensitive to RF attenuation and dielectric effects, are perhaps the most obvious approach to eliminating flip angle errors. However, these pulses suffer lengthy pulse durations (potentially doubling imaging time) and can have high energy deposition (limiting their use at high field strengths or in pediatric populations). Quantitative measurement, or calibration, of the transmitted flip angle has received increased attention with the move to higher magnetic field strengths and has benefited from a proliferation of rapid techniques. The most common of these techniques, the double-angle approach (45) acquires two spin-echo images with flip angles α and 2α and through a trigonometric relationship determines α through the ratio of the signal intensities. Rapid volumetric approaches, such as (46–49), also provide robust calculation of the flip angle field and are fast enough to be used as a calibration step before a T1 or T2 experiment. A less common approach improving flip angle spatial uniformity is parallel transmission (50). Here the RF power of the independent elements of modern multi-element RF coils is tuned and optimized to produce an overall uniform pulse profile throughout the object. This approach has the added advantage of reducing the overall RF power deposition, an important consideration at higher magnetic field strengths or in certain clinical populations. 3.1.4.2. Residual or Incoherent Transverse Magnetization

When the time between successive RF pulses (TR) is greater than 5 × T2 and T2 ∗ , there is little coherence among the proton spins in the transverse plane (i.e., the transverse magnetization is naturally spoiled or de-phased). However, in accelerated T1 measurement methods, TR is generally much less than T2 and T2 ∗ . Thus,

88

Deoni

there is residual transverse magnetization, which left unchecked, will introduce unwanted T2 weighting into the signal and result in significant errors in the T1 estimates (51). To eliminate this residual transverse magnetization, a combination of RF and gradient spoiling is necessary and is an essential feature of SPGR and spoiled FLASH sequences. Though spoiling is not used in SSFP, the related phenomenon of incoherent transverse magnetization is a significant confound. The simplified balanced SSFP signal model provided by Freeman (52) (Eq. [10]) assumes the transverse magnetization is perfectly recycled at the end of the TR interval. Unexpected precession of the magnetization, caused by unbalanced gradient errors or, more commonly, susceptibility-induced off-resonance, results in deviations of the signal from this theoretical value. In areas of susceptibility-induced field gradients, a well-known banding artifact can appear (Fig. 4.21), corrupting T1 and T2 estimates made with IR-SSFP or DESPOT2. A common approach to deal with this artifact, presented by Zur et al., is RF phase cycling (53). An RF pulse acts to rotate the magnetic moment around an axis. A 90◦ pulse, for example, rotates the longitudinal magnetization around the y-axis, yielding a magnetic moment oriented along the x-axis (Fig. 4.22a). While we may assume that each RF pulse in an imaging sequence is identical, rotating the magnetization around the same axis of rotation, this is generally not the case. Rather, in practice the axis of rotation is incremented around the XY-plane (Fig. 4.22b) by either a constant or a variable angle and the angle between the axis of rotation and the x-axis is termed the phase ( ) of the RF pulse. An RF pulse that rotates the magnetization around the y-axis, for example, has 0◦ phase, while an RF pulse that rotates the magnetization around the x-axis has a phase of 90◦ . In the SSFP sequence, each RF pulse is applied with a constant increment added to the RF phase (such as 180◦ ). Changing this increment has an unexpected effect on the acquired image, shifting the spatial location of the SSFP banding artifact in the image. An artifact-free SSFP image can, therefore, be obtained by combining two images acquired with different RF phases, usually 0 and 180◦ (Fig. 4.23). Deoni et al. (54) has shown how this technique can also be used to calculate artifact-free T2 maps using DESPOT2. 3.1.4.3. Ensuring Steady State

Most accelerated T1 and T2 relaxation measurement techniques utilize steady-state imaging sequences, where the magnetization is driven to a dynamic equilibrium that depends on the time between RF pulses, flip angle, and T1 and T2 . Care must be taken to ensure this dynamic steady-state condition is established before data acquisition, a process that can take several seconds (generally at least a time equal to T1 ). In practice, this condition

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

89

Fig. 4.22. Rotation of the magnetization by an RF pulse around an arbitrary axis of rotation. Conventionally, we assume a single constant axis of rotation (such as the Y-axis) that each RF pulse tilts the magnetization around (a). In practice, the axis of rotation is incremented around the XY-plane on each pulse (b) by some phase angle, . In SSFP, the RF pulse is commonly incremented by = 180◦ , so that the tipped magnetization alternates between pointing in the +X and –X directions (c).

Fig. 4.23. Illustration of RF pulse phase cycling in SSFP. Incrementing the phase of each RF pulse shifts the spatial location of signal bands (yellow arrows). The maximum intensity projection of two SSFP images acquired with different phase-cycling patterns (phase angles) can produce an artifact-free image (right panel).

90

Deoni

is achieved by playing out a series of ‘dummy’ pulses, without signal acquisition, before imaging. Though this process may add several seconds to the acquisition, it is essential to correct T1 or T2 estimation. 3.1.4.4. Movement and Flow

Beyond bulk motion artifacts (i.e., ghosting, blurring), which can have devastating effects on image appearance and the derived T1 and T2 measures, physiological motion can produce more subtle effects. Introduced in the above section, most accelerated techniques require the establishment of a steady state. In the case of T1 , thorough spoiling of the transverse magnetization is also necessary. For moving tissues, such as flowing blood, these conditions may be violated. Depending on flow rate and the extent of the excited image volume, flowing blood may have exited the volume before it has reached steady state. Further, the movement of blood protons through the applied spoiling gradients may result in incomplete spoiling or worse, refocused magnetization. In either case, the estimated T1 values will be erroneous. In large volume three-dimensional applications, this effect may be subtle, as the flowing magnetization will evolve into steady state as it navigates the image volume, leaving only subtle artifact near the edge of the volume where flow initiates. Application of a large spoiling gradient can further assist in eliminating unwanted residual transverse magnetization, even in the presence of rapid flow. In single and multiple slice 2D applications, this artifact can be far more insidious, requiring the use of saturation bands, which null the flowing signal immediately outside the slice of interest. It is also unlikely flowing blood will achieve adequate steady state before exiting the slice, making 2D approaches ill suited to quantifying blood T1 or T2 .

3.2. Exchange and Multiple Component Relaxation

Throughout this chapter, we have assumed the relaxation of signal in each image voxel is characterized by a single T1 and T2 . This is analogous to assuming only a single water environment within each image voxel. In tissue, water is compartmentalized into multiple distinct micro-anatomical environments, each with unique biophysical and biochemical properties and, therefore, distinct T1 and T2 characteristics. Further, if the boundaries between these compartments are permeable to water, protons may easily exchange between them. When averaged over the spatial dimensions of a voxel, this exchange and differential relaxation result a complex signal decay. ill-described by a single T1 or T2 value. Such values are more correctly termed effective relaxation times and represent a weighted average of the compartmental T1 and T2 values and exchange rate (16). The distinct relaxation properties of each compartment, however, provide a potential means for discriminating and

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

91

isolating the individual signal. Within brain and spinal cord, multicomponent analysis of T2 and, more recently, T1 and T2 ∗ relaxation data provides evidence of at least two distinct and reproducible water environments (55–59). These are broadly attributed to (1) the intra- and extra-cellular water and (2) water trapped between the hydrophobic lipid bilayers of the myelin sheath (55). Quantification of this latter component provides a non-invasive means of measuring and monitoring myelin content and has been applied to the study of known and suspected de- and dys-myelinating white matter disorders (such as multiple sclerosis (60) and schizophrenia (61)). An important consideration in multicomponent analysis is the temporal timescales of relaxation and the magnetization exchange rate between water compartments. Zimmerman and Britten coined the concept of exchange regime (62) to describe this phenomenon. In the fast exchange regime, exchange is rapid relative to the relaxation timescale. Due to the quick mixing of the environments, an averaged, mono-exponential relaxation curve is produced, masking the information from each compartment. In contrast, in the slow exchange regime where the exchange time is slow relative to the T1 and T2 of the individual components, a multi-exponential relaxation curve is observed and the individual compartments can be interrogated. It is conventionally assumed that in brain tissue, exchange is fast relative to T1 , but slow relative to T2 and T2 ∗ . 3.2.1. Measurement of Multicomponent Relaxation

The gold standard approach to measuring multicomponent relaxation remains the multi-echo T2 approach championed by McKay, Whittall, and colleagues (58). Assuming the proton and magnetization exchange time between water compartments is slow relative to T2 , the spin-echo signal from a multicomponent system is a general expansion of Eq. [5]: S=ρ

N 

fi e−TE/T2 ,

[14]

i=1

where fi and T2,i are the volume fraction and T2 is the relaxation time of the ith water component. To derive estimates of each components volume fraction and T2 , a Poon–Henckelman CPMG sequence (63) is used to sample upward of 32 uniformly spaced echo times between 10 and 3,200 ms with TR kept long to mitigate T1 effects. Discrete two, three, or N (i.e., continuous distribution) component models are fit to this data to derive the desired volume fraction and T2 values (Fig. 4.24). In the most commonly presented implementation, single-slice data can be acquired with scan times on the order of 16 min. Current implementations, however, can provide 8–12 contiguous

92

Deoni

Fig. 4.24. Multicomponent relaxation theory and practice. A simple model of brain tissue contains two water components: free intra- and extra-cellular water (blue) and water trapped between the lipid bilayers of the myelin sheath (green). The measured MR signal contains contributions from each of these water pools. MCR aims to reconstruct these individual contributions and quantify the volume of the myelin water pool (right).

2D slices in a similar time-frame. More recently, Du et al. (64) have proposed the alternative use of a multi-echo gradient echo sequence to derive similar information based on component T2 ∗ differences. Finally, building on the multicomponent T1 work of Kreis et al. (65) and Spencer and Fishbein (66), Deoni et al. (67) have recently presented a combined multicomponent relaxation technique that models both T1 and T2 effects, taking into account proton exchange between the water compartments. 3.3. Analysis of Relaxation Time Data

The qualitative nature of conventional T1 -and T2 -weighted images makes it difficult to perform direct quantitative comparisons between data acquired of different subject groups (i.e., patient and healthy control groups); longitudinally acquired data (for example, of a single subject group when establishing treatment efficacy); or across multiple imaging centers (in order to obtain sufficient data to study a rare disorder or phenotype).

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

93

Differences in scanner hardware, scanner drift, software upgrades, pulse sequence implementation differences, coil geometry, and placement differences all but make it impossible to directly compare qualitative T1 -and T2 -weighted data. A common approach to side-step this obstacle is to derive quantitative metrics from the inherently qualitative data and to perform comparative studies using these derivative metrics. The usual approach, referred to as voxel-based morphometry (68), is to segment the weighted image into maps of white matter and gray matter percentages (also termed white and gray matter density maps) to investigate changes in gray or white matter volume with pathology. Derivation and comparison of cortical thickness, through cortical gray matter segmentation, are also increasingly used (69). However, while these approaches may highlight regions of difference, they provide little information regarding the underlying basis of identified changes. Further, results are inherently sensitive to tissue contrast, which depends nonlinearly upon ρ, T1 , and T2 (and other non-tissue affects). For example, an increase in white matter T1 along a gray/white matter boundary could present as a corresponding increase in white matter density, despite there being no actual change in the amount of white matter. Alternatively, opposing changes in ρ and T1 (as seen in edema or inflammation) mask each other in a conventional T1 -weighted scan and may not present as a change in white matter density, despite there being a very real tissue change. An alternative to these indirect assessments of tissue structure, volume and density are direct comparison of the T1 and T2 relaxation times. Depending on the spatial extent and resolution of the acquired maps, comparisons can be at the whole brain, hemispheric, regional, white matter tract, or voxel-wise levels. In addition to group comparisons, a powerful attribute of relaxation data is the ability to perform single subject comparisons against population norms without requiring correction for scanner hardware, acquisition strategy, etc. (70). 3.3.1. Group-Wise Comparisons

The overwhelming majority of clinical and research structural neuroimaging studies involve normal vs. pathological comparisons to determine (1) if there is a difference in brain structure associated with the condition; (2) where in the brain those differences are manifested; and (3) how identified differences correlate with degree or severity of pathology. To address these basic questions, a number of analysis approaches have been developed and refined. Here we cover the more commonly used approaches, describing them in order to increasing spatial specificity.

3.3.2. Histogram-Based Comparisons

Histogram-based analysis offers a straightforward means of addressing the most basic question: is there a difference between

94

Deoni

normal and pathological tissue? (71) The method is advantageous in that it requires no spatial normalization (alignment of images from each participant) and it does not require an a priori hypothesis as to where changes might be expected. Indeed, no spatial information is provided. Further, histograms provide an intuitive and tangible medium for visualizing group differences. Calculation of T1 or T2 histograms, simply the frequency of binned values, is straightforward. Correction for brain volume is accomplished by normalizing the bin frequencies by the total number of voxels included in the histogram (or area under the histogram curve). Averaged patient and control histograms can be visually and statistically compared, with standard metrics of comparison including mean, median, mode, skewness, kurtosis, peak height, peak location, among others. An example of histogrambased comparison of white matter T1 and T2 in healthy adolescents and adolescents with autism is shown in Fig. 4.25.

Fig. 4.25. Histogram-based comparison of whole-brain white matter T1 (left) and T2 (right) in healthy young adults and those with autism. The T1 histogram reveals a global increase in T1 in autism.

The lack of any spatial information, however, is the primary disadvantage of histogram-based analysis. Further, the approach may be ill posed in cases where subtle, small scale, or regional differences are expected. Subtle or focal differences may unfortunately be obscured when included with whole-brain data. 3.3.3. Voxel-Based Comparisons

If the primary disadvantage of histogram analysis is the lack of spatial information, region of interest analysis provides an effective alternative. Such analysis may take the form of either using pre-defined regions, or more generally, to treat each voxel as an independent region of interest and to perform voxel-wise comparisons (68). Following linear or nonlinear spatial normalization of patient and control image data (68, 72), voxel-wise t-tests (or a nonparameter equivalent) with appropriate correction for multiple comparisons (58, 73) can identify regions of group difference. The ability to examine the whole brain, without requiring consideration of a priori hypotheses, offers tremendous potential for

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

95

speculative or exploratory studies, where affected regions may not be known before hand, or when the spatial progression of the pathology is the topic of study. An example of voxel-based comparison of T1 and T2 in schizotypy is shown in Fig. 4.26, demonstrating hemispheric differences in the relaxation times in normal vs. high schizotypy groups.

Fig. 4.26. Voxel-based T1 comparison of medium and high schizotypy patients. Voxels showing a significant difference following multiple comparison correction are shown in the far right panel (1–p-value map).

Potential pitfalls of voxel-based comparisons, however, include mis-registration or poor spatial alignment, too little or too much spatial smoothing, and inadequate correction for multiple comparisons (or a lack thereof). Each of these effects combines to make reproducibility across different research groups challenging [Ref Derek]. 3.3.4. Tract-Based Comparisons

A classic theme among neuroimaging researchers is the use of a connectionist approach to understand neurological or psychiatric disorders (74). This approach implies consideration of the white matter tracts which connect the disparate brain regions that comprise the integrated brain systems and networks. While voxel-based approaches can elucidate regions of differences, these regions may contain multiple independent white matter pathways connecting different gray matter regions. Thus, additional information may be gleaned by considering the T1 and T2 characteristics along specific tracts of interest (75). Two common approaches for isolating specific white matter pathways are the use of digitized atlases (76, 77) or the combined acquisition of relaxation and diffusion tensor imaging (DTI) data (78). As will be discussed later, diffusion imaging provides estimates of local fiber orientation (78). By stitching these independent orientation measures together (tractography), three-dimensional representations of the white matter paths may be reconstructed (79). Using either atlas or tractography data to supply regions of interest, T1 and T2 values within these regions can be isolated

96

Deoni

Fig. 4.27. Tract-specific comparison of T1 and T2 in the right and left superior longitudinal fasciculi (shown as the green volume rendering superimposed on the anatomical images) in healthy young adults and those with autism, respectively. Histograms of values along these tracts demonstrate substantive alteration in T1 and T2 .

and statistical comparisons made. As an example, Fig. 4.27 shows comparison of T1 and T2 histogram data for the left and right superior longitudinal fasiculi in patients with autism and healthy age, sex, and IQ-matched controls. 3.3.5. Comparisons with Population Norms

In a variety of disorders, group-wise comparisons are difficult or ill suited. For example, multiple sclerosis is characterized by acute focal white and gray matter lesions occurring throughout the brain and spinal cord (80). From a disease monitoring, prognosis or predictive perspective, it is not necessarily the lesions themselves that are of interest, but the surrounding ‘normal appearing’ white matter (81). Voxel-wise comparisons across groups are not appropriate given the near-random location of lesions. While histogram-based approaches are useful, they are void of the spatial information necessary to pin-point affected ‘normal appearing’ brain regions. Here, then, it is preferable to perform subjectspecific analysis, comparing each subject with a matched population normal (or average) template to identify affected areas. Similar to group-wise comparisons, a population template (comprising both the mean and the variance) can be determined by spatially normalizing healthy participant data (matched for age, sex, handedness, etc. as required) and calculating the voxel-wise mean and standard deviation. Identification of voxels or regions that differ substantively from the population norm can be determined through straightforward voxel-wise z-tests. Illustrated in Fig. 4.28 is a comparison of T1 data from an MS patient against a matched population average, demonstrating how this form of analysis can reveal affected white matter regions that appear normal on conventional clinical T1 - and T2 -weighted scans.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

97

Fig. 4.28. Single subject T1 analysis of an MS (relapsing remitting, EDSS score of 4) patient. The population average and variance were calculated from 18 healthy age-matched controls. Z-score analysis reveals significant distribution in cerebral white matter that appears normal in the patient’s clinical FLAIR image.

4. Applications of T1 and T2 Quantification

4.1. Improving Tissue Contrast

The fundamental relationship linking tissue biochemistry and micro-structure with T1 and T2 underpins the use of T1 and T2 to identify, investigate, diagnose, and monitor pathology. However, while T1 and T2 are exquisitely sensitive to tissue alteration, they are notoriously non-specific, requiring careful interpretation. Over the following section, a summary of prominent relaxation time measurement applications in neuroimaging is provided, highlighting some of the more robust neurological and psychiatric findings. A central goal in neuroimaging research is improved visualization of subtle structure. This goal can be achieved through a combination of increased spatial resolution and enhanced tissue contrast. Increases in spatial resolution have been made possible through the proliferation of high field strength imaging systems (i.e., 3 T), coupled with parallel reception techniques (82). However, the relationship between T1 , field strength, and T1 -weighted image contrast often means a trade-off between spatial resolution and contrast (Fig. 4.29). As T1 increases, T1 -weighted contrast decreases since the signal is related to exp(−TR/T1 ). Acquisition of T1 maps can circumvent this issue, providing near optimum contrast that increases with field strength, since the T1 difference between tissues also increases with field strength. Figure 4.30 shows a comparison of T1 -weighted and T1 maps acquired at 1.5 and 3 T using similar acquisition strategies. Variations in T1 , T2 , and ρ should not be considered independently as these parameters are inherently inter-related. However, these relationships can vary on a tissue-by-tissue and pathologyspecific basis. For example, the formation of the myelin sheath in white matter causes not only a decrease in T1 but also the displacement of free water decreases ρ. Associated changes in relaxation

98

Deoni

Fig. 4.29. Comparison of 1.5 and 3 T T1 -weighted and T1 map images acquired with similar imaging parameters. While image contrast is slightly worse in the 3 T-weighted image, the increased gray–white matter T1 difference provides improved contrast in the map image, relative to 1.5 T.

Fig. 4.30. Comparison of a T1 -weighted, T1 , and ρ map of the cerebellum. The increased T1 and ρ of the dentate nucleus (purple arrows) relative to the surrounding cerebellar white matter obscures the structure in the weighted image. However, the structure is easily visualized on the map images.

times and proton density can reduce T1 - and T2 -weighted image contrast and obscure tissue differences and boundaries. To illustrate this, Fig. 4.30 contains a high SNR T1 -weighted image of the cerebellum alongside corresponding T1 and ρ maps. While the deep cerebellar dentate nucleus is clearly visible on the map images (83), it cannot be distinguished from the surrounding cerebellar white matter in the weighted image. This masking is due to the prolonged T1 and increased proton density of the dentate

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

99

nucleus relative to the surrounding white matter. These associated changes effectively ‘cancel’ each other and the measured T1 weighted signal is identical to that of the white matter. In addition to the cerebellum, T1 maps show improved tissue contrast throughout the brain, revealing subtle myelo and cyto-architectural differences unseen on conventional T1 - and T2 weighted scans, as illustrated in Fig. 4.31.

Fig. 4.31. High spatial resolution T1 map images of different brain regions, demonstrating visualization of subtle myelo and cyto-architecture. Identified structures include the internal and external globus pallidus (GPi, GPe), anterior ventral lateral nucleus (AVL), ventral anterior nucleus (VA), ventral lateral nucleus (VL), medial dorsal nucleus (MD), sub-thalamic nucleus, substantia nigra (SN), line of Gennari, and cortical striations of the hippocampus.

4.2. Neurological Disorders 4.2.1. Multiple Sclerosis

Multiple sclerosis (MS) is a neurodegenerative and neuroinflammatory disorder characterized by focal white and gray matter lesions in the brain and spinal cord. The lipid myelin sheath within and surrounding these lesions has been damaged, reduced, or lost (80, 81, 84). Though gray matter lesions are also present, their role in the disorder is less well established, and until the recent proliferation of high-field (i.e., 7 T) scanners, have been detectable only through histological and histochemical approaches (86). The destruction of the myelin sheath, and replacement by inflammatory cells, free water, and other proteins, leads to substantive changes in T1 and T2 . Lesion areas commonly present hypo-intensely on T1 or hyper-intense on T2 -weighted scans (80). Based on this association between pathology and relaxation times, T1 and T2 have been proposed as surrogate markers of disease activity for monitoring and therapeutic trial outcome purposes.

100

Deoni

Though the principal presentations of MS on MR images are focal lesions, correlations between lesion number and extent with disease severity (EDSS) and activity have, unfortunately, been week (86). Recent interest has, therefore, turned to investigating the surrounding ‘normal appearing’ white matter (NAWM) as well as diffusely abnormal ‘dirty appearing’ white matter (DAWM). Investigations of NAWM and DAWM have revealed global alterations in T1 (81) and T2 (87) suggestive of wide-scale tissue disruption and have again suggested T1 and T2 may be suitable biomarkers. As MS is inherently a disease of myelin, direct investigation of myelin content is desirable (Fig. 4.32). The use of MCR, therefore, has played a significant albeit, to date, research-only role in characterizing myelin damage in MS (88). In addition to expected focal reductions in myelin water fraction within lesions, MCR has revealed significant alterations within NAWM (89) and shown potential to discriminate between acute, chronic and active, and chronic and inactive lesion subtypes (90).

Fig. 4.32. Example MWF image in MS compared with a conventional clinical T2 -weighted FLAIR image.

4.2.2. Stroke

Diffusion-weighted imaging is the de facto standard in clinical stroke imaging, providing clear visualization of stroke extent. However, the diffusion signal is unable to disambiguate salvageable from unsalvageable tissue, an important clinical distinction. The significant alteration of tissue biochemistry and microstructure immediately following stroke results in substantial increases in both T1 and T2 (primarily stemming from the influx of free water into the region). The magnitude of T1 and T2 prolongation has been shown to differentiate infarcted regions from the surrounding affected tissue (91) and that T1 is superior to either T2 or diffusion in detecting early ischemic change.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

101

The time course of relaxation time changes is also informative of underlying tissue pathology and of diagnostic utility. The early and rapid T1 and T2 increases are followed by gradual declines to a non-normative equilibrium 2–3 days later (92). Measurement of these early changes has been demonstrated to accurately predict final infarct area (93). 4.2.3. Epilepsy

Atrophy of the hippocampus (hippocampal sclerosis, HS) is the most common cause of temporal lobe epilepsy. Though HS is commonly associated with an increased T2 -weighted signal, due to a prolongation of the T2 relaxation time, the ambiguous nature of T2 -weighted signal changes make definitive diagnosis challenging (94). As an adjunct to conventional spin-echo acquisitions, quantitative estimates of T2 in normal and pathologic hippocampal tissue are an effective method for the detection and monitoring of hippocampal structural changes (95). Evidence of T2 differentiation and prolongation in epilepsy is shown in Fig. 4.33.

Fig. 4.33. Comparison of hippocampal T2 maps from healthy (top) and epileptic (bottom) individuals showing the prolonged T2 of the HS patient.

In addition to T2 alterations, temporal lobe T1 values have also been shown to be significantly longer in epileptic patients compared with healthy controls (96). Further, T1 values throughout the hemisphere containing the seizure focus were also prolonged, intimating a potential role for quantitative T1 mapping in conjunction with EEG for identifying seizure foci. 4.3. Psychiatric Disorders: Dementia and Alzheimer’s Disease

T1 and T2 measurements have been studied in most dementia subtypes, including vascular dementia, dementia with Lewy bodies, and Alzheimer’s disease (AD). The classic neuropathological hallmark of AD is the presence of iron-containing beta amyloid plaque deposits. The iron composition of these plaques has been

102

Deoni

suggested as a means of permitting their visualization; however, to date, direct plaque visualization has only been possible in animal models or in vitro specimens (97) or at ultrahigh field strengths (i.e., 7 T). Measurement of more diffuse changes in T2 within the hippocampus, and basal ganglia, caused by aggregate accumulation of plaques, however, may reduce the need for direct plaque visualization (98). In addition to plaque deposits, white matter hyper-intensities are also commonly observed in AD (99). While the exact mechanisms behind these white matter changes remain unknown (though they may be associated with vascular changes), a recent disease model forwarded by Bartzokis (100) suggests white matter and demyelination may play an underlying role in this traditionally gray matter-centric disorder. Indirect support for this myelin hypothesis has been the observation of T2 increases (presumably due to decreased myelin and increased free water content) throughout the white matter of subjects reporting memory loss and confirmed AD patients (101). 4.4. Neurodevelopment

Abnormal brain maturation and neurodevelopment is a hypothesized substrate in a number of neurological and psychiatric disorders, including autism (102). Neurodevelopment is marked by a wide range of biophysical and biochemical changes, including axonal sprouting, synaptic pruning, and the establishment of the lipid myelin layer around axons (myelination). The influx of lipids, proteins, and other macromolecules associated with these processes can be indirectly monitored through T1 and T2 measurements. The influence of myelin precursory proteins and the sheath itself on T1 has been well established (7), and acquisition of T1 maps throughout neurodevelopment can provide a non-invasive window into the dynamic myelination process. Reduction of T1 throughout the white matter tracts broadly mirrors the histologically established time course of myelination. Alongside conventional (single-component) T1 measurement, multicomponent relaxometry can provide a more direct and quantitative assessment of myelination throughout neurodevelopment. Figure 4.34 shows an example of myelin water fraction maps obtained from six healthy infants from 3 through 8

Fig. 4.34. Myelin development throughout infancy. MWF maps acquired throughout the first 8 months of infancy were high pass filtered at the level of 3.5. A surface was fit to the remaining voxels.

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

103

months of age. Figure 4.34 provides a powerful yet intuitive view of white matter development during infancy and corresponds to spatiotemporal pattern derived from prior histological studies. While alternative imaging methods, including diffusion tensor and magnetization transfer imaging, have been proposed as surrogate markers of myelination, these techniques provide information that is only related to myelin content and not specific to it.

5. Toward a More Complete Picture of Tissue: Combining Relaxation Data with Other Forms of Imaging Information

The development of non-invasive imaging, including computed tomography (CT), positron emission tomography (PET), and MRI, has made possible the significant gains in our understanding of brain development, function, and the pathology that affects them. Within the field of MRI, different imaging and acquisition techniques provide complementary information that, when fused, can provide a more holistic view of tissue microstructure. Diffusion tensor imaging provides voxel-wise information regarding local fiber orientation, axonal density, and axonal size. T1 and T2 provide indirect measures of myelin and free water content, as well as lipid, protein, macromolecule, and paramagnetic material concentrations. Multi-component relaxometry provides a direct assessment of myelin content. Additional techniques, such as magnetic resonance spectroscopy, can provide further information related to metabolism, function, and integrity. Cumulatively, these data are informative of the principal facets of brain tissue microstructure, integrity, and function. An imaging protocol combining these elements would be well positioned to address crucial questions related to tissue alteration in pathology. The continued gains in imaging technology, including ever increasing magnetic field strengths, continued reductions in acquisition time through improvements in gradient technology and parallel transmission and reception, and the development of novel imaging pulse sequences, have brought the realization of such a multi-parametric, quantitative imaging protocol ever closer. Based on current state of the art, a combined whole-brain diffusion tensor/quantitative multi-component relaxometry protocol requires approx. 30 min. Single-slice chemical shift spectroscopic imaging can be completed in a further 10 min. Many functional MRI paradigms require less than 5 min with good statistical power. Thus, a single 45 min protocol could provide a new mechanism for investigating structure–function relationships and associated alterations in pathology. An example of a combined relaxometry/diffusion study is shown in Fig. 4.35. Though technically more challenging than conventional T1 - or T2 -weighted imaging, quantitative relaxation time

104

Deoni

Fig. 4.35. Combination of diffusion tensor tractography and relaxation time measurement. The colormap along the tracts shows the corresponding T1 and T2 values.

measurement affords a number of advantages. As outlined throughout this chapter, diagnostic observations in weighted images are non-linear and complex functions of the underlying relaxation times, acquisition strategy, and scanner hardware. Relaxation time measurement cleanly separates these individual contributors, providing a standardized basis for comparison. This not only provides an ideal basis for large, multi-center, and longitudinal research studies but also has clinical utility in diagnosis and treatment monitoring. Comparisons with population-based norms, as shown in Fig. 4.28, may be a crucial next step in bridging the gap between research studies and clinical adoption, allowing not only disease progression or treatment response to be monitored and quantified on an individual basis but also providing a more sensitive diagnostic tool.

Acknowledgments A special thanks is extended to those who provided clinical examples and imaging data used herein: Prof. Derek Jones, Dr. Shannon Kolind, Dr. Janneke Zinkstok, Dr. Marco Catani, Dr. Emma Burkus, Dr. Mark Richardson, Katrina McMullin, Catherine Traynor, and Sarah Kwan. A debt of gratitude is owed to Dr. David Lythgoe, Dr. Fernando Zelya, Astrid Pauls, and Katrina McMullen for proof reading and comments. References 1. Blinkov, S. M., Glezer, I. I. The Human Brain in Figures and Tables. New York, NY: A Quantitative Handbook. Plenum Press; 1968. 2. Damadian, R. V. Tumor detection by nuclear magnetic resonance. Science 1971;171:1151. 3. Rabi, I. I., Zacharias, J. R., Millman, S., Kusch, P. A new method of measuring

the nuclear magnetic moment. Phys Rev 1938;53:318. 4. Bloembergen, N., Purcell, E. M., Pound, R. V. Relaxation effects in nuclear magnetic resonance absorption. Phys Rev 1948;73:679–715. 5. Stanisz, G. J., Odrobina, E. E., Pun, J., Escaravage, M., Graham, S. J. et al. T1,T2

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

6.

7.

8.

9.

10.

11.

12.

13.

14.

15. 16.

relaxation and magnetization transfer in tissue at 3t. Magn Reson Med 2005;54: 507–512. Bottomley, P. A., Hardy, C. J., Argersinger, R. E., Allen-Moore, G. A review of 1 H nuclear magnetic resonance relaxation in pathology: Are T1 and T2 diagnostic? Med Phys 1987;14:1–37. Paus, T., Collins, D. L., Evans, A. C., Leonard, G., Pike, B., Zijdenbos, A. Maturation of white matter in the human brain: A review of magnetic resonance studies. Brain Res Bull 2001;54:255–266. Gelman, N., Ewing, J. R., Gorell, J. M., Spickler, E. M., Solomon, E. G. Interregional variation of longitudinal relaxation rates in human brain at 3.0t: Relation to estimated iron and water contents. Magn Reson Med 2001;45:71–79. Gelman, N., Gorell, J. M., Barker, P. B., Savage, R. M., Spickler, E. M. et al. MR imaging of the human brain at 3.0t: Preliminary report on transverse relaxation rates and relation to estimated iron content. Radiology 1999;210:759–767. Naoko, S., Sakai, O., Ozonoff, A., Jara, H. Relaxo-volumetric multispectral quantitative magnetic resonance imaging of the brain over the human lifespan: Global and regional aging patterns. Mag Reson Imaging 2009;27:895–906. Williams, L. A., Gelman, N., Picot, P. A., Lee, D. S., Ewing, J. R. et al. Neonatal brain: regional variability of in vivo MR imaging relaxation rates at 3.0t – initial experience. Radiology 2005;235:595–603. Hoque, R., Ledbetter, C., Gonzalez-Toledo, E., Misra, V., Menon, V. et al. The role of quantitative neuroimaging indices in the differentiation of ischemia from demyelination: An analytical study with case presentation. Int Rev Neurobiol 2007;79:491–519. Li, Y., Srinivasan, R., Ratiney, H., Lu, Y., Chang, S. M., Nelson, S. J. Comparison of T1 and T2 metabolite relaxation times in glioma and normal brain at 3t. J Magn Reson Imaging 2008;28:342–350. Carr, H. Y., Purcell, E. M. Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys Rev 1954;94: 630–638. Meiboom, S., Gill, D. Modified spin-echo method for measuring nuclear relaxation times. Rev Sci Instrum 1958;29:688–691. Whittall, K. P., MacKay, A. L., Li, D. K. B. Are mono-exponential fits to a few echoes sufficient to determine T2 relaxation for in vivo human brain? Magn Reson Med 1999;43:1255–1257.

105

17. Powell, M. J. D. A tolerant algorithm for linearly constrained optimization calculations. Math Programming 1989;45:547–566. 18. Nelder, J. A., Mead, R. A simplex method for function minimization. Comput J 1965;7:308–313. 19. Levenberg, K. A method for the solution of certain non-linear problems in least squares. Appl Math 1944;2:164–168. 20. Holland, J. H. Adaptation in Natural and Artificial Systems. Ann Arbor, MI: The University of Michigan Press; 1975. 21. Kirkpatrick, S., Gelatt, C. D., Vecchi, M. P. Optimization by simulated annealing. Science 1983;220:671–680. 22. Clerc, M., Kennedy, J. The particle swarmexplosion, stability, and convergence in a multidimensional complex space. IEEE Trans Evol Comput 2002;6:58–73. 23. Berger, M. F., Silverman, H. F. Microphone array optimization by stochastic region contraction. IEEE Trans Signal Proc 1991;38:2377–2386. 24. Look, D. C., Locker, D. R. Time saving in measurement of NMR and EPR relaxation times. Rev Sci Instrum 1970;41: 250–251. 25. Brix, G., Schad, L. R., Deimling, M., Lorenz, W. J. Fast and precise T1 imaging using a TOMROP sequence. Magn Reson Imaging 1990;8:351–356. 26. Henderson, E., McKinnon, G., Lee, T. Y., Rutt, B. K. A fast 3d look-locker method for volumetric T1 mapping. Magn Reson Imaging 1999;17:1163–1171. 27. Gai, N. D., Butman, J. A. Modulated repetition time look-locker (MORTLL): A method for rapid high resolution threedimensional T1 mapping. J Magn Reson Imaging 2009;30:640–648. 28. Freeman, A. J., Gowland, P. A., Mansfield, P. Optimization of the ultrafast look-locker echo-planar imaging T1 mapping sequence. Magn Reson Med 1998;16:765–772. 29. Christensen, K. A., Grand, D. M., Schulman, E. M., Walling, C. Optimal determination of relaxation times of Fourier transform nuclear magnetic resonance. Determination of spin-lattive relaxation times in chemically polarized species. J Phys Chem 1974;78: 1971–1977. 30. Homer, J., Beevers, M. S. A re-evaluation of a rapid ‘new’ method for determining NMR spin-lattice relaxation times. J Magn Reson 1985;63:287–297. 31. Wang, H. Z., Riederer, S. J., Lee, J. N. Optimizing the precision in T1 relaxation estimation using limited flip angles. Magn Reson Med 1987;5:399–416.

106

Deoni

32. Homer, J., Roberts, J. K. Routine evaluation of mo ratios and T1 values from driven equilibrium NMR specta. J Magn Reson 1990;87:265–272. 33. Deoni, S. C. L., Rutt, B. K., Peters, T. M. Rapid combined T1 and T2 mapping using gradient recalled acquisition in the steady state. Magn Reson Med 2003;49: 515–526. 34. Deoni, S. C. L., Peters, T. M., Rutt, B. K. Determination of optimal angles for variable nutation proton magnetic spin-lattice, T1, and spin-spin, T2, relaxation times measurement. Magn Reson Med 2004;51: 194–199. 35. Chang, L. C., Koay, C. G., Basser, P. J., Pierpaoli, C. Linear least-squares method for unbiased estimation of T1 from SPGR signals. Magn Reson Med 2008;60:496–501. 36. Carr, H. Y. Steady-state free precession in nuclear magnetic resonance. Phys Rev 1958;112:1693–1701. 37. Hinshaw, W. S. Spin mapping: The application of moving gradients to NMR. Phys Lett A 1974;48:87–88. 38. Scheffler, K., Hennig, J. T1 quantification with inversion recovery truefisp. Magn Reson Med 2001;45:720–723. 39. Deimling, M., Heid, O. (1994). Magnetization Prepared True FISP Imaging. In: Proceedings of the 2nd annual meeting of the ISMRM. San Fransico, p. 495. 40. Schmitt, P., Griswold, M. A., Jakob, P. M., Kotas, M., Gulani, V. et al. Inversion recovery truefisp: Quantification of T1, T2 and spin density. Magn Reson Med 2004;51: 661–667. 41. Parker, G. J., Barker, G. J., Tofts, P. S. Accurate multislice gradient echo T1 measurement in the presence of non-ideal RF pulse shape and RF field nonuniformity. Magn Reson Med 2001;45:838–845. 42. Pauly, P., Le Roux, P., Nishimura, D., Macovski, A. Parameter relations for the shinnar-le roux selective excitation pulse design algorithm. IEEE Trans Med Imaging 1991;10:53–65. 43. Madhuranthakam, A. J., Busse, R. F., Brittain, J. H., Rofsky, N. M., Alsop, D. C. B1-insensitive fast spin-echo using adiabatic square wave enabling of the echo train (SWEET) excitation. Magn Reson Med 2008;59:1386–1393. 44. Garwood, M., Ugurbil, K., Rath, A. R., Bendall, M. R., Ross, B. D., Mitchell, S. L., Merkle, H. Magnetic resonance imaging with adiabatic pulses using a single surface coil for RF transmission and signal detection. Magn Reson Med 1989;9:25–34.

45. Insko, E. K., Bolinger, L. Mapping of the radiofrequency field. J Magn Reson Ser A 1993;103:82–85. 46. Jiru, F., Klose, U. Fast 3D radiofrequency field mapping using echo-planar imaging. Magn Reson Med 2006;56:1375–1379. 47. Morrell, G. R. A phase-sensitive method of flip angle mapping. Magn Reson Med 2008;60:889–894. 48. Cunningham, C. H., Pauly, J. M., Nayak, K. S. Saturated double-angle method for rapid B1+ mapping. Magn Reson Med 2006;55:1326–1333. 49. Deoni, S. C. L. High resolution T1 mapping of the brain at 3T with driven equilibrium single pulse observation of T1 with highspeed incorporation of RF field inhomogeneities (DESPOT1-HIFI). J Magn Reson Imaging 2007;26:1106–1111. 50. Katscher, U., Bornert, P. Parallel RF transmission in MRI. Nmr Biomed 2006;19:393–400. 51. Preibisch, C., Deichmann, R. Influence of RF spoiling on the stability and accuracy of T1 mapping based on spoiled FLASH with varying flip angles. Magn Reson Med 2009;61:125–135. 52. Freeman, R., Hill, H. D. W. Phase and intensity anomalies in Fourier transform NMR. J Magn Reson 1971;4:366–383. 53. Zur, Y., Wood, M. L., Neuringer, L. J. Motion-insensitive steady-state free precession imaging. Magn Reson Med 1990;16:444–459. 54. Deoni, S. C. L. Transverse relaxation time (T2) mapping in the brain with off-resonance correction using phase-cycled steady-state free precession imaging. J Magn Reson Imaging 2009;30:411–417. 55. MacKay, A., Laule, C., Vavasour, I., Bjarnason, T., Kolind, S., Madler, B. Insights into brain microstructure from the T2 distribution. Magn Reson Imaging 2006;24:515–525. 56. Kroeker, R. M., Henkelman, R. M. Analysis of biological NMR relaxation data with continuous distributions of relaxation times. J Magn Reson 1986;69:218–235. 57. Menon, R. S., Rusinko, M. S., Allen, P. S. Multiexponential proton relaxation in model cellular systems. Magn Reson Med 1991;20:196–213. 58. Whittal, K. P., MacKay, A. L., Graeb, D. A., Nugent, R. A., Li, D. K., Paty, D. W. In vivo measurement of T2 distributions and water contents in normal human brain. Magn Reson Med 1997;37:34–43. 59. Cheng, K. H. In vivo tissue characterization of human brain by chisquares

Magnetic Resonance Relaxation and Quantitative Measurement in the Brain

60.

61.

62.

63. 64.

65.

66.

67.

68. 69.

70.

parameter maps: Multiparameter proton T2relaxation analysis. Magn Reson Imaging 1994;12:1099–1109. Laule, C., Leung, E., Lis, D. K., Traboulsee, A. L., Paty, D. W. et al. Myelin water imaging in multiple sclerosis: Quantitative comparison with histopathology. Mult Scler 2006;12:747–753. Flynn, S. W., Lang, D. J., MacKay, A. L., Goghari, V., Vavasoir, I. M. et al. Abnormalities of myelination in schizophrenia detected in vivo with MRI, and post-mortem with analysis of oligodendrocyte proteins. Mol Psychiatry 2003;8:811–820. Zimmerman, J. R., Britten, W. E. Nuclear magnetic resonance studies in multiple phase systems: Lifetime of a water molecule in an adsorbing phase on a silica gel. J Phys Chem 1957;61:1328–1333. Poon, C. S., Henkelman, R. M. Practical T2 quantitation for clinical applications. J Magn Reson Imaging 1992;2:541–553. Du, Y. P., Chu, R., Hwang, D., Brown, M. S., Kleinschmidt-DeMasters, B. K. et al. Fast multislice mapping of the myelin water fraction using multicomponent analysis of T2∗ decay at 3t: A preliminary postmortem study. Magn Reson Med 2007;58:865–870. Kreis, R., Fusch, C., Boesch, C. (1992). In vivo characterization of three water compartments in human white matter using a single voxel technique with short TE. In: Proceedings of the 11th Annual Meeting of SMRM, Berlin, Germany, p. 1963. Spencer, R. G. S., Fishbein, K. W. Measurement of spin-lattice relaxation times and concentrations in systems with chemical exchange using the one-pulse sequence: Breakdown of the Ernst model for partial saturation in nuclear magnetic resonance spectroscopy. J Magn Reson 2000;142:120–135. Deoni, S. C. L., Rutt, B. K., Arun, T., Pierpaoli, C., Jones, D. K. Gleaning multicomponent T1 and T2 information from steady-state imaging data. Magn Reson Med 2008;60:1372–1387. Ashburner, J., Friston, K. J. Voxel-based morphometry – the methods. Neuroimage 2000;11:805–821. Tosun, D., Rettmann, M. E., Han, X., Tao, X., Xu, C. et al. Cortical surface segmentation and mapping. Neuroimage 2004;23(S1):S108–S118. Deoni, S. C. L., Williams, S. C. R., Jezzard, P., Suckling, J., Murphy, D. G., Jones, D. K. Standardized structural magnetic resonance imaging in multicentre studies using quantitative T1 and T2 imaging at 1.5T. Neuroimage 2008;40:662–671.

107

71. van Buchem, M. A., McGowan, J. C., Grossman, R. I. Magnetization transfer histogram methodology: Its clinical and neuropsychological correlates. Neurology 1999;53: S23–S28. 72. Jenkinson, M., Bannister, P. R., Brady, J. M. Improved optimization for the robust and fast linear registration and motion correction of brain images. Neuroimage 2002;17: 825–841. 73. Voormolem, E. H., Wei, C., Chow, E. W., Bassett, A. S., Mikulis, D. J., Crawley, A. P. Voxel-based morphometry and automated lobar volumetry: The trade-off between spatial scale and statistical correction. Neuroimage 2009;18 (ahead of print). 74. Catani, M., ffytche, D. H. The rises and falls of disconnection syndromes. Brain 2005;128:2224–2239. 75. Kanaan, R. A., Shergill, S. S., Barker, G. K., Catani, M., Ng, V. W. et al. Tract-specific anisotropy measurements in diffusion tensor imaging. Psychiatry Res 2006;146:73–82. 76. More, S., Oishi, K., Jiang, L., Li, X., Akhter, K. et al. Stereotactic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage 2008;40:570–582. 77. Catani, M., Thiebaut de Schotten, M. A diffusion tensor imaging tractography atlas for virtual in vivo dissections. Cortex 2009;44:1105–1132. 78. Basser, P. J. Inferring microstructural features and the physiological state of tissues from diffusion-weighted images. Nmr Biomed 1995;8:333–344. 79. Jones, D. K., Simmons, A., Williams, S. C., Horsfield, M. A. Non-invasive assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI. Magn Reson Med 1999;42:37–41. 80. Bakshi, R., Thompson, A. J., Rocca, M. A., Pelletier, D., Dousset, V. et al. MRI in multiple sclerosis: Current status and future prospects. Lancet Neurol 2008;7:615–625. 81. Manfredonia, F., Ciccarelli, O., Khaleeli, Z., Tozer, D. J., Saste-Garriga, J. et al. Normal appearing brain T1 relaxation time predicts disability in early primary progressive multiple sclerosis. Arch Neurol 2007;64: 411–415. 82. Wang, Yi. Description of parallel imaging in MRI using multiple coils. Magn Reson Med 2000;44:495–499. 83. Deoni, S. C. L., Catani, M. Visualization of the deep cerebellar nuclei using quantitative T1 and rho magnetic resonance imaging at 3 Tesla. Neuroimage 2007;37:1260–1266. 84. Ffrench-Constant, C. Pathogenesis of multiple sclerosis. Lancet 1994;29:271–275.

108

Deoni

85. Be, L. The histopathology of grey matter demyelination in multiple sclerosis. Acta Neurol Scand Suppl 2009;189:51–57. 86. Gasperini, C., Horsfield, M. A., Thorpe, J. W., Kidd, D., Barkr, G. J. et al. Macroscopic and microscopic assessments of disease burden by MRI in multiple sclerosis: Relationship to clinical parameters. J Magn Reson Imaging 1996;6:580–584. 87. Ropele, S., Strasser-Fuchs, S., Augustin, M. et al. A comparison of magnetization transfer ratio, magnetization transfer rate, and the native relaxation time of water protons related to relapsing-remit- ting multiple sclerosis. AJNR Am J Neuroradiol 2000;21:1885–1889. 88. Laule, C., Vavasour, I. M., Moore, G. R., Oger, J., Li, D. K., Paty, D. W., MacKay, A. L. Water content and myelin water fraction in multiple sclerosis. A T2 relaxation study. J Neurol 2004;251:284–293. 89. Laule, C., Vavasour, I. M., Kolind, S. H., Traboulsee, A. L., Moore, G. R. et al. Long T2 water in multiple sclerosis: What else can we learn from multi-echo T2 relaxation? J Neurol 2007;254:1579–1587. 90. Kolind, S. H., Laule, C., Vavasour, I. M., Li, D. K., Traboulsee, A. L. et al. Complementary information from multi-exponential T2 relaxation and diffusion tensor imaging reveals differences between multiple sclerosis lesions. Neuroimage 2008;40:77–85. 91. DeWitt, L. D., Kistler, J. P., Miller, D. C., Richardson, E. P., Buonanno, F. S. NMRneuropathologic correlation in stroke. Stroke 1987;18:342–351. 92. Kaur, J., Tuor, U. I., Zhao, Z., Peterson, J., Jin, A. Y., Baber, P. A. Quantified T1 as an adjunct to apparent diffusion coefficient for early infarct detection: A high-field magnetic resonance study in a rat stroke model. Int J Stroke 2009;4:159–168. 93. Bernarding, J., Braud, J., Hohmann, J., Mansmann, U., Hoehn-Berlage, M. et al. Histogram-based characterization of healthy and ischemic brain tissues using multiparametric MR imaging including apparent diffusion coefficient maps and relaxometry. Magn Reson Med 2000;43:52–61.

94. Reutens, D. C., Stevens, J. M., Kingsley, D., Kendall, B., Mose- ley, I. et al. Reliability of visual inspection for the detection of volumetric hippocampal asymmetry. Neuroradiology 1996;38:221–225. 95. Jackson, G. D., Connelly, A., Duncan, J. S., Gruenewald, R. A., Gadian, D. G. Detection of hippocampal pathology in intractable partial epilepsy: Increased sensitivity with qualitative magnetic resonance T2 relaxometry. Neurology 1993;43:1793–1799. 96. Conlon, P., Trimble, M. R., Rogers, D., Callicot, C. Magnetic resonance imaging in epilepsy: A controlled study. Epilepsy Reson 1988;2:37–43. 97. Jack, C. R., Jr., Wengenack, T. M., Reyes, D. A., Garwood, M., Curran, G. L. et al. In vivo magnetic resonance microimaging of individual amyloid plaques in Alzheimer’s transgenic mice. J Neurosci 2005;25:10041–10048. 98. Schenck, J. F., Zimmerman, E. A. Highfield magnetic resonance imaging of brain iron: Birth of a biomarker? NMR Biomed 2004;17:433–445. 99. Duan, J. H., Wang, H. Q., Xu, J., Lin, X., Chen, S. Q. et al. White matter damage of patients with Alzheimer’s disease correlated with decreased cognitive function. Surg Radiol Anat 2006;28:150–156. 100. Barkzokis, G., Lu, P. H., Mintz, J. Human brain myelination and amyloid beta deposition in Alzheimer’s disease. Alzheimers Dement 2007;3:122–125. 101. House, M. J., St. Pierre, T. G., Foster, J. K., Matrins, R. N., Clarnette, R. Quantitative MR imaging R2 relaxometry in elderly participants reporting memory loss. AJNR 2006;27:430–439. 102. Akshoomoff, N., Pierce, K., Courchesne, E. The neurobiological basis of autism from a developmental perspective. Dev Psychol 2002;14:613–634. 103. Saito, N., Sakai, O., Ozonoff, A., Jara, H. Relaxo-volumetric multispectral quantitative magnetic resonance imaging of the brain over the human lifespan: Global and regional aging patterns. Magn Reson Imaging 2009;27:895–906.

Chapter 5 Magnetic Resonance Brain Image Processing and Arithmetic with FSL William R. Crum Abstract Medical imaging has been transformed by a move from qualitative to quantitative approaches where image processing is used to enhance visual information and image analysis is used to derive structural and functional measurements. The ideal quantitative analysis methods are automatic and require no user intervention, and so-called image analysis pipelines exist for some applications. However, in the majority of cases automatic methods seldom live up to their name, may fail when prior assumptions are not met, and may not exist at all for new applications. The identification and careful use of well-known image processing and analysis techniques is a vital part of imaging and invaluable when problems arise with automatic methods. Here a number of key image analysis tasks in brain imaging are presented with particular reference to the freely available FMRIB Software Library. Key words: Medical image analysis, FSL, image segmentation, image registration, image arithmetic.

1. Introduction Medical image analysis has changed many aspects of clinical research and is finding application in clinical practice. Acquisition and analysis of single images in isolation are useful for answering very specific questions, e.g. “How big is this tumour?” “What is the lesion load?” “How big is the hippocampus?” Derived measurements such as lengths, areas, volumes, shape measures, and texture (1) can be analysed within and across groups and have been the basis of many successful scientific studies. Computer processing-based applications in neuroimaging, particularly in group-based neuroscience research studies, M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_5, © Springer Science+Business Media, LLC 2011

109

110

Crum

are amongst the best developed and fastest evolving uses of image analysis in medicine. Examples include Statistical Parametric Mapping (2) (http://www.fil.ion.ucl.ac.uk/spm), FreeSurfer (http://surfer.nmr.mgh.harvard.edu) (3), and the FMRIB Software Library (http://www.fmrib.ox.ac.uk/fsl) (4). With the growth in methods and applications comes a potentially bewildering collection of algorithms and tools to choose from. Some common analysis tasks benefit from automated approaches, particularly in studies involving large numbers of scans. However, automated analysis techniques can often fail when assumptions are broken (for instance when an atypical MR imaging sequence is used or when a study group has unusual brain structure or function) or image preparation (preprocessing) is not performed adequately. Specific applications or research studies can require preprocessing, measurements, or analysis steps not anticipated in automated pipelines. Therefore there remains a fundamental need for user-driven image processing, and analysis algorithms which are accessible to the non-specialist can be tuned for new applications and which when used effectively can ensure the maximum use is made of precious images. In this chapter we describe how to successfully apply some of the most common and useful brain image processing analysis steps using freely available software. We will separately consider noise reduction, automated skull and scalp removal, brain tissue classification, brain image registration, and image arithmetic. Finally we will show how simple techniques can be combined into sophisticated analysis pipelines by showing how to implement the Brain Boundary Shift Integral algorithm for brain atrophy measurement.

2. Materials 2.1. Software

There are many software packages available which can perform some or all of the tasks described in this chapter. We focus particularly on the FMRIB Software Library (FSL) (4, 5) as it is freely available for non-commercial use, extremely well supported via a well-supported email discussion forum (www.jiscmail.ac.uk/lists/fsl.html), and does not depend on other non-free software typically used as a platform for medical image analysis (e.g. MATLAB). Readers should note that FSL has many more tools providing much more functionality than we can cover in this chapter.

2.1.1. FSL System Requirements

FSL (the FMRIB Software Library) is developed and made available (http://www.fmrib.ox.ac.uk/fsl/) by the Oxford Centre

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

111

for Functional MRI of the Brain (http://www.fmrib.ox.ac.uk/) licensed (http://www.fmrib.ox.ac.uk/fsl/fsl/licence.html) via Isis Innovation (http://www.isis-innovation.com/). It is a collection of computer programs and scripts which provide access to stand-alone processing tools and more sophisticated analysis pipelines. FSL is hardware compatible with PC, Mac, Sun, Silicon Graphics. Memory requirements will vary depending on the analysis and image and group data sizes, but 512 MB minimum available RAM is required (see Note 1). Swap space of at least 2 GB is recommended and should always at least equal the available RAM. Disk space of 10 times the size of the images to be analysed is recommended. The recommended operating system is Linux. Precompiled binaries are available for Centos (www.centos.org) and Debian (www.debian.org) Linux flavours. On Windows PC XP/Vista, Linux FSL can be run using VMware (www.vmware.com) by following the instructions here (http://www.fmrib.ox.ac.uk/fsl/fsl/windows.html). Source code is available for compiling on other operating systems. To validate the installation, an associated test suite of data and scripts, the FSL Evaluation and Example Data Suite (FEEDS) (http://www.fmrib.ox.ac.uk/fsl/feeds/doc/), should be downloaded and applied after a new installation. 2.1.2. FSL Tools

susan – non-linear noise reduction bet – brain extraction tool fast – brain tissue classification flirt – FMRIB’s Linear Image Registration Tool fslmaths – image arithmetic and operations fslstats – report summary intensity statistics All are part of the standard installation.

2.2. Example Image Data

We use the FSL Evaluation and Example Data Suite (FEEDS) (http://www.fmrib.ox.ac.uk/fsl/feeds/doc/) for most of the examples in this chapter (see Note 2).

3. Methods Analysis of MRI brain images involves some standard tasks which are independent of scanner and sequence type such as noise reduction, brain extraction, tissue classification, image registration, and image statistics. Generic algorithms have been developed which successfully tackle these problems. However, most of

112

Crum

these algorithms require parameters to be specified. In some cases default parameters are suitable for a wide range of images and applications. In other cases, careful tuning of parameters on a set of test data will be required. 3.1. Software Usage

The methods in this chapter use the command-line versions of the software which are invoked by typing commands into a terminal window. This form facilitates batch processing by including a series of commands in a shell script in Linux or other Unix variant. To see the format and allowed parameters for each command simply type its name without any supplied arguments or parameters. Where the software can also be launched with a graphical user interface which allows user interaction with a computer mouse this is indicated in the text.

3.2. Noise Reduction with Susan 3.2.1. Basic Usage and Parameters

The task is to reduce the appearance of noise in the image while preserving intensity gradients and region boundaries. See also Note 3. susan

<use_median>

= image to be de-noised = brightness threshold

= spatial scale of de-noising (mm) = planar (2) or volumetric (3) de-noising <use_median> = do (0) or don’t (1) apply median filter to point noise = 0

= de-noised image

3.2.2. Example Usage

To denoise the image called structural.nii.gz (see Notes 4 and 5) in 3D with brightness threshold = 2,000, spatial scale = 2.0 mm, without using the median filter and producing a new de-noised image called structural-denoised.nii.gz use susan structural.nii.gz 10 2.0 3 0 0 structural-denoised.nii.gz

3.2.3. Parameter Setting

Figure 5.1 shows an example calibration run where the brightness threshold and spatial scale parameters were varied for different amounts of added Rician noise (6) (Note 6). See also Note 7 regarding computational requirements.

3.2.4. GUI Version

Susan

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

113

Fig. 5.1. Examples of applying susan for noise reduction of the image structural.nii.gz while varying the brightness threshold, bt, and the spatial scale, dt. Clockwise from top left: original, best result (bt = 2,000, dt = 2), brightness threshold too high (bt = 4,000, dt = 2), brightness threshold too high and spatial scale too large (bt = 4,000, dt = 4).

3.3. Brain Extraction Using Bet 3.3.1. Basic Usage

The task is to start with a brain scan which includes whole head and neck and produce a brain image with non-brain tissues such as neck, scalp, eyes, etc., removed (7). bet

= original image

= original image with non-brain tissues removed

3.3.2. Example Usage

To remove non-brain tissues from structural.nii.gz and generate a new image called structural_bet.nii.gz use bet structural.nii.gz structural_bet.nii.gz To also generate a brain mask called structural_bet_mask. nii.gz suitable for image registration use bet structural.nii.gz structural_bet.nii.gz -m

3.3.3. Advanced Usage

There are several parameters which are used to customise bet for non-standard images. bet

-c -f <x y z > -g -r -c <x y z> = coordinates of “centre of gravity” (in voxels) of initial brain mesh

114

Crum

-f = fractional intensity threshold (range [0, 1], default = 0.5). Smaller values of f result in larger brain outline estimates -g = vertical gradient in fractional intensity threshold (range [–1, +1], default = 0). Positive (negative) values give larger (smaller) superior brain outline and smaller (larger) inferior brain outline. -r = head radius (in mm) which is used to set the size of the initial brain mesh 3.3.4. Example Advanced Usage

To remove non-brain tissues from structural.nii.gz and generate a new image called structural_bet.nii.gz while encouraging a larger brain outline with large initial estimate of head radius use bet structural.nii.gz structural_bet.nii.gz –f 0.7 -r 250.0

3.3.5. Parameter Setting

Figure 5.2 shows the use of bet and the different representations of the extracted brain.

Fig. 5.2. The different output options available for brain extraction of structural.nii.gz using bet. Clockwise from top left: original image, default BET output, original overlaid with binary mask, overlaid BET outline.

3.3.6. GUI Version

Bet

3.4. Brain Tissue Classification Using Fast 3.4.1. Basic Usage and Prerequisites

The task is to produce a brain image where voxels which are predominantly grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) are set to indicative values (background = 0,

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

115

CSF = 1, GM = 2, WM = 3) (8). To ensure non-brain voxels with similar intensities are not classified as brain tissue, brain extraction (Section 3.3) should be applied before brain tissue classification. Brain tissue classification is by default integrated with a correction for MRI intensity inhomogeneity (9). fast --nopve -o

--nopve prevents additional partial volume estimation being performed = brain-extracted image (see Note 8)

= base-name for output classification images 3.4.2. Example Usage

To create a simple brain tissue classification called structural_fast_seg.nii.gz from structural_bet.nii.gz use fast --nopve -o structural_fast structural_bet.nii.gz To additionally create a partial volume estimated brain tissue classification called structural_fast_pveseg.nii.gz (see also Note 9) use fast -o structural_fast structural_bet.nii.gz

3.4.3. Advanced Usage

There are several parameters which can be used to generate more detailed information. fast -B -o

-B additionally outputs an intensity inhomogeneity-corrected image called

_restore.nii.gz Multi-channel data can also be classified provided it is preregistered (see Note 10). fast -S -o

. . . -S = number of supplied image channels (default = 1) The tissue classification can be made spatially smoother by increasing the -H and -R parameters. fast -H -R -o

-H = initial segmentation spatial smoothness (default = 0.1) -R = PVE spatial smoothness (default = 0.3)

3.4.4. Example Advanced Usage

To perform tissue classification using partial volume estimation on registered T1-weighted and T2-weighted brain-extracted image volumes, outputting the intensity inhomogeneity-corrected image and forcing a smooth PVE segmentation use fast -S 2 -B -H 0.4 -o image_fast image_T1_bet.nii.gz image_ T2_bet.nii.gz

116

Crum

The resulting images will be called image_fast_seg.nii.gz, image_fast_pveseg.nii.gz, image_fast_pve_0.nii.gz, image_ fast_pve_1.nii.gz, image_fast_pve_2.nii.gz and image_fast_ restore.nii.gz. 3.4.5. Classification Types

Figure 5.3 shows examples of the images generated by fast.

Fig. 5.3. The different output options available for brain tissue classification with fast. Clockwise from top left: original structural image, standard fast segmentation overlaid on original, partial volume estimation (PVE) fast segmentation, overlaid on original, the estimated intensity bias field, the mixel map showing the different tissue compartments used in the analysis, the original image corrected for intensity bias.

3.4.6. GUI Version

Fast

3.5. Brain Image Registration Using Flirt 3.5.1. Basic Usage and Prerequisites

The task is to determine and apply an affine coordinate transformation which maps corresponding features (10) from an input brain scan onto a reference scan (11) (see Note 11).

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

117

flirt –in -ref -out

= brain scan to be transformed = reference brain scan

= transformed image 3.5.2. Example Usage

To register image.nii.gz with reference.nii.gz optimising the match over the reference brain (-refweight) (Note 12) and outputting a copy of the transformation matrix (-omat) use flirt -refweight reference_weight.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To correct only for positional differences (three rotations and three scalings) between the brains use flirt -dof 6 -refweight reference_bet_mask.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To correct for positional differences and global scalings (Note 13) between the brains use flirt -dof 9 -refweight reference_bet_mask.nii.gz -in image.nii.gz -ref reference.nii.gz -out image_reg –omat matrix.mat To apply an existing registration result to an additional image, image_add.nii.gz, use flirt -in image_add.nii.gz -ref reference.nii.gz -init matrix.mat -applyxfm -out image_add_reg.nii.gz

3.5.3. Advanced Usage

There are several parameters which affect the way the registration is optimised. -nosearch switch off the global rotational optimisation -init matrix_init.mat = supply an initial estimate of the transformation matrix -searchrx <min_angle> <max_angle> = specify angular search range around the x-axis in degrees (default is –90 90). Also –searchry and –searchrz. -coarsesearch <delta_angle> = specify coarse angular search increment in degrees (default is 60) -finesearch <delta_angle> = specify coarse angular search increment in degrees (default is 18) There are several parameters which affect the way the registration is assessed. -cost costfn = specify function used to assess registration, (default is corratio) -searchcost costfn (default is corratio)

118

Crum

In both cases, costfn should be one of (mutualinfo, corratio, normcorr, normmi, leastsq, labeldiff) (see Note 14). 3.5.4. Example Advanced Usage

To register two images with weighting images applied to both using normalised mutual information as the cost function and restricting the angular search around the z-axis use flirt -dof 6 –searchz 0 60 –cost normmi -searchcost normmi –inweight iweight.nii.gz -refweight rweight.nii.gz –in input.nii.gz -ref reference.nii.gz -out output.nii.gz -omat output.mat The resulting image will be called output.nii.gz with an associated rigid body transformation matrix called output.mat.

3.5.5. Registration Example

Figure 5.4 shows examples of images registered using flirt.

Fig. 5.4. Example of affine registration of T1 and T2 axial images with flirt. Top row: left = T1-weighted image, middle = misregistered T2-weighted image, right = registered T2-weighted image. The registration is relatively subtle with some in-plane positional displacement and some through-plane misregistration visible around the left medial cortex. Bottom row: left = overlay of misregistered T2-weighted and T1-weighted images, right = overlay of registered T2-weighted and T1-weighted images. Misregistration is particularly evident around the ventricles where light and dark boundaries show the regions of mismatch and around the cortex. These misregistrations are clearly resolved on the overlaid registered images.

3.5.6. GUI Version

Flirt

3.6. Image Operations, Statistics and Arithmetic 3.6.1. Basic Usage and Prerequisites

The FMRIB Software Library provides some general purpose utilities such as fslmaths (image arithmetic and other operations) and fslstats (image statistics) which, in conjunction with bet, fast,

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

119

flirt, enable sophisticated image analysis to be performed. This is demonstrated in Section 3 by showing how fslmaths and fslstats can be used to implement the Brain Boundary Shift Integral (12, 13) algorithm for measuring brain atrophy in longitudinal imaging. First some common analysis steps involving brain tissue masks, arithmetic operations on pairs of images and image intensity histogram construction are detailed. fslmaths [operations]

[operations] are typically of the form where data is a number or another image fslstats [options] [options] are typically single letter arguments, e.g. “-m” outputs the mean intensity 3.6.2. Example Usage

To create a binary GM mask (where non-GM voxels have intensity 0 and GM voxels have intensity 1) from the output of fast (see Notes 15–17) use fslmaths structural_fast_pveseg.nii.gz -thr 2 -uthr 2 –div 2 structural_gm_mask.nii.gz To create a binary brain mask (where non-brain voxels have intensity 0 and brain voxels have intensity 1) from the output of fast (see Note 15, 16) use fslmaths structural_fast_pveseg.nii.gz -thr 2 -min 1 structural_brain_mask.nii.gz To create a subtraction image, image12_diff.nii,gz, highlighting differences between two images, image1.nii.gz and image2.nii.gz (see Notes 18, 19), use fslmaths image1.nii.gz diff.nii,gz

-sub

image2.nii.gz

image12_

To compute the entries of a 64-bin intensity histogram for structural.nii.gz (Fig. 5.5) use fslstats structural.nii.gz -h 64 3.6.3. Advanced Usage

Computing the Brain Boundary Shift Integral, a measure of brain volume change which has occurred in the time between two scans image1.nii.gz and image2.nii.gz being acquired. See Fig. 5.6 for a flow chart of the main image processing operations. First perform brain extraction using bet (Section 3.3) generating image1_bet.nii.gz, image1_bet_mask.nii.gz, image2_bet. nii.gz, image2_bet_mask.nii.gz. Perform tissue classification using fast (Section 3.4) on image1_bet.nii.gz and image2_bet.nii.gz. The images generated are image1_fast_pveseg.nii.gz and image2_fast_pveseg.nii.gz.

120

Crum

Fig. 5.5. A graph showing the intensity histogram with 64 bins for the example image structural.nii.gz. The data were generated from the command fslstats structural.nii.gz –h 64. Note that for illustrative purposes the log of the entries in each intensity bin is plotted.

Generate a binary brain mask from each pveseg image generated in step 3b using fslmaths (Section 3.6, Section 2b) producing image1_brain_mask.nii.gz and image2_brain_mask.nii.gz. Generate a combined dilated (expanded) brain and CSF mask of the first scan for registration with foreground/background intensity ratio = 250/1. fslmaths image1_bet_mask.nii.gz –dilM –mul 249 –add 1 image1_bet_mask_weight.nii.gz Register image2.nii.gz to image1.nii.gz as in Section 3.5 (using the option -refweight image1_brain_mask_weight.nii.gz) to produce image2_reg.nii.gz and matrix.mat. Use flirt with the -interp nearestneighbour option to apply the registration result to the corresponding brain mask to produce image2_brain_mask_reg.nii.gz (Section 3.5.2) (Note 20). Compute the intersection of the two registered brain masks (Note 21). fslmaths image1_brain_mask.nii.gz –mul fslmaths image2_ brain_mask_reg.nii.gz brain_mask_int.nii.gz Compute the union of the two registered brain masks (Note 22). fslmaths image1_brain_mask.nii.gz -add fslmaths image2_ brain_mask_reg_thresh.nii.gz –min 1 brain_mask_uni. nii.gz Apply binary erosion to the intersection brain mask and binary dilation to the union brain mask:

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

121

Fig. 5.6. A schematic showing the image processing steps which generate the brain– CSF boundary mask for the Brain Boundary Shift Integral calculation. (i) Scan 2 is registered to scan 1 and transformed to give scan 2 , (ii) the brain is extracted from scan 1 and scan 2 using bet, (iii) brain tissue classification is performed on scan 1 and scan 2 using fast, (iv) the fast tissue maps are manipulated using fslmaths to produce brain masks, and the brain mask on scan 2 is transformed using the registration result giving mask 2 , (v) the registered brain masks are combined using fslmaths to produce the intersection (inner) and union (outer) brain masks, (vi) the inner brain mask is eroded and the outer brain mask is dilated, both using fslmaths, (vii) the brain boundary mask is the difference between the dilated outer mask and the eroded inner mask, again computed using fslmaths. See text for a full description of all these steps.

fslmaths brain_mask_uni.nii.gz -dilM brain_mask_uni_ d.nii.gz fslmaths brain_mask_int.nii.gz -ero brain_mask_int_e.nii.gz Generate a mask covering the boundary of brain and CSF by computing the difference between the dilated union and the eroded intersection images (Note 23). fslmaths brain_mask_uni_d.nii.gz -sub brain_mask_int_e. nii.gz image_csf_bnd.nii.gz Normalise the original images over the internal region so the mean intensity is 100: #!/bin/csh # See Note 24. # First, compute and store the mean brain intensity

122

Crum

set mean1 = `fslstats image1.nii.gz -k brain_mask_int_e.nii.gz –m` set mean2 = `fslstats image2_reg.nii.gz -k brain_mask_int_ e.nii.gz –m` # Then normalise the image intensities to have average intensity = 100 fslmaths image1.nii.gz –div ${mean1} -mul 100 image1_ norm.nii.gz fslmaths image2_reg.nii.gz –div ${mean2} -mul 100 image2_ reg_norm.nii.gz Clip the maximum and minimum normalised intensities to an appropriate window (Note 25) and compute the difference image. fslmaths image1_norm.nii.gz -max 25 -min 75 image1_ norm_clip.nii.gz fslmaths image2_reg_norm.nii.gz -max 25 -min 75 image2_ reg_norm_clip.nii.gz fslmaths image1_norm.nii.gz –sub image2_reg_norm.nii.gz image12_sub.nii.gz Finally compute the boundary shift integral by summing the clipped intensity difference across the brain boundary (see Fig. 5.7).

Fig. 5.7. The final stages of the Brain Boundary Shift Integral calculation. Left: the normalised subtraction image of two scans of a subject showing bright areas of atrophy in brain. Middle: a mask of the brain–CSF boundary. Right: the mask overlaid on the subtraction image showing the parts of the subtraction image which contribute to the integral.

#!/bin/csh set bbsi = `fslstats image12_sub.nii.gz -k brain_csf_bnd.nii.gz –V -M` set pix = `fslinfo image12_sub.nii.gz | egrep pixdim` # See Note 26 concerning then next line set bbsiml = `echo scale=2\; \($bbsi[3] ∗ $bbsi[1] ∗ $pix[2] ∗ $pix[4] ∗ $pix[6] \)/\(1000 ∗ 50 \) | bc` echo BBSI is ${bbsiml} ml

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

123

4. Notes 1. A “512 MB computer” will have significantly less than 512 MB RAM available for analysis because of memory required by the operating system and other processes. 2. Images may be viewed using another FSL program called fslview. The general use is fslview image1.nii.gz image2.nii.gz. . . imagen.nii.gz. 3. Susan is most useful for visualisation or for subsequent manually driven measurements (e.g. manual segmentation). Many automated analysis techniques incorporate a noise model, and susan should then not be applied as it may change the assumed noise properties of the image. 4. The image called structural.nii.gz is a three-dimensional structural MRI brain volume supplied with the FEEDS testing and evaluation suite. 5. We assume images are stored in the NIfTI-1format (http://nifti.nimh.nih.gov) and then compressed using the common Unix tool gzip. Thus structural.nii indicates an image called structural stored in the NIfTI format and structural.nii.gz is obtained by issuing the command gzip structural.nii on Unix/Linux systems. Note that the FSL software reads compressed files transparently, so there is usually no need for the user to compress or uncompress (using gunzip) images by hand. 6. The figures show single slices for illustrative purposes, but it should be remembered that MR images are usually comprised of multiple slices which either are collected to be approximately contiguous or are reconstructed to be contiguous from a 3D data volume. The image analysis techniques in this book operate on all slices in the volume unless otherwise specified. 7. The computational time for noise reduction with susan rises significantly with increasing values of the spatial scale parameter

. 8. The brain-extracted image supplied to fast is assumed to be T1 weighted. When a T2 or proton density-weighted image is used the –t option specifies the image type. fast –t 1 –o image_T1_fast image_T1_bet.nii.gz fast –t 2 –o image_T2_fast image_T2_bet.nii.gz fast –t 3 –o image_PD_fast image_PD_bet.nii.gz 9. By default, as well as generating a partial volume segmented image, fast also outputs individual partial volume estimates of each tissue class. For an output base name specified as

124

Crum

-o image_fast, the default output images are image_fast_ seg.nii.gz, image_fast_pveseg.nii.gz, image_fast_pve_0.nii. gz, image_fast_pve_1.nii.gz and image_fast_pve_2.nii.gz. In principle these individual pve images, which contain fractional estimates of tissue occupancy per voxel, can be used for more accurate tissue volume estimation. 10. Multi-channel data can produce more robust tissue classification. The classic MRI example is using T1-weighted images in conjunction with T2-weighted images. However, post acquisition, these images will not generally be aligned and usually have different voxel dimensions and/or fields of view. Therefore a registration and resampling procedure must be performed to ensure that the T2-weighted brain image is aligned with the T1-weighted image and has the same voxel dimensions; flirt can be used for this purpose. 11. Brain image registration is almost never performed without other options because by default, no distinction is made between foreground and background voxels or between brain and non-brain voxels. The optimum registration computed for a T1-weighted volume including air, neck, scalp, etc., in addition to brain is highly unlikely to produce a good alignment of the brain, even when the scans are of the same person, because of differences in shape and position of the different tissues across scans. 12. When using weight images, it is tempting to set the background weight to 0, e.g. by simply using the binary mask output by bet as a weighting image. However, this can destabilise the registration by removing too much information about surrounding tissues. It is far better to use a weights image with a high foreground value and a low background value. The simplest way to accomplish this starting with a binary brain mask is to use fslmaths to scale and add a small constant value (e.g. 1) to every point in the input mask (binary_mask.nii.gz) to produce a suitable weights_image.nii.gz: fslmaths binary_mask.nii.gz –mul 99 –add 1 weights_image.nii.gz. 13. “dof” stands for degrees of freedom and specifies the number of independent parameters in the affine transformation matrix. The default is 12 corresponding to 3 translations, 3 rotations, 3 scalings and 3 skews. Other commonly used dofs are six (three translations, three rotations, often called “rigid body”) and nine (three translations, three rotations and three scalings). 14. When the image intensity characteristics are not well characterised (the usual scenario unfortunately) it is safest to choose costfn as normmi (normalised mutual information) or corratio (correlation ratio) which assume an

Magnetic Resonance Brain Image Processing and Arithmetic with FSL

125

unknown probabilistic or functional relationship between voxel intensities in the two images, respectively. When the images to be registered are known to have very similar intensity characteristics (i.e. they are structurally similar and were collected in the same way on the same scanner) then costfn can be normcorr (normalised correlation) which assumes a linear relationship between voxel intensities; this can be more robust but is also more susceptible to deviations from those assumptions. 15. The standard segmentation of a T1-weighted volume output by fast has voxel intensities 0 (background), 1 (CSF), 2 (GM) and 3 (WM). 16. Operations are executed in order. –thr 2 sets any voxels of intensity less than 2 to 0. –uthr 2 sets any voxels of intensity greater than 2 to 0. –div 2 divides each voxel intensity by 2. The result is to leave voxels which were originally of intensity 2 (corresponding to GM) set to 1 and all other voxels set to 0. Other operations include –min x (replace each voxel with the lower of its original intensity or x) and –max x (replace each voxel with the higher of its original intensity or x). 17. Substitute “3” for “2” to get a WM mask instead of a GM mask. 18. In brain imaging scans would typically first be aligned using flirt and then intensity normalised as in the previous example before creating the difference image. 19. Images can also be added, multiplied or divided by replacing –sub with –add, –mul or –div, respectively. 20. During registration and image transformation it is necessary to interpolate one image to match the voxel boundaries of another. The –interp option allows the intensity interpolation method to be specified. In order of increasing computing time and accuracy the options are nearestneighbour, trilinear and sinc; trilinear is sufficient unless very high accuracy is required and nearestneighbour is most often used to interpolate binary region masks. 21. The intersection of two binary voxels is equivalent to a logical AND (i.e. both voxels must be set). Therefore a simple implementation is to multiply voxels together – if either is 0 the result will be 0. 22. The union of two binary voxels is equivalent to a logical OR (i.e. either voxel is set). Therefore a simple implementation is to add voxels together and then reset positive results to 1. 23. To expand the region, add more –dilM and –ero terms, respectively, to the first two fslmaths commands.

126

Crum

24. These commands make use of Unix csh (“c-shell”) syntax. They should be put in a data file called normalise-script and saved to the current directory. Then do chmod +x ./normalise-script to make the script executable. Then do ./normalise-script to run the commands. 25. The intensity window limits of 25 and 75 follow the original reference with width = centre = 50 for intensities scaled by a factor of 100. The optimum values will be application dependent and can be determined using procedures outlined in (12, 13). 26. This line converts the Boundary Shift Integral into sensible units using the inbuilt Unix calculator bc to two decimal places (scale = 2). It computes the raw integral equal to mean masked intensity ($bbsi[3]) times the number of voxels ($bbsi[1]) converted to mm3 by multiplying by the voxel dimensions $pixdim[1], $pixdim[2], $pixdim[3]. The integral must be normalised by dividing by the intensity window width (50) and converted to ml (1,000). References 1. Crum, W. R. Shape and texture. Quantitative MRI of the Brain: Measuring Changes Caused by Disease, P. Tofts, (ed.) Chichester: Wiley; 2004, pp. 559–579. 2. Friston, K. J., Ashburner, J. T., Kiebel, S. J., Nichols, T. E., Penny, W. D., (eds.). Statistical Parametric Mapping: The Analysis of Functional Brain Images. Amsterdam: Academic Press; 2007. 3. Fischl, B., Liu, A., Dale, A. M. Automated manifold surgery: Constructing geometrically accurate and topologically correct models of the human cerebral cortex. IEEE Trans Med Imaging 2001;20(1):70–80. 4. Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E. J., Johansen-Berg, H., Bannister, P. R., De Luca, M., Drobnjak, I., Flitney, D. E., Niazy, R., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J. M., Matthews, P. M. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 2004;23(S1):208–219. 5. Woolrich, M. W., Jbabdi, S., Patenaude, B., Chappell, M., Makni, S., Behrens, T., Beckmann, C., Jenkinson, M., Smith, S. M. Bayesian analysis of neuroimaging data in FSL. Neuroimage 2009;45:S173–S186. 6. Gudbjartsson, H., Patz, S. The Rician distribution of noisy MRI data. Magn Reson Med 1995;34(6):910–914.

7. Smith, S. M. Fast robust automated brain extraction. Hum Brain Mapp 2002;17(3):143–155. 8. Zhang, Y., Brady, M., Smith, S. Segmentation of brain MR images through a hidden Markov random field model and the expectation maximization algorithm. IEEE Trans Med Imaging 2001;20(1): 45–57. 9. Vovk, U., Pernus, F., Likar, B. A review of methods for correction of intensity inhomogeneity in MRI. IEEE Trans Image Process 2007;26(3):405–421. 10. Crum, W. R., Hartkens, T., Hill, D. L. G. Non-rigid registration, theory and practice. Br J Radiol 2004;77: S140–S153. 11. Jenkinson, M., Smith, S. M. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5(2):143–156. 12. Fox, N. C., Freeborough, P. A. Brain atrophy progression measured from registered serial MRI: Validation and application to Alzheimer’s disease. J Magn Reson Imaging 1997;7:1069–1075. 13. Freeborough, P. A., Fox, N. C. The boundary shift integral: An accurate and robust measure of cerebral volume changes from registered repeat MRI. IEEE Trans Med Imaging 1997;16(5):623–629.

Chapter 6 Diffusion Tensor Imaging Derek K. Jones and Alexander Leemans Abstract Diffusion tensor MRI (DT-MRI) is the only non-invasive method for characterising the microstructural organization of tissue in vivo. Generating parametric maps that help to visualise different aspects of the tissue microstructure (mean diffusivity, tissue anisotropy and dominant fibre orientation) involves a number of steps from deciding on the optimal acquisition parameters on the scanner, collecting the data, pre-processing the data and fitting the model to generating final parametric maps for entry into statistical data analysis. Here, we describe an entire protocol that we have used on over 400 subjects with great success in our laboratory. In the ‘Notes’ section, we justify our choice of the various parameters/choices along the way so that the reader may adapt/modify the protocol to their own time/hardware constraints. Key words: Diffusion tensor, MRI, sampling schemes, pulse sequence, optimal, data quality.

1. Introduction Diffusion tensor MRI (DT-MRI), developed in the early- to mid1990s (1, 2), provides a means for non-invasively characterising the properties of soft tissue on a microstructural scale. It works by sensitising the MRI signal to the random molecular motion of water molecules (diffusion) by addition of ‘diffusionencoding gradients’ to a standard MR pulse sequence (3). At the typical resolution of a DT-MRI experiment (2–3 mm voxel sizes), in the grey matter and cerebro-spinal fluid, the diffusionweighted signal is independent of the direction in which the gradients are applied, and the diffusion appears to be isotropic (2). In white matter, water molecules diffuse more freely along the dominant fibre orientation than across them (4). This anisotropy of diffusion provides insights into the microstructural organisation M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_6, © Springer Science+Business Media, LLC 2011

127

128

Jones and Leemans

of the white matter. The simplest model that encapsulates this anisotropic diffusion is the diffusion tensor (1). By applying diffusion-encoding gradients in at least six non-collinear and noncoplanar orientations, one can estimate the six unknown elements of the diffusion tensor (1) and thus characterise the anisotropy. Further, the direction in which the diffusion-weighted signal has the greatest attenuation gives an indication of the dominant fibre orientation – which can be used to create voxel-wise maps of fibre orientation (e.g. 5) or pieced together to reconstruct continuous trajectories throughout the white matter (i.e. ‘tractography’) (e.g. 6–10). Such information has previously been unavailable in vivo – and so it is understandable that the technique has attracted huge interest from, and has enjoyed rapid uptake by, clinical and neuroscientific communities. Here we present a protocol that begins by bringing the subject to the scanner room and preparing them for the diffusion tensor imaging data acquisition, through the acquisition of the data and pre-processing, through to estimation of the diffusion tensor in each voxel and subsequent computation of quantitative parametric maps. We stress the importance of checking the data at each stage of the pipeline to ensure that the data are robust. It should be noted that what is ‘optimal’ for one laboratory – which can only allocate 10 min to DT-MRI on a 1.5 T system made by manufacturer X, for example, will be of little interest to the group that can allocate 30 min on a 3 T system made by manufacturer Y. Therefore, in the Notes section, details will be provided on how to proceed in choosing what is optimal for a given set of circumstances.

2. Materials MRI Scanner: General Electric HDx 3.0 T system (see Note 1). a. Gradients: Twin-speed gradient system with gradient strength = 40 mT/m and maximum slew rate = 150 T/m/s (see Note 2). b. RF Coils: Whole-body birdcage coil used for RF transmit; eight-channel head coil (made by MRI Devices Corp.) used for RF receive (see Note 3). c. Scanner Software Capability: Software to provide diffusion tensor imaging capability (see Note 4). d. Peripherals: Adequate padding for the head (wedge cushions, etc.); hearing protection (ear plugs); a peripheral pulse-oximeter; a squeeze-bulb (for the participant to communicate to the operator).

Diffusion Tensor Imaging

129

3. Methods 1. Preparing the Participant: After successful completion of standard MR screening and appropriate informed consent forms, the participant is led to the MR magnet room. They are given earplugs before being placed onto the scanner bed in the supine position. The pulse-oximeter is then placed onto the subject’s forefinger, and they are given the squeezebulb. We take special care to warn the participant that the diffusion tensor imaging part of the protocol is ‘louder than the other scans’ and that they ‘can expect the bed to vibrate quite a lot’ (see Note 5). We also warn the participant that ‘there will be irregularly timed knocking noises – and these will appear to move about as the scan progresses. This is fully expected’ (see Note 5). 2. Scanning: The integrated laser alignment system is used to landmark on the nasion, and the participant slid into the magnet, taking particular care not to trap the squeeze-bulb/pulse-oximeter leads during the process. As an optional extra, we provide the participant with the option of watching a subtitled movie of their choice in the scanner via a rear projection onto a periscope mounted on the head coil (see Note 6). The sequence is a twice-refocused spin-echo EPI sequence (11) (see Note 7), with a parallel imaging (ASSET) factor of 2 (see Note 8). Sixty axially oriented slices are prescribed to cover the entire head (see Note 9). The field of view is 230 mm, with an acquisition matrix of 96 × 96 and a slice thickness of 2.4 mm (see Note 10). A total of 66 images are acquired (see Note 11) at each of 60 slice locations. Six images are acquired with no diffusion-weighting gradients applied, and 60 diffusion-weighted images are acquired at a b-value of 1,000 s/mm2 (see Note 12). The diffusion-weighted images are acquired with encoding gradients applied along 60 non-collinear directions (see Table 6.1 for the configurations – and see Note 13). The echo time is 87 ms (see Note 14), and the sequence is triggered to the cardiac cycle via a pulse-oximeter placed on the participant’s forefinger (see Note 15). We select a minimal trigger delay (see Note 16), and the effective TR that is set depends on the heart rate estimated from the pulse-oximeter trace. Each image is initially stored in DICOM format. We then convert the separate DICOM images into a 4D data set (with ‘time’ or ‘diffusion-weighted measurement’ as the fourth dimension) in the NIFTI imaging format.

130

Jones and Leemans

Table 6.1 60 Electrostatically arranged sampling vectors (21) optimally ordered according to Cook et al. (24) 0.1706

0.3255

0.9300

0.3933

0.9154

−0.0853

−0.6263

−0.2862

0.7252

−0.6058

0.7763

0.1742

0.9204

−0.0040

0.3911

0.2903

−0.6194

0.7294

−0.7022

0.3532

0.6182

0.3712

0.7702

0.5187

−0.9738

−0.0801

0.2126

0.2158

−0.9371

0.2743

−0.0604

−0.2236

0.9728

0.8147

0.5611

0.1462

−0.2150

0.7211

0.6587

0.7166

−0.5846

0.3804

0.5354

−0.1672

0.8279

−0.2157

−0.7327

0.6454

0.6549

0.3557

0.6668

−0.0601

0.9726

0.2244

−0.7389

−0.5195

0.4290

−0.9432

0.3042

0.1338

−0.4428

0.2032

0.8733

−0.1629

−0.9173

0.3634

0.0988

0.6380

0.7637

0.7879

−0.2745

0.5513

−0.3122

−0.4225

0.8509

3. Initial Quality Check: The data are loaded into FSLview (part of the FSL software package) (www.fmrib.ox.ac.uk/fsl) and viewed in all three planes simultaneously – in ‘cine’ mode – looping through the separate volumes to check for any obvious artefacts in the data (see Fig. 6.1a and Note 17). 4. Correcting Motion/Distortion: We register the 2nd through 66th diffusion-weighted volumes to the 1st diffusionweighted volume using a global, 12 degrees of freedom affine deformation, with normalised mutual information as the cost function (see Note 18). We extract the rotational component of the transformation for each volume and apply this rotation to the gradient encoding tables (see Note 19).

Diffusion Tensor Imaging

131

Fig. 6.1. Example of the data quality assessment through visual inspection. a Visualisation of the raw diffusion-weighted data. This is done as a cine loop to quickly view all diffusion-weighted volumes. Abnormal data values, such as signal dropouts (i) or hyperintensities (ii), are easily detectable on the planes orthogonal to the acquisition plane (axial), i.e. the coronal and sagittal image views. b Looking at the data residuals to the diffusion tensor fit may reveal other data abnormalities, for instance, in the form of hyperintensities around the rim of the brain (iii), which can also be observed in the FA maps (c). This high anisotropy rims suggest image misalignment across the different diffusion-weighted images due to subject motion or geometric distortions. In (d), the FA map is shown after performing this correction procedure.

132

Jones and Leemans

5. Re-inspecting the Data: We re-inspect the data in three orthogonal planes in a cine format to ensure that the motion/distortion correction has been performed correctly and that no additional artefacts have been introduced into the data. 6. Fitting the Diffusion Tensor: We perform an initial ordinary least squares fit of a single Gaussian diffusion tensor to the data to generate a starting estimate for subsequent entry into a nonlinear least squares (Levenberg–Marquardt) algorithm to estimate the tensor in each voxel (see Note 20). 7. Further Inspection of the Data: We generate a 4D data set of the residuals to the tensor fit (Fig. 6.1b) and again view these in three orthogonal planes as a movie to look for obvious outliers/artefacts (see Note 21). 8. Computation of Parametric Maps: The tensor in each voxel is diagonalised to derive the eigenvectors and eigenvalues, and parametric maps of the mean diffusivity and fractional anisotropy (2) are computed (see Note 22). We also compute the directionally encoded colour (DEC) map showing fibre orientation (5). 9. Final Inspection of the Data: Before passing any maps into further analyses, we visualise the FA, MD and DEC maps in three orthogonal orientations to check for any obvious artefacts, such as rims of high anisotropy at the edge of the brain (see Fig. 6.1c and Note 23).

4. Notes 1. In most research environments for human imaging, the field strength is 1.5 or 3.0 T with a few exceptions. The advantage of higher field is higher signal-to-noise ratio (SNR) per unit time – allowing higher resolution for fixed scan time, shorter scan time for the same resolution or higher precision in the data if all scan parameters are kept the same. The disadvantage is that the standard approach to DT-MRI (1, 2) uses echo-planar imaging (EPI) – which renders the images to be very sensitive to differences in magnetic susceptibility (such as at air–tissue interfaces) – leading to either a stretch or a compression of the image along the phase encode direction (usually aligned with the anterior–posterior axis of the head), and these distortions become worse at higher field strength. 2. The imaging gradients should also ideally have the capability of producing gradient amplitudes of above 20 mT/m. The stronger the gradient amplitudes, the better for

Diffusion Tensor Imaging

133

diffusion MRI. The key factors in determining the amount of diffusion weighting, characterised by the ‘b-factor’, b, are the amplitude (G), the duration (δ) and the temporal separation ( ), given by the so-called Stejskal–Tanner equation (3):  δ , − 3

 2

2 2

b=γ G δ

where γ is the gyromagnetic ratio. Increasing the gradient amplitude means that smaller values of δ and can be employed, which in turn means shorter echo times and therefore increased SNR. 3. A multi-channel head coil is preferable for improved SNR and the possibility to employ parallel imaging, which in turn helps to reduce the EPI-based distortions (12). Eightchannel head coils are prevalent in neuroimaging research centres, but again – more channels, if available, are preferred with 12-channel and even 32-channel coils becoming purchasable options. Again – these will boost the SNR – which is most definitely beneficial for DT-MRI. Further, more channels permit the use of higher speed-up factors in parallel imaging acquisition strategies, which is advantageous for DT-MRI (12). 4. The scanner should provide the capability of applying diffusion-encoding gradients in at least six different orientations (1), although more is better (see Note 13). The capability to choose one’s own encoding directions (normally facilitated under a research agreement with the manufacturer) is preferred, particularly if the sampling orientations are to be optimally ordered (see Paragraph 2 of Note 13). 5. Past experience has shown that it is beneficial to alert the participant to the fact that the vibrations can be quite severe. This is particularly true with the twice-refocused pulse sequence – as reported in (13). Moreover, it is useful to warn the participant that there will be spatially varying irregularly spaced knocking which refers to the fact that the TR will be non-uniform due to the cardiac gating, and the knocking is the gradients being played out in different orientations. 6. The use of a video projection system, at least anecdotally, appears to reduce head motion as it engages the participant and reduces the likelihood of them looking around (which may result in additional head motion) or being distracted by other thoughts over what can be a considerable duration of scan (sometimes up to 30 min). It is not expected that there will be functionally dependent changes in the signal

134

Jones and Leemans

acquired with a standard DT-MRI protocol that will impact on standard DT-MRI analyses. 7. The most commonly used acquisition scheme for DT-MRI is now the twice-refocused spin-echo EPI sequence (11) and is provided by most manufacturers as part of the DT-MRI package that they sell. The twicerefocused sequence will markedly reduce the effects of eddy current-induced distortions resulting from the rapid switching of the diffusion-encoding gradients. 8. For EPI-based DT-MRI acquisitions, there is a clear benefit to the use of parallel imaging (12) using approaches such as the image domain SENSE approach (14) or k-space domain approaches such as GRAPPA (15). The choice of which to use normally depends on availability and/or software provided by the manufacturer. A parallel imaging factor of 2 seems to provide a reasonable compromise between SNR, distortion reduction and speed-up factor. 9. Pure axial orientation gives the best quality data in terms of ghosts and distortions. 10. For most purposes, it is highly desirable to have isotropic imaging voxels for DTI so that there is no preferential averaging of fibre orientations along a particular axis. This is particularly important for tractography applications. The limitations to consider are (1) The slice profile: On some scanners (e.g. General Electric), the issue of the fat–water frequency shift, particularly problematic in EPI, is addressed through the use of a spatially and spectrally selective pulse – which leaves the fat (in the scalp, for example) unexcited. While this is an alternative to implicitly turning on ‘FatSat’, it does tend to limit the minimal slice thickness that can be achieved (around 2.5 mm is typical). (2) The signal-to-noise ratio: It is worth bearing in mind that the average SNR in the diffusion-weighted image is on the order of 30% of that in the non-diffusion weighted image. If the non-diffusion weighted signal intensity is I0 then the diffusion-weighted signal is I0 exp(-bD), where D is the apparent diffusion coefficient along the axis of the applied encoding gradient. As discussed below, the b-value is typically on the order of the reciprocal of D so that I = I0 exp(–1) = 0.33I0 . For isotropic media, all diffusion-encoding directions will have the same attenuation. It is important that the SNR in the diffusion-weighted images does not go below approximately 3:1. This is the domain where the Rician-distributed data begin to look non-Gaussian (16, 17), and one encounters problems with the rectified noise floor that cause underestimation of dif-

Diffusion Tensor Imaging

135

fusivities (18) and corrupt estimates of diffusion anisotropy (19), among other problems. For anisotropic media, say with a fractional anisotropy (20) of 0.7, the largest eigenvalue will be twice the mean diffusivity. This, in turn, means that the signal attenuation will be equal to exp(–2bD) = 0.13. Therefore to void the rectified noise floor issue, the SNR in the non-diffusion weighted image should be at least 3:0.13∼ = 22:1. It is therefore advisable, when setting up the protocol, to make some SNR measurements (with the coil, parallel imaging factor, etc., for the experiment already chosen and in place) before settling on the final resolution. In summary, one should go for the highest resolution possible that achieves isotropic resolution but ensures that the SNR in white matter in the non-diffusion image is greater than 20. The field of view should be sufficient to cover the entire head and selected in consideration of the image acquisition matrix and slice thickness to ensure isotropic voxels. 11. The more measurements are made, the more accurate and precise the parameters derived from DTI will be. The total number of measurements will therefore be dictated by the time allocated in the protocol. The ratio of measurements made at the higher b-value to those with no diffusion weighting should be around 8–10:1 (21); it is highly recommended that the images with b = 1,000 s/mm2 are acquired with gradients distributed as uniformly in space as possible (21, 22 – see Note 13). If time permits, 30 directions (with two or three non-diffusion weighted scans) will give measures that are statistically rotationally invariant (22) – but if time permits even more measurements to be made, then one should take advantage and acquire more points. 12. It is recommended that just two different diffusion weightings are used: one close to zero and one close to b = 1,000 s/mm2 , the latter being a compromise between ensuring sufficient attenuation for precise estimation of diffusivities and balancing this against the penalty of increasing the echo time to accommodate the diffusion-encoding gradients, which in turn leads to signal loss through T2 relaxation (21). 13. For estimating the tensor, it is beneficial to space out the sampling gradients as uniformly in space as possible (21, 22). This helps to reduce the dependence of the reproducibility (variance) of measurements on the orientation of the material being imaged (22). The de facto standard, adopted by most manufacturers and research groups, is now to use an ‘electrostatic repulsion’ algorithm (21) to

136

Jones and Leemans

space out the gradients. This is done by simulating what happens if the gradient sampling vectors are all rods passing through the origin (so that each rod points in the direction of a sampling vector and its antipode) and if there are point charges placed at the ends of each rod. The orientations of the rods are then changed until the sum of repulsive forces between all possible pairs of charges is minimised (this is effectively what happens in the formation of crystals in nature – so it is no surprise that, for appropriate numbers of sampling vectors, the orientations that are produced by this algorithm produce the regular polyhedric arrangements – such as the tetrahedral arrangement of sp3 hybrid orbitals seen in crystals (23)). If the scan is run to completion, then the order in which the different sampling orientations are played out is irrelevant – as the complete data set is available and all points go into the estimation of the tensor. However, in subjects where it is anticipated that there may be a curtailment (for example, due to excessive motion/claustrophobia or general non-compliance), stopping a randomly ordered acquisition before completion could lead to, for example, lots of measurements being made along similar orientations – with large portions of the sampling space left out (which is sub-optimal). To counter this, Cook et al. (24) and Dubois et al. (25) have independently proposed ways of re-ordering the directions derived from the electrostatic repulsion algorithm so that, if (for example) a 30-direction sampling scheme is interrupted half-way through, the 15 directions that have been collected are still pretty much uniformly distributed. This seems an eminently sensible strategy and, if the full data set is collected anyway, has no impact on the data since, as stated before, the ordering is immaterial. Table 6.1 lists the example of 60 electrostatically arranged sampling vectors that have been ordered according to this principle and are used in our protocol. For the sake of completeness, in Tables 6.2 and 6.3, we tabulate the optimal point ordering for 18 and 30 unique sampling orientations, respectively. These were obtained with the Camino package (26), which could be used to generate other numbers of ordered point sets. 14. One is always battling with SNR issues in DT-MRI (27). Therefore, one should strive to minimise additional attenuation – consequently, whenever possible, having selected the b-value on the scanner console, one should select the ‘minimum TE’ setting. This can force the acquisition into a partial k-space acquisition, with, for example, 8 lines of k-space acquired before the echo and 48 lines acquired

Diffusion Tensor Imaging

137

Table 6.2 18 Electrostatically arranged and optimally ordered sampling vectors 0.7371

−0.5680

0.3662

0.7958

0.4311

0.4253

−0.8225

0.3677

0.4339

0.0006

0.9856

0.1692

0.2290

0.1508

0.9617

−0.4124

−0.7535

0.5120

−0.3586

0.2328

0.9040

−0.8912

−0.4176

0.1768

0.3199

−0.4987

0.8056

0.3099

0.6677

0.6769

0.5797

−0.8070

−0.1124

−0.2096

−0.3585

0.9097

0.9907

−0.1123

0.0774

0.1533

−0.9033

0.4008

0.5302

0.8454

0.0651

−0.2829

0.7167

0.6374

0.7201

−0.0527

0.6919

−0.7339

−0.1786

0.6554

afterwards. In some cases, it may be necessary to increase the number of lines of k-space before the centre of k to avoid corruption of the data due to excessive vibration. 15. It has been shown in several studies that diffusion MRI measurements can be severely corrupted by the effect of cardiac pulsation (28–31). As the heart beats – the pressure wave is carried through to the brain – and one gets both local deformation of the tissue (32, 33), which results in local misregistration between successive images, and intra-voxel dephasing due to diffusion (30). The latter will be interpreted as increased diffusion measured at the point a particular encoding gradient is applied – which, in turn, will lead to biases in the estimate of the diffusion tensor, resulting in inaccurate estimates of both anisotropy and fibre orientation (31). To avoid this, it is recommended that the acquisition be timed so as to avoid acquiring data during the time that the brain tissue is susceptible to this pulsation. 16. We have previously found that corruption of data due to cardiac pulsation can be avoided by waiting until at least

138

Jones and Leemans

Table 6.3 30 Electrostatically arranged and optimally ordered sampling vectors 0.0559

−0.9920

−0.1134

−0.6669

−0.6780

0.3091

0.1637

0.5330

0.8301

−0.4320

−0.0898

0.8974

0.8865

0.2190

0.4076

0.7671

−0.6107

0.1964

0.3840

−0.2940

0.8753

−0.5464

0.6192

0.5639

−0.9500

0.0308

0.3108

0.0973

−0.7119

0.6955

−0.7308

−0.6541

−0.1950

0.0777

0.0950

0.9924

−0.3115

0.9086

−0.2781

0.2961

0.8430

0.4491

−0.7523

−0.3014

0.5858

0.7882

−0.2137

0.5771

0.8458

−0.4786

−0.2358

−0.1722

−0.9086

0.3806

−0.2838

0.3816

0.8797

0.5428

0.1339

0.8291

−0.3989

−0.6000

0.6934

0.5072

−0.8486

−0.1505

−0.9375

−0.3410

0.0689

0.1467

−0.8120

−0.5649

−0.9757

0.1655

−0.1434

220 ms after the onset of the R-wave in the ECG trace recorded from chest leads (30). In turn, we have also found that the arrival of the pulse wave on a peripheral pulse-oximeter placed on the finger is 249 ± 17 ms (mean ± SD) after the R-wave of the ECG (30). Consequently, the delay needed after the peak on a peripheral oximeter trace to avoid DT-MRI data corruption is minimal. It is therefore beneficial to use a peripheral oximeter – as there is less ‘dead time’ in each pulse-to-pulse interval, and it is undoubtedly more convenient for the participant. One point to note is that since the repetition time will then vary due to natural variations in the cardiac cycle, it is important that the effective TR is at least five times the T1 of the tissue

Diffusion Tensor Imaging

139

of interest (to avoid partial recovery effects). However, in practice – this condition is almost always satisfied – as one can rarely squeeze in more than three or four slices into an R–R interval and, for a typical 60-slice acquisition, this means an effective TR of 15 or 20 R–R intervals. We use a look-up table to rapidly determine the optimal effective TR (Table 6.4).

Table 6.4 Prescribed effective TR as a function of participant heart rate Participant heart rate

Effective TR

<50 bpm

15 R–R intervals

50–64 bpm

20 R–R intervals

65–95 bpm

30 R–R intervals

>95 bpm

60 R–R intervals

17. The first and most important thing that must be done is to examine the raw data. With pressures to get studies completed quickly, this is an often overlooked step in many laboratories. However, given that one is going to take multiple images and use them to compute the diffusion tensor at each voxel (1) – it is important to check that there are no corruptions in any of the individual data points. It can be extremely informative to view the data in three orthogonal planes simultaneously (see Fig. 6.1a). There are multiple tools available for doing this (for example, the popular FSLview from the FSL software library). This allows for rapid identification of slice-to-slice intensity variations (for an axially acquired data set, these will be visible on the coronal and sagittal planes). With the data set stored in a 4D format (the fourth dimension being the number of the diffusion-weighted scan), viewing the data set as a looping movie is a very efficient way of checking the data for unexpected signal dropouts (which will be visible as unusually dark horizontal bands on the sagittal and coronal slices) or other artefacts (see Fig. 6.1a). 18. It is extremely important to correct the data for subject motion and eddy current-induced distortions (e.g. 34, 35). Although the twice-refocused spin-echo sequence will ameliorate much of the eddy currents (11), there may still be residual distortions that need to be taken care of and it is unlikely that the participant will have remained perfectly still. The ‘industry standard’ approach is to use a global affine registration (12 degrees of freedom – with translation, rotation, magnification and shear along each of the

140

Jones and Leemans

principal axes) to register each diffusion-weighted volume to the volumes collected without any diffusion-weighting applied (the ‘b = 0 s/mm2 images’). Given that the image contrast is so different between the b = 0 image and those acquired at b = 1,000 s/mm2 , cost functions for the registration such as cross-correlation tend to fail and much better results are obtained with entropy-based metrics, such as the mutual information index (and its normalised version) (36, 37). Again, there are many software packages that cater to such global affine-based registration requirements, allowing one to specify the cost function. 19. Estimation of the tensor matrix requires exact knowledge of the orientation of the diffusion-encoding gradient with respect to the participant’s head (1). If the participant moves their head during the acquisition – this can be corrected with image realignment methods – as just discussed. However, simple naïve application of ‘off-the-shelf’ registration software will not account for the fact that, during such a rotation, the angle between the participant’s head and the pre-selected gradient sampling vectors will change. Failing to account for this can lead to substantial errors in estimates of anisotropy and of fibre orientation (38). Therefore, it is desirable that, when available, the gradient table that is used as input into the tensor estimation routine is modified accordingly. The amount of diffusion encoding for a particular combination of gradients is characterised by the b-matrix (1). The effect of rotation can be handled by deriving the rotational part of the transformation required to realign the images and subsequently applying this rotation to the b-matrix prior to estimation of the tensor. Figure 6.2 shows the effect of neglecting to perform this step on estimates of anisotropy and fibre orientation. 20. There are three widely used approaches to estimating the diffusion tensor from the b-matrix and the diffusionweighted data: ordinary linear least squares (OLLS), weighted linear least squares (WLLS) (1) and nonlinear least squares (NLLS) (39). (Note that there are lots more approaches in the literature – but these are the most common.) It is outside the scope of this chapter to go into the fine detail of these different approaches – but standard mathematical processing packages provide access to all of these. For the first two approaches, i.e. linear fitting, the diffusion-weighted signal intensities are first log transformed (1). In OLLS, each observation contributes equally to the fit – and thus a set of simultaneous equations relating the log of the signal to the unknown elements of the tensor are set up – and a simple matrix

Diffusion Tensor Imaging

141

Fig. 6.2. Directionally encoded fibre orientation maps without (a) and with (b) the required b-matrix rotation prior to estimating the diffusion tensor. The differences in orientation (colour) and fractional anisotropy (intensity) between the images is clearly visible, as shown, for instance, with the enlarged view of the posterior region of the corpus callosum.

inversion yields the unknown elements of the tensor. Given the rapid nature of this approach, it is extremely popular and is employed in several popular software packages (e.g. FSL). The consequence, however, of taking the log transformation is that, although the noise/random errors in the signal prior to the log transformation are uniform (i.e. homoscedastic) – after the log transformation, the variance in the (log-transformed) signal becomes a function of the signal itself (1, 40) and so the errors are heteroscedastic. To properly address this, a weighted linear regression (WLLS) is required, where one has to compute a covariance matrix – deriving the relationship between the variance in the logtransformed and non-log transformed data – and include it in the regression step (1). Although this makes computation longer, the results are far more robust and WLLS is to be preferred over OLLS (39). In NLLS, on the other hand, there is no log transformation of the signal – and thus the errors remain homoscedastic, so the covariance matrix is a multiple of the identity matrix and can effectively be factored out of the analysis, which is an advantage over the linear framework approaches. While NLLS is attractive in that it fits the model to the data directly, producing results that are superior to WLLS (and therefore to OLLS) (39), the computational time is considerably longer – and special care has to be taken that the fitting algorithm has not been trapped in a local extremum. 21. Once one has fitted the model, it is expected that any differences between the observed signal and those predicted by the model (i.e. the ‘residuals’) should be random and

142

Jones and Leemans

should not contain any structure. Deviations will rapidly identify points that are corrupt (41) (see Fig. 6.1b). Therefore, before proceeding with further analyses, we recommend inspecting a map of the residuals to ensure that there is nothing unexpected in the data. 22. The final stage in a standard DT-MRI pipeline is to derive parameters of interest from the diffusion tensor. These are invariably the mean diffusivity (or trace of the diffusion tensor) (2), a measure of anisotropy (27) – and the principal diffusion orientation, which can be used for creation of directionally encoded colour (DEC) maps (5) or for fibre-tracking analyses (6–10). With regard to the choice of anisotropy index, there is a plethora of indices in the literature (20). However, by far the most popular is the fractional anisotropy, derived from the three eigenvalues of the diffusion tensor. One can show that this has a better signalto-noise ratio characteristic than other popular indices such as relative anisotropy (42). 23. It is always prudent to look carefully at the resultant parametric maps before inputting them into any form of analysis. Values of anisotropy greater than 1 are physically nonmeaningful, since they are designed to lie between 0 and 1 – and will result when one or more of the eigenvalues is negative. Again, this is non-physically meaningful – but can occur when the diffusion-weighted signal is higher in intensity than the non-diffusion weighted signal. In turn, this may arise in regions that are particularly noisy – or where there is insufficiently corrected misregistration. A similar tell-tale sign for artefact is to examine the rim of the anisotropy maps. A high anisotropy around the rim of the adult mammalian brain is not expected in practice (Fig. 6.1c). Given that high anisotropy means that the DW signal varies rapidly with the direction of the diffusionencoding gradient, the appearance of such a bright rim would be consistent with misregistration of, for instance, CSF and gray matter with either white matter or perhaps the edge of the brain. References 1. Basser, P. J., Mattiello, J., LeBihan, D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66: 259–267. 2. Pierpaoli, C., Jezzard, P., Basser, P. J., Barnett, A., Di Chiro, G. Diffusion tensor MR imaging of the human brain. Radiology 1996;201:637–648.

3. Stejskal, E. O., Tanner, J. E. Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J Chem Phys 1965;42:288–292. 4. Moseley, M. E., Cohen, Y. C., Kucharczyk, J., Asgari, H. S., Wendland, M. F., Tsuruda, J., Norman, D. Diffusion-weighted MR imaging of anisotropic water diffusion

Diffusion Tensor Imaging

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

in cat central nervous system. Radiology 1990;176:439–445. Pajevic, S., Pierpaoli, C. Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: Application to white matter fiber tract mapping in the human brain. Magn Reson Med 1999;42:526–540 (Erratum in: Magn Reson Med 2000 43:921). Mori, S., Crain, B. J., Chacko, V. P., van Zijl, P. C. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol 1999;45: 265–269. Jones, D. K., Simmons, A., Williams, S. C., Horsfield, M. A. Non-invasive assessment of axonal fiber connectivity in the human brain via diffusion tensor MRI. Magn Reson Med 1999;42:37–41. Conturo, T. E., Lori, N. F., Cull, T. S., Akbudak, E., Snyder, A. Z., Shimony, J. S., McKinstry, R. C., Burton, H., Raichle, M. E. Tracking neuronal fiber pathways in the living human brain. Proc Natl Acad Sci USA 1999;96:10422–10427. Basser, P. J., Pajevic, S., Pierpaoli, C., Duda, J., Aldroubi, A. In vivo fiber tractography using DT-MRI data. Magn Reson Med 2000;44:625–632. Parker, G. J. M., Haroon, H. A. et al. A framework for a streamline-based probabilistic index of connectivity (PICo) using a structural interpretation of MRI diffusion measurements. J Magn Reson Imag 2003;18:242–254. Reese, T. G., Heid, O., Weisskoff, R. M., Wedeen, V. J. Reduction of eddy-currentinduced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson Med 2003;49:177–182. Bammer, R., Auer, M., Keeling, S. L., Augustin, M., Stables, L. A., Prokesch, R. W., Stollberger, R., Moseley, M. E., Fazekas, F. Diffusion tensor imaging using single-shot SENSE-EPI. Magn Reson Med 2002;48:128–136. Hiltunen, J., Hari, R., Jousmäki, V., Müller, K., Sepponen, R., Joensuu, R. Quantification of mechanical vibration during diffusion tensor imaging at 3 T. Neuroimage 2006;32: 93–103. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. Griswold, M. A., Jakob, P. M., Heidemann, R. M., Nittka, M., Jellus, V., Wang, K., Kiefer, B., Haase, A. Generalized autocalibrating partially parallel

16.

17. 18.

19.

20.

21.

22.

23.

24.

25.

26.

27. 28.

143

acquisitions (GRAPPA). Magn Reson Med 2002;47:1202–1210. Edelstein, W. A., Bottomley, P. A., Pfeifer, L. M. A signal-to-noise calibration procedure for NMR imaging systems. Med Phys 1984;11:180–185. Gudbjartsson, H., Patz, S. The Rician distribution of noisy MRI data. Magn Reson Med 1995;34:910–914. Dietrich, O., Heiland, S., Sartor, K. Noise correction for the exact determination of apparent diffusion coefficients at low SNR. Magn Reson Med 2001;45:448–453. Jones, D. K., Basser, P. J. “Squashing peanuts and smashing pumpkins”: How noise distorts diffusion-weighted MR data. Magn Reson Med 2004;52:979–993. Basser, P. J., Pierpaoli, C. Microstructural and physiological features of tissue elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209–219. Jones, D. K., Horsfield, M. A., Simmons, A. Optimal strategies for measuring diffusion in anisotropic systems by magnetic resonance imaging. Magn Reson Med 1999;42: 515–525. Jones, D. K. The effect of gradient sampling schemes on measures derived from diffusion tensor MRI: A Monte Carlo study. Magn Reson Med 2004;51:807–815. Conturo, T. E., McKinstry, R. C., Akbudak, E., Robinson, B. H. Encoding of anisotropic diffusion with tetrahedral gradients: A general mathematical diffusion formalism and experimental results. Magn Reson Med 1996;35:399–412. Cook, P. A., Symms, M., Boulby, P. A., Alexander, D. C. Optimal acquisition orders of diffusion-weighted MRI measurements. J Magn Reson Imaging 2007;25: 51–58. Dubois, J., Poupon, C., Lethimonnier, F., Le Bihan, D. Optimized diffusion gradient orientation schemes for corrupted clinical DTI data sets. Magma 2006;19: 134–143. Cook, P. A., Bai, Y., Nedjati-Gilani, S., Seunarine, K. K., Hall, M. G., Parker, G. J., Alexander, D. C. (2006) Camino: OpenSource Diffusion-MRI Reconstruction and Processing, 14th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, Seattle, WA, USA, p. 2759, May 2006. Pierpaoli, C., Basser, P. J. Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 1996;36:893–906. Turner, R., Le Bihan, D., Maier, J., Vavrek, R., Hedges, L. K., Pekar, J. Echo-planar

144

29.

30.

31.

32.

33.

34.

35.

Jones and Leemans imaging of intravoxel incoherent motions. Radiology 1990;177:407–414. Skare, S., Andersson, J. L. On the effects of gating in diffusion imaging of the brain using single shot EPI. Magn Reson Imaging 2001;19:1125–1128. Pierpaoli, C., Marenco, S., Rohde, G., Jones, D. K., Barnett, A. S. (2003). Analyzing the contribution of cardiac pulsation to the variability of quantities derived from the diffusion tensor in “Proc. ISMRM 11th Annual Meeting, Toronto” p. 70. Jones, D. K., Pierpaoli, C. (2005) The contribution of cardiac pulsation to variability in tractography results. In “Proc. ISMRM 13th Annual Meeting, Miami”, p. 225. Poncelet, B. P., Wedeen, V. J., Weisskoff, R. M., Cohen, M. S. Brain parenchyma motion: Measurement with cine echoplanar MR imaging. Radiology 1992;185: 645–651. Enzmann, D. R., Pelc, N. J. Brain motion: Measurement with phase-contrast MR imaging. Radiology 1992;185: 653–660. Rohde, G. K., Barnett, A. S., Basser, P. J., Marenco, S., Pierpaoli, C. Comprehensive approach for correction of motion and distortion in diffusion-weighted MRI. Magn Reson Med 2004;51:103–114. Andersson, J. L., Skare, S. A modelbased method for retrospective correction of geometric distortions in diffusionweighted EPI. Neuroimage 2002;16: 177–199.

36. Maes, F., Collignon, A., Vandermeulen, D., Marchal, G., Suetens, P. Multimodality image registration by maximization of mutual information. IEEE Trans Med Imaging 1997;16:187–198. 37. Studholme, C., Constable, R. T., Duncan, J. S. Accurate alignment of functional EPI data to anatomical MRI using a physics-based distortion model. IEEE Trans Med Imaging 2000;19:1115–1127. 38. Leemans, A., Jones, D. K. The B-matrix must be rotated when motion correcting diffusion tensor imaging data. Magn Reson Med 2009;61:1336–1349. 39. Koay, C. G., Chang, L. C., Carew, J. D., Pierpaoli, C., Basser, P. J. A unifying theoretical and algorithmic framework for least squares methods of estimation in diffusion tensor imaging. J Magn Reson 2006;182: 115–125. 40. Bevington, P. R., Robinson, D. K. Data Reduction and Error Analysis for the Physical Sciences, 2nd ed. New York, NY: McGrawHill; 1992. 41. Leemans, A., Evans, C. J., Jones, D. K. (2008). Quality assessment through analysis of residuals of diffusion image fitting. In “Proc. ISMRM 16th Annual Meeting, Toronto”. p. 3300. 42. Hasan, K. M., Alexander, A. L., Narayana, P. A. Does fractional anisotropy have better noise immunity characteristics than relative anisotropy in diffusion tensor MRI? An analytical approach. Magn Reson Med 2004;51:413–417.

Chapter 7 Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) Cynthia A. Massaad and Robia G. Pautler Abstract The use of manganese ions (Mn2+ ) as an MRI contrast agent was introduced over 20 years ago in studies of Mn2+ toxicity in anesthetized rats (1). Manganese-enhanced MRI (MEMRI) evolved in the late nineties when Koretsky and associates pioneered the use of MEMRI for brain activity measurements (2) as well as neuronal tract tracing (3). Currently, MEMRI has three primary applications in biological systems: (1) contrast enhancement for anatomical detail, (2) activity-dependent assessment and (3) tracing of neuronal connections or tract tracing. MEMRI relies upon the following three main properties of Mn2+ : (1) it is a paramagnetic ion that shortens the spin lattice relaxation time constant (T1 ) of tissues, where it accumulates and hence functions as an excellent T1 contrast agent; (2) it is a calcium (Ca2+ ) analog that can enter excitable cells, such as neurons and cardiac cells via voltage-gated Ca2+ channels; and (3) once in the cells Mn2+ can be transported along axons by microtubule-dependent axonal transport and can also cross synapses trans-synaptically to neighboring neurons. This chapter will emphasize the methodological approaches towards the use of MEMRI in biological systems. Key words: MEMRI, rodents, manganese, central nervous system, contrast agent, MRI.

1. Introduction Mn2+ is a trace element essential for normal body function and development throughout the lifespan of mammals (4). Most notably Mn2+ is an essential cofactor for several enzymes responsible for a wide variety of physiological body functions (4). Such enzymes include manganese superoxide dismutase (5) which is essential for oxidative stress prevention, pyruvate carboxylase (6) which plays a critical role in gluconeogenesis, arginase (7) which is involved in urea production by the liver, and glutamine synthetase M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_7, © Springer Science+Business Media, LLC 2011

145

146

Massaad and Pautler

(8), an astrocyte-specific enzyme regulated by about 80% of brain Mn2+ . The importance of Mn2+ is illustrated by studies linking disruption of Mn2+ homeostasis to disease occurrence (5, 9). Mn2+ deficiency has been linked to deficient bone metabolism in rats (10), as well as skin lesions, bone malformation, epileptic seizures, and increased Ca2+ and phosphorous levels in humans (11). Although Mn2+ deficiency is clearly associated with adverse effects, the aforementioned studies were achieved with an artificially induced Mn2+ -deficient diet (11). No naturally occurring Mn2+ -deficiency related diseases have been observed. However, Mn2+ is more frequently of toxicological concern. Although it is considered the least toxic of all essential elements (12), excessive exposure to the metal leads to central nervous system toxicity (4). It has been shown that Mn2+ can enter the central nervous system either directly via the olfactory receptor neurons or through the blood brain barrier by diffusion or active transport (13, 14). Once in the nervous system, Mn2+ is transported along neurons by microtubule-dependent axonal transport (15, 16) and can traverse synapses to accumulate in neighboring neurons (17, 18). The resulting neurotoxicity preferentially targets the striatum leading to Parkinson’s disease-like symptoms, including generalized bradykinesia, widespread rigidity, tremors, hallucinations, and memory loss (4, 19, 20). In addition to multiple roles in normal physiology, Mn2+ is also a Ca2+ analog and can enter excitable cells via several types of Ca2+ channels such as voltage-gated Ca2+ channels and the Na+ /Ca2+ exchanger (21–24). Mn2+ also accumulates in mitochondria via the mitochondrial Ca2+ uniporter (25, 26). The analogy of Mn2+ with Ca2+ resulted in the use of Mn2+ as a fura-2 quencher and hence Ca2+ indicator in biological systems by fluorescence microscopy (27–31). Another very important feature of Mn2+ is that it is paramagnetic and produces MR contrast by causing a strong reduction in the T1 relaxation times of water (32–35). Positive contrast is detected in T1 -weighted images of tissues where Mn2+ accumulates (32–35). The combined physical and biological properties of Mn2+ make it a useful contrast agent for anatomical and functional imaging in multiple systems. Indeed, manganese-enhanced MRI (MEMRI) has been gaining growing interest in the past few years (2, 3, 36, 37) and currently has three main applications for biological systems. First, owing to its contrast-enhancing properties, systemic Mn2+ injections are used for enhancement of the brain cytoarchitecture for anatomical studies (38–44). This technique has been used in adult, as well as in young developing organisms. Its use has further been extended to studying the development of embryos in utero (45). Second, given that Mn2+ can enter cells via voltage-gated Ca2+ channels, it is used as a

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

147

marker of activity in specific protocols that promote its accumulation in active brain areas (2, 46–52). This use of MEMRI is termed activation-induced MEMRI or AIM-MRI. AIM-MRI also has applications in the heart because of the high concentration of Ca2+ channels (53). This protocol, however, will emphasize MEMRI applications in the nervous system. The third and last application of MEMRI is tract tracing; given that Mn2+ is transported by microtubule-dependent axonal transport and can cross synapses to reach post-synaptic neurons, MEMRI has been used as a neuronal tract tracer for several neuronal pathways including the visual, olfactory, and somatosensory pathways, in a variety of animal models, such as mice, rats, monkeys, and birds (3, 54–63). This review will focus on MEMRI applications in rodents. The versatility of MEMRI is also demonstrated by the development of methods for dynamic Mn2+ transport imaging, which are proving as useful markers of disease and related therapy (64, 65). The following chapter will expand upon each of the three applications of MEMRI with special emphasis on techniques related to each application.

2. Materials 2.1. Anatomical Contrast Enhancement 2.1.1. Intravenous MnCl2

1. MnCl2 as a source of Mn2+ 2. Sterile water 3. Sterile saline 4. Beaker of warm water for tail warming and dilation of the tail vein 5. 27- or 30-gauge needle 6. Forceps 7. 1-ml syringes 8. Tubing suitable to attach to a 27-gauge needle 9. Pre-anesthetic (e.g., glycopyrrolate) 10. Anesthetic (e.g., isoflurane) 11. Analgesic (e.g., bupivicaine) 12. Sterile saline 13. Tape 14. 4-gauge nylon suture 15. Syringe pump

148

Massaad and Pautler

16. Warming blanket 17. Small animal-monitoring system complete with rectal temperature probe and respiration sensor 18. Neuromuscular blocking agent (pancuronium bromide or gallamine triethiodide) 19. Ventilator 2.2. ActivationInduced MRI 2.2.1. Intravenous MnCl2

see Section 2.1.1.

2.2.2. Blood Brain Barrier (BBB) Disruption 2.2.2.1. Hyperosmolar Mannitol Infusion:

2.2.2.2. Hyperosmolar Mannitol Injection Through the External Carotid Artery:

• This requires the infusion of mannitol through the tail vein (or femoral vein). Materials will be identical to Section 2.1 with the exception of using a 5–10% mannitol solution instead of the MnCl2 solution (66–69). 1. Anesthesia (e.g., isoflurane, urethane or α-chloralose) 2. Tape 3. Hair clipper (Note 1) 4. Surgical tools (blade with holder, hemostat, forceps, scissors etc.) 5. Microvascular clips 6. Disinfecting solutions (e.g., betadine, chlorhexidine, alcohol) 7. Polyethylene tubing PE-50, thinned to an outer diameter of ∼0.4 mm (Note 2) 8. PE90 tube attached to the hub of a needle 9. 6-0 nylon suture 10. Mannitol solution (20%) 11. Small metal laryngoscope 12. Muscular blocking agent (pancuronium bromide or gallamine triethiodide) 13. Ventilator

2.2.3. Intraperitoneal MnCl2 – Visual and Auditory Activation

1. MnCl2 as a source of Mn2+ (66 mg/kg in saline – for intraperitoneal injections) (48, 70). 2. 1-ml syringe. 3. For auditory activation studies: auditory isolation box enabled for auditory stimulation with the addition of a sound synthesizer, audio amplifier, and speakers.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

149

4. For visual activation studies: visual stimulation box consisting of four walls made up of 14-inch computer screens. Remaining areas consist of black-painted wood protected by aluminum mesh. 5. Anesthesia (isoflurane or urethane). 2.2.4. Intranasal MnCl2 – Olfactory System Activation

1. MnCl2 as a source of Mn2+ (10 mM in H2 O – for intranasal administration of 7 μl/naris or 1.5 M in H2 O for aerosolized Mn2+ administration with a vaporizer) (52, 61) 2. Pipette – 10 μl 3. Anesthesia (e.g., isoflurane, urethane) 4. Heating pad 5. Odorant for olfactory stimulation (e.g., 1:100 amyl acetate, 1:10 octanal, 1:10 carvone etc.) 6. Vaporizer 7. Fume hood

2.3. Tract Tracing 2.3.1. Tract Tracing – Visual System

1. MnCl2 as a source of Mn2+ (1 M in H2 O – for intravitreal injections) (36, 62) 2. 27-gauge needle 3. Polyethylene tubing (0.4 mm diameter) 4. A 5-μl Hamilton syringe 5. Anesthesia (isoflurane or ketamine/xylazine combination or pentobarbital sodium, see Note 11) 6. Heating pad 7. Dissecting microscope

2.3.2. Tract Tracing – Olfactory System

1. MnCl2 as a source of Mn2+ (3.79 M in H2 O – for intranasal administration) (61, 65) 2. Pipette – 2 μl or 10 μl 3. Anesthesia (e.g., isoflurane) 4. Heating pad

2.3.3. Tract Tracing – Deep Brain Structures

1. MnCl2 as a source of Mn2+ (5 mM in H2 O – for intracranial injections) (36, 63, 71, 72) 2. Anesthesia (e.g., ketamine/xylazine combination as a preoperative followed by 2% isoflurane for maintenance) 3. Surgical tools (blade with holder, hemostat, forceps, scissors etc.) 4. Disinfecting solution (chlorhexidine, betadine, alcohol) 5. Small sharp scissors and/or rodent hair clipper

150

Massaad and Pautler

6. Small drill with associated bits (similar to a dental drill) 7. 6-0 nylon suture 8. Mouse/rat brain atlas 9. Capillary tube puller 10. Quartz capillary tubes with filament 11. Surgical area including stereotaxic holder, dissecting microscope and gaseous anesthesia line 12. Surgical tool sterilizer (e.g., glass beads electric sterilizer) 13. Picospritzer with holder and push/pull options (to fill injection needles and subsequently inject solution out of them) 14. Heating pad 15. Sterile cotton swabs 16. Eye ointment 17. Leveling tool (small fork-shaped metallic tool that can be used to ascertain 2D horizontal leveling of the mouse/rat head in the stereotaxic holder) 18. Calibrated volume gauge (Note 3)

3. Methods 3.1. Anatomical Contrast Enhancement

Anatomical contrast enhancement by systemic Mn2+ injection has been studied in rodents (38–40, 43, 44), birds (57, 58), and primates (54, 55). The methods presented here are specifically designed for rat brain visualization based upon work developed in Koretsky’s laboratory (44). The following methods can be adapted for use with any organism, provided reasonable optimization is conducted on the organism of interest as well as magnetic field strength.

3.1.1. Preparation of the MnCl2 Solution

Different concentrations of MnCl2 can be used in systemic injections for positive contrast in T1 -weighted images (37, 44). Optimization, with regards to the animal model used, as well as available MRI hardware, should be performed for the best results. Also, when preparing MnCl2 , care should be taken as to the tonicity and pH of the final solution. The body fluid has an osmolarity of 300 mOsm/l. One mole of MnCl2 is equivalent to 3 Osm. Therefore, concentrations in the range of 100 mM should be used to insure proper tonicity when large amounts of MnCl2 are to be infused to the animals. When adjusting the pH of MnCl2 solution to a physiological pH of 7.4, bicine buffer, equilibrated with

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

151

NaOH, is a good buffer. Following are guidelines for the preparation of 100 mM MnCl2 at pH 7.4, suitable for imaging of rat cytoarchitecture by systemic MnCl2 injection. 1. Dissolve 1.63 g bicine (FW=163.17) in 100 ml water to obtain a 100 mM bicine solution. 2. Bring solution to pH 7.4 using NaOH. 3. Sterilize solution (either by autoclaving or by filtering). 4. Dissolve 98.95 mg of MnCl2 .4H2 0 (FW=197.91) into 5 ml of sterile bicine solution. Depending on the weight of the animals used, 2–4 ml will be enough for imaging one animal; increase the volume of solution according to the number of animals to be imaged. 3.1.2. Intraperitoneal MnCl2 Injection

3.1.3. Setting Up a Tail Vein Line

Systemic administration of MnCl2 by intraperitoneal injection consists of one injection of 100 mM MnCl2 at a dose of 66 mg/kg. Imaging can be performed as early as 3 h and up to 24 h post-injection. 1. Anesthetize the rat with 4% isoflurane in O2 initially and then keep it anesthetized with 1.5–2% isoflurane using a facemask. 2. Using forceps, break the metallic part of the 27-gauge needle away from its plastic base connector. Take care to avoid causing the needle to get blocked. Connect the metallic part of the needle to its plastic base using a piece of tubing long enough to allow you to comfortably place the connected syringe onto the pump. 3. Fill 1-ml syringe with sterile physiological saline and attach it to the needle/tubing combination. 4. Immerse the tail in warm water to dilate the tail vein. 5. Insert the tip of the needle into the vein; proper insertion is confirmed by the backflow of blood from the tail into the saline-filled attached tubing. 6. To fix the needle in proper place, first use tape over the tubing to loosely hold everything in place. Then, using the nylon suture, tie a knot around the tail/metallic part of the needle. The suture will not go through the skin. 7. Carefully remove the saline-filled syringe and replace it with a MnCl2 pre-filled syringe. 8. Place the MnCl2 syringe into the holder of the syringe pump and set the infusion rate to 1.8 ml/h. Do not start the infusion yet.

3.1.4. MnCl2 Infusion

1. Inject 0.01 mg/kg glycopyrrolate intramuscularly. Glycopyrrolate is a muscarinic cholinergic blocker used as a

152

Massaad and Pautler

pre-anesthetic medication to diminish the risk of vagal inhibition to the heart. 2. Insert a rectal probe into the rat, and maintain the temperature at 37.5◦ C during the infusion using an animal-heating system (e.g., warming blanket or heated air). 3. Keep the anesthesia light during the infusion (0.5–1% isoflurane). 4. The goal MnCl2 concentration is 175 mg/kg, which amounts to approximately 2 ml of total volume per animal (Note 4). 5. To avoid dehydration, inject sterile saline subcutaneously (6.7 ml/100 g) immediately and 6 h after the MnCl2 infusion. 6. Keep the animals under controlled temperature for up to 24 h post-infusion. It is normal for the animals to display lethargic behavior at the end of the MnCl2 infusion. Their behavior will gradually improve to normal by 24 h postinfusion. 3.1.5. Animal Preparation for MRI

1. Anesthetize rats with 4% isoflurane initially. 2. Intubate the animals and keep them ventilated with 1.5% isoflurane in O2 (see Section 3.2.1.1. for detailed intubation protocol). 3. Maintain body temperature at 37.5◦ C using an animalheating system. 4. Monitor temperature, blood pressure, and respiration rate with a small animal physiological monitoring system. 5. Inject the animals with pancuronium bromide (2.5 mg/kg) intraperitoneally to suppress motion (an alternative neuromuscular blocking agent is gallamine triethiodide 80 mg/kg i.v.).

3.1.6. Imaging Parameters

It should be noted that imaging protocols and parameters will vary considerably depending on the field strength to be used. Reported below are the optimal imaging parameters for proton MRI on an 11.7 T magnet based upon the work done by Aoki et al (44). The following parameters can be used as a starting point; however, further optimization should be performed for different field strengths and imaging protocols. Two-dimensional multi-slice multi-echo (MSME-2D) Repetition time (TR) = 300 ms Echo time (TE) = 10.5 ms Matrix size = 256 × 256

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

153

Field of view (FOV) = 25.6 × 25.6 mm Slice thickness (ST) = 1 mm Number of averages (NEX) = 8 An inversion recovery sequence could be used to acquire T1 weighted images as well. The parameters are as follows: Inversion time = 1,100 ms TR = 4,000 ms TE = 11.2 ms Matrix size = 512 × 256 FOV = 38.4 × 19.2 mm ST = 1 mm NEX = 1 Three-dimensional spin echo (SE-3D) TR = 250 ms TE = 7.3 ms Matrix size = 256 × 256 × 128 FOV = 19.2 × 19.2 × 9.6 mm NEX = 2 Total acquisition time = 273 min 3.1.7. Expected Results

The expected results are an increase in positive contrast enhancement in the central nervous system (Fig. 7.1). The pattern of enhancement can be obtained over a wide range of MnCl2 concentrations, with regions of the brain lacking a blood brain barrier (BBB), such as the pituitary gland, exhibiting stronger enhancement at low doses. Regions with an intact BBB, such as the hippocampus and cortex show a dose-dependent increase in contrast enhancement. Higher doses may even allow the detection of finer details of the neuroarchitecture, such as cortical laminae structure.

3.2. AIM-MRI

Activation-induced MRI (AIM) is essentially a Mn2+ -based functional MRI paradigm. It was introduced in 1997 by Lin and Koretsky as a blood-flow independent alternative to functional MARI (fMRI) (2). This method is based on the following two essential properties of Mn2+ : (1) Mn2+ can enter the brain parenchyma from the blood via a disrupted (or leaky) blood brain barrier (BBB) (66) and (2) Mn2+ is a Ca2+ analog that can enter neurons via voltage-gated Ca2+ channels and accumulates in neurons in an activity-dependent manner (30, 73–75). Mn2+ ions cannot efficiently enter the brain parenchyma through an intact BBB (66). Some diffusion may occur at the blood/CSF interface in choroid plexuses, but the amount of Mn2+ entering the brain is minimal compared to the cases where the BBB is disrupted

154

Massaad and Pautler

Fig. 7.1. T1 -weighted MRI after systemic MnCl2 administration in the rat. T1 -weighted MRI of a control rat (column A) and a rat 1 day after IV infusion of MnCl2 solution (column B). Top row shows transverse slices at the level of the olfactory bulb (OB, Bregma: +7 mm). The middle row shows horizontal slices including the hippocampal formation (Bregma: –6 mm). The bottom row shows sagittal slices. The signal intensity of the T1 weighted MRI was enhanced prominently 1 day after systemic administration of MnCl2 in the rat. There were characteristic signal enhancements that were large in the olfactory bulb (OB), hippocampus, cerebellum, and pituitary. Reprinted from Aoki et al. (44), copyright 2004, with permission from Elsevier.

(1, 76, 77). As a result, most AIM studies to date were performed in conjunction with BBB disruption (66). Some studies on the activation of the auditory (48, 78, 79) and visual pathways (70, 80) following auditory and visual stimulation respectively were performed in mice without disruption of the BBB. Also, a subset of functional studies in the olfactory system, capitalized on the active entry of Mn2+ through Ca2+ channels and its trans-synaptic

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

155

transport (61). These studies were a spin-off from tract-tracing studies and were also performed without BBB disruption. Several groups have used this technique since 1997 with certain variations regarding the route of administration of MnCl2 as well as the paradigm followed to break the BBB. Protocols are presented here for intravenous or intraperitoneal MnCl2 administration with hyperosmolar mannitol injections and infusions to break the BBB. Protocols involving olfactory, visual, or auditory activation without BBB disruption are also presented. 3.2.1. Intravenous Infusion of MnCl2 with BBB Disruption

3.2.1.1. Direct Oral/Tracheal Intubation for Artificial Ventilation (Note 5)

The protocol for preparing MnCl2 and setting up a tail vein line for intravenous MnCl2 infusion was described in Sections 3.1.1– 3.1.3. Here, a protocol for disrupting the BBB in conjunction with MnCl2 infusion is described. 1. Anesthetize the mouse/rat with an intraperitoneal injection of α-chloralose and urethane combination (Note 6). 2. Upon lack of toe pinch reflex, place the animal in dorsal position on a pre-heated warming pad. 3. Tape the limbs down. 4. Pull the head back by placing 4-0 nylon behind the upper incisors. 5. Use a small metal laryngoscope to pull the lower jaw down and expose the tracheal opening. 6. Insert a ∼2-cm long PE-90 tubing attached to the hub of a needle about 3 mm into the trachea. The tip of the tube should be beveled in the direction of the natural bend of the tubing to avoid any tissue damage during insertion (Note 7). 7. Attach the needle hub to a ventilator and set it to 80–100 breaths/min. (Do not turn the ventilator on yet; this will occur when imaging begins.)

3.2.1.2. Catheterization of the External Carotid Artery for Hyperosmolar Mannitol Injection

1. While the animal is still in dorsal position from the previous step, shave the neck area with a hair clipper (for rats) or sharp small scissors (for mice) (Note 1). 2. Disinfect the operating field with betadine and 75% alcohol. 3. Make a vertical 5-cm incision and expose the arteries. 4. Temporarily clamp the right common and internal carotids using the microvascular clamps. 5. Carefully insert the PE50 tubing into the external carotid artery through a small puncture in the retrograde direction (Note 8).

156

Massaad and Pautler

6. Secure the tubing in place with 6-0 nylon suture. Tighten the sutures around the tube very well to prevent any further bleeding from the artery. 7. Remove the microvascular clamps. 8. Sew the wound back roughly keeping the tubes well in place. 9. The mouse is now ready to be placed in the magnet. 3.2.1.3. Animal Preparation for MRI

1. Place the animal in a suitable holder with bite bars. 2. Set up a tail vein line for MnCl2 infusion as described in Section 3.1.2 (Note 9). 3. Extend all tubing outside of the magnet room. 4. Maintain anesthesia with 2% isoflurane. 5. Inject the animal with pancuronium bromide (2.5 mg/kg) intraperitoneally to suppress motion (an alternative neuromuscular blocking agent is gallamine triethiodide 80 mg/kg i.v.). 6. Start the ventilator for artificial breathing (80–100 breaths/min).

3.2.1.4. Experimental Protocol

Keeping in mind that there are several different variations of experimental protocols, a typical setting for AIM experiments is given below. 1. Acquire a series of baseline scans. 2. Start the MnCl2 infusion (typically infusions last about 1 h). Care should be taken in selecting the concentration of MnCl2 for AIM studies. MnCl2 causes toxicity to the heart due to Mn2+ homology to Ca2+ . A very common initial effect of MnCl2 is a drop in blood pressure that is typically recovered to normal within 10 min. Therefore, the concentration of MnCl2 may not have drastic effects in long-term studies, but it does affect short-term studies such as AIM. A concentration of 0.2 mmol/kg infused over an hour has been shown to work well. 3. About halfway through the infusion, inject a bolus of 25% mannitol (5–7 ml/kg) through the external carotid catheter. Keep the room temperature higher than 25◦ C and pre-warm all tubing and attachments related to the mannitol injection. This is necessary to prevent mannitol recrystallization and subsequent micro-infarcts (Note 10). Unilateral injection usually results in the opening of the BBB on the side of the injection, with the contralateral receiving less Mn2+ . The inhomogeneous opening of the BBB is not very well understood, but modulating some factors such as dose of mannitol, injection rate, distance of the injection site from the artery, age of the animal etc. plays an important role

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

157

in obtaining sufficient BBB disruption for suitable contrast achievement. 4. At the completion of the MnCl2 infusion, administer the activating signal specific for the AIM study at hand (pharmacological as well as behavioral stimuli have been described). 5. Begin T1 -weighted image acquisitions at different time intervals to measure activation of specific brain areas involved in your study. 3.2.1.5. Imaging Parameters

Different imaging paradigms are possible for nervous system activity measurements using MEMRI in conjunction with BBB disruption. The imaging parameters will depend on the area of interest, organism used, as well as magnetic field strength. Optimization with regards to all of these variables is required for best results. Following is an example of imaging parameters, excerpted from Weng et al (50), for imaging cortical activity in rats following whisker stimulation at 3T. Multi-slice spin-echo sequence TR = 500 ms TE = 10 ms In-plane resolution = 187 μm ST = 1.5 mm

3.2.1.6. Expected Results

Depending upon the system under study, one can expect to detect increased signal enhancement in areas of the brain involved in the activity studied. For example, following whisker stimulation, Weng et al. observed signal enhancement in the right cortical barrels of the rat brain (Fig. 7.2) (50).

Fig. 7.2. Three consecutive slices of the averaged Mn2+ -enhanced T1 WIs under urethane anesthesia a. Mn2+ enhancement was observed in right cortical barrels. The color maps of the averaged Mn2+ -enhanced T1 WIs b. Reprinted from Weng et al. (50), copyright 2007, with permission from Elsevier.

158

Massaad and Pautler

3.2.2. Activation Studies Without BBB Disruption – Visual and Auditory Systems 3.2.2.1. Activation Studies in the Auditory System 3.2.2.1.1. MnCl2 Administration

1. Administer 66 mg/kg MnCl2 as an intraperitoneal injection. 2. Place the animals in normal isolation cages with free access to food and water. 3. Place the isolation cage (with the animal in it) inside an auditory isolation box. 4. Subject the animals to different sound stimulation protocols over a period of 24 h. 5. Upon completion of the auditory stimulation paradigm, initially anesthetize the animals with 5% isoflurane followed by maintenance with 2% isoflurane and acquire T1 -weighted images. 6. A target region of interest for MEMRI quantification is in the auditory nuclei.

3.2.2.1.2. Imaging Parameters

Different imaging paradigms are possible for activity measurements from the auditory system using MEMRI. Following are optimal imaging parameters for the acquisition of T1 -weighted images of the mouse brain at 7T. These parameters are adapted from the work of Daniel Turnbull and associates (48) studying nerve activity in the auditory pathway of mice and are meant to be a starting guide. Further optimization is required depending upon the animal model used as well as the magnetic field strength. 3D gradient echo pulse sequence TR = 50 ms TE = 4 ms Flip angle = 65◦ Total imaging time = 1 h 50 min per mouse This sequence yields a volumetric image set covering the whole brain, with an isotropic resolution of 100 μm

3.2.2.1.3. Expected Results

With the described auditory stimulation protocol combined with a systemic MnCl2 administration, one can expect to observe significant Mn2+ enhancement in structures of the auditory system, including the auditory nuclei in the brainstem and the thalamus (Fig. 7.3). Although the auditory cortex would also be expected to show enhancement, this particular protocol does not lead to

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

159

Fig. 7.3. MEMRI enhancement in brainstem auditory nuclei was altered in mice with conductive hearing loss (CHL). Comparisons of individual mice with bilateral CHL (bi-CHL) (a), mice with unilateral CHL (uni-CHL) (b), and normal mice (c) demonstrated marked differences in MEMRI signals in the cochlear nucleus (CN) (arrow heads) and inferior colliculus (IC) (arrows), but not in non-auditory caudate putamen (CPu). Adapted by permission from (48), copyright 2005.

signal enhancement in the auditory cortex, perhaps due to less Mn2+ reaching its remote location (48). Nonetheless, this technique is a very useful approach for studying neural activation in the auditory system and can be applied to different animal models of diseases involving the auditory system (e.g. Fig. 7.3). 3.2.2.2. Activation Studies in the Visual System 3.2.2.2.1. MnCl2 Administration

1. Prior to the start of the procedure, house the animals in darkness for 8–12 h. All subsequent procedures and animal handling should be performed in darkness or under very dim red light. 2. Administer 66 mg/kg MnCl2 as an intraperitoneal injection. 3. Place the animal in the stimulation chamber for 8 h. Stimulation consists of a moving square wave (as opposed to constant diffuse light) to avoid habituation. 4. Upon completion of the stimulation paradigm, anesthetize the animals with isoflurane or urethane and acquire T1 weighted MR images.

3.2.2.2.2. Imaging Parameters

Different imaging paradigms are possible for activity measurements from the visual system using MEMRI. Following are two sets of optimal imaging parameters for the acquisition of T1 weighted images of the mouse or rat brain at 4.7T. These parameters are adapted from the work of Bruce Berkowitz and associates (70, 80) studying visual system activity in awake animals and are meant to be a starting guide. Further optimization is required depending on the animal model used as well as the magnetic field strength.

160

Massaad and Pautler

3.2.2.2.3. Mouse Brain

Adiabatic spin echo sequence TR = 350 s TE = 16.7 ms Number of acquisitions = 16 Sweep width = 61,728 Hz Matrix size = 512 × 512 ST = 620 μm FOV = 12 × 12 mm2

3.2.2.2.4. Rat Brain

Rapid Acquisition with Relaxation Enhancement (RARE) sequence TR = 330 ms TE = 16.6 ms RARE factor = 8 Number of acquisitions = 2 Matrix size = 256 × 256 × 173 FOV = 3.84 × 3.84 cm2 ST = 150 μm Time = 80 min/image

3.2.2.2.5. Expected Results

With the aforementioned protocol, one can expect to detect Mn2+ enhancement consistent with layer-specific visual cortex activity in awake and free-moving animals. Layers of a given cortical region respond differently to sensory stimulation and this MEMRI protocol appears to be sensitive enough to detect subtle changes in layer-specific activity (70, 80).

3.2.3. Activation Studies Without BBB Disruption – Olfactory System 3.2.3.1. MnCl2 Administration

Two current Mn2+ exposure modalities exist for activation studies from the olfactory system: Intranasal administration of MnCl2 1. Anesthetize the animal with 5% isoflurane. 2. Pipet 7 μl of a 10 mM MnCl2 solution in each naris. 3. Allow the animal to recover on a heating pad. 4. Place the animal in a clean cage and drop 7 μl of odorant solution in each of the four corners of the cage. 5. Allow odorant exposure for 20 min. 6. Anesthetize the animal with 5% isoflurane for imaging. Exposure to aerosolized MnCl2

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

161

1. Prepare a 1.5 M solution of MnCl2 in H2 O. 2. Place the solution in the heating chamber of a humidifier or vaporizer. For experiments involving the exposure to an odor, mix the odor within the MnCl2 solution. Use a different humidifier for every odor used to avoid crosscontamination between experiments. 3. Place the vaporizer inside a fume hood to avoid exposure to Mn2+ vapors. 4. Animal exposure to the aerosolized MnCl2 with or without odor is performed on either awake or anesthetized animals and is done as follows: • For awake animals: place the animal in the same plastic box housing the humidifier in the hood. Turn the humidifier on for 30 min. Keep the mouse in the box for 1.5 h after the humidifier has been turned off. It is important not to open the box during that time because of possible exposure to the aerosolized Mn2+ still present. • For anesthetized animals: anesthetize the animal with 20 mg/kg urethane and secure it on top of the humidifier with a restraining device. The exposure paradigm involves two sequences of 5-min on and 5-min off, and then the animal is kept for 1.5 h in the chamber with the humidifier off. Again, it is important not to open the box during that time because of possible exposure to the aerosolized Mn2+ still present. 3.2.3.2. Imaging Parameters

Different field strengths will dictate different imaging parameters for best results. The following parameters are adapted from the work of Alan Koretsky and associates (52) for studying olfactory activation in mice at 11.7T. These imaging parameters are meant to be a starting guide; further optimization with regards to field strength and organism used is required. T1 -weighted images acquired by a 3D RARE sequence TR = 300 ms TE = 10 ms Matrix size = 128 × 128 × 64 RARE factor = 2 Isotropic spatial resolution = 100 μm

3.2.3.3. Expected Results

The expected results from the activity-dependent olfactory tract tracing are a gradual increase in signal enhancement ranging from the olfactory epithelium to the olfactory bulbs. Signal enhancement will follow a region-specific enhancement depending on the stimulating odorant used (Fig. 7.4).

162

Massaad and Pautler

Fig. 7.4. Detecting odor-dependent Mn2+ enhancement in mouse olfactory bulb by MRI. MEMRI maps after stimulation by acetophenone, carvone, octanal, and control in four mice, respectively, show distributed enhancement in the glomerular layer with each odorant having its own distinct spatial pattern. High signal change at the interface between the olfactory nerve layer and olfactory turbinates (arrow) indicates where Mn2+ flowed in. Scale bars represent 1 mm. Reprinted from Chuang et al. (52), copyright 2009, with permission from Elsevier.

3.3. Tract Tracing

Tract tracing takes advantage of the following two properties of Mn2+ : (1) it is transported along neurons by microtubuledependent axonal transport and (2) it can traverse synapses and reach second-order neurons leading to contrast enhancement of the whole neuronal system in question. Tract-tracing studies have been performed in several systems such as the visual and olfactory systems as well as from deep brain structures such as the hippocampus and amygdala.

3.3.1. Tracing the Visual Pathway

3.3.1.1. MnCl2 Administration

1. Anesthetize the animal initially with 5% isoflurane and then maintain it with 2% isoflurane. 2. Place the animal in the prone position on a heating pad to maintain body temperature 3. Gently detach the metallic piece of the 27-gauge needle from its plastic hub using forceps. Connect the metallic portion of the needle to the Hamilton syringe via a small piece of polyethylene tubing. 4. Insert the tip of the needle into the vitreous with the aid of a microscope. A good injection site is about 2 mm posterior to the dorsal limbus. 5. Inject 0.1 μl of the MnCl2 solution over 5 min. The volume injected can be gauged by the advancement of the meniscus in the polyethylene tube using the scale of the Hamilton syringe. 6. Leave the injection needle in the eye for at least 15 min and then withdraw it very slowly to minimize the loss of MnCl2 through leakage from the injection site. This waiting time is necessary to insure homogenous distribution of the MnCl2

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

163

inside the eye as well as for the intraocular pressure to reach equilibrium. 7. Terminate anesthesia and return the animals to their cages. 8. Following the MnCl2 injection and prior to imaging, it is advised to check the integrity of the eyes. Successful injection can be ascertained by a bright-looking vitreal humor on a T1 -weighted image. 3.3.1.2. Imaging Parameters

Imaging parameters will vary depending on the animal model used as well as on the magnetic field strength. Following are the optimal imaging parameters used by Watanabe et al. to trace the visual pathway of rats at 2.35T (62). These parameters should be used as a starting guide and further optimization with regards to field strength performed for best results. T1 -weighted 3D FLASH gradient echo sequence TR = 15 ms TE = 4.2 ms Flip angle = 25◦ FOV = 50 × 50 × 16 mm Matrix size = 256 × 256 × 128 NEX = 8 Acquisition time = 65.5 min.

3.3.1.3. Expected Results

Signal enhancement is expected to be seen in the entire visual pathway, starting from the eye and extending to the superior colliculus. An example of such enhancement is illustrated in Fig. 7.5 (62).

3.3.2. Tracing the Olfactory Pathway 3.3.2.1. MnCl2 Administration

1. Anesthetize the animal with 5% isoflurane. 2. Hold animal in a vertical position by slightly pinching the hair in the back of the head. 3. Using a 10-μl pipette, administer 2 μl of a 3.9 M MnCl2 solution to each naris. The 2 μl can be either administered as 2 × 1 μl or at once. It is normal to observe some bubbling from the nose following the nasal lavage. 4. Place the animal on a heating pad to accelerate recovery. Usually it only takes a few minutes for the animal to regain consciousness. 5. Return the animal to the housing cage. 6. Proper lavage can be ascertained by very dark-looking turbinates on a T1 -weighted image (due to T2 effects of the concentrated Mn2+ solution).

164

Massaad and Pautler

Fig. 7.5. Signal enhancement of the rat visual pathway (24 h after Mn2+ -injection into the left eye) in oblique sections 235◦ (top left), 210◦ (top right), 15◦ (bottom left), and 137.5◦ (bottom right) relative to a transverse reference plane. Enhanced structures are (1) left retina, (2) left optic nerve, (3) optic chiasm, (4) right optic tract, (5) right lateral geniculate nucleus, (6) right brachium of the superior colliculus, (7) right pretectal region, and (8) right superior colliculus. Reprinted from (62), copyright 2001, with permission from John Wiley & Sons, Inc.

3.3.2.2. Imaging Parameters

Based upon the work of Pautler et al (3), the optimal imaging parameters for tracing the olfactory system of the mouse at 7T are as follows: T1 -weighted multi-slice spin echo sequence TR = 307 ms TE = 12.7 ms FOV = 2.5 cm ST = 1 mm Matrix size = 128 × 128 Higher resolution 3D scans can be acquired using the following parameters: TR = 300 ms TE = 8.7 ms

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

165

FOV = 2.5 × 2.5 × 2.5 cm Matrix size = 128 × 128 × 128 3.3.2.3. Expected Results

The expected results are a positive contrast enhancement in the olfactory bulbs as well as the primary olfactory cortex as illustrated in Fig. 7.6 (3).

Fig. 7.6. a Three sagittal slices of a mouse treated with Mn2+ in the naris from a representative 3D T1 -weighted MRI sequence. Note the highlighting of the olfactory bulb as well as the primary olfactory cortex leading from the bulbs. b Four axial slices from the same mouse treated with Mn2+ in the naris from a 3D T1 -weighted MRI sequence. Note the highlighting of the outer layers of the olfactory bulbs where the olfactory glomeruli are located. In addition, the enhanced contrast continues caudally into the primary olfactory cortex. Due to the length of the scan, mice were sacrificed before 3D imaging. Reprinted from (3), copyright 1998, with permission from John Wiley & Sons, Inc.

3.3.3. Tracing of Deep Brain Structures

3.3.3.1. Injection Site Coordinates

1. Identify the brain region you wish to inject.

3.3.3.2. Preparing the Injection Needle

1. Using a pipet puller and a quartz capillary tube with filament, pull the tube to create the injection needle. A long, fine-tip needle is needed. Micropipette pullers use a

2. Utilizing a stereotaxic brain atlas, determine the stereotaxic coordinates of the region of interest. This region will most likely encompass multiple slices. Be sure to choose the coordinates that correspond with the largest region in the structure of interest if possible. Additionally, it should be noted that the stereotaxic coordinates will vary based upon sex, age, and animal strain.

166

Massaad and Pautler

combination of heat (laser output power), filament (scan size) velocity (determines the point at which the heat is turned off), delay (the time between deactivation of the laser and the application of a hard pull), and pull (final hard pull applied on the capillary tube) to create needles of different shapes and lengths. Keeping in mind that different pullers operate under different settings (same pullers may even differ depending on the type of filament they have), the following parameters are good starting points: Heat=700; Filament=3; Velocity=60; Delay=140; Pull=175 2. Using fine forceps, gently break the tip of the needle to open it (the needle comes out sealed from the puller). 3. Place the needle in the picospritzer holder. 4. Set the picospritzer to the “pull” option and slowly fill the needle with the MnCl2 solution. 5. Set the prepared needle aside and proceed to preparing the animal for surgery. 6. It is advisable to pull and fill several needles, in case one breaks during the surgery. Store the filled needles in a humidified chamber to prevent crystallization of the MnCl2 solution. 3.3.3.3. Surgery

1. Anesthetize the animal with pentobarbital sodium or ketamine/xylazine combination (Note 11). 2. Upon lack of toe pinch reflex, place the animal on a warming blanket and clip the hair on the back of the head (the area extending from between the ears to the start of the back) (see Note 1). 3. Fix the animal’s head in a stereotaxic holder complete with a bite bar and cheek/ear bars. 4. Maintain anesthesia with 2% isoflurane. 5. Clean the operating field with the disinfecting solution chlorhexidine alternating with sterile water (three times). 6. Make a vertical incision extending from the nose to the start of the back; hold the skin open with hemostats. 7. With the help of the leveling tool and the microscope, make sure that the head is leveled properly both in the longitudinal and horizontal directions. 8. Using the stereotaxic device, determine the coordinates of your animal’s Bregma. 9. Calculate the placement of your region of interest with regards to the Bregma coordinates. For example, if your region of interest mesolateral position was –4.2 and your

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

167

animal’s Bregma was located at the mesolateral coordinate of 34.6, then you will need to place your holder at the 30.4 position (34.6–4.2). The same type of calculation applies to the other coordinates. 10. Mark the 2D location of the injection site by making a small scrape with the tip of a 27-gauge needle. 11. Begin drilling carefully at that location. Be sure to only drill a hole into the skull (e.g., mouse skull is less than 1 mm thick). The drill bit should not go through the dura or brain tissue. 12. Attach the volume gauge to the pre-filled needle with a small piece of modeling clay. 13. Lower the filled needle into the drilled hole. Carefully let the tip of the needle pierce the dura. Keep lowering until you reach the pre-calculated depth coordinate of your region of interest. 14. Adjust the microscope focus on the gauge so that the injected volume can be monitored. 15. Set the injection pressure to 20 psi approximately and begin with an injection time of 5 ms. Gradually increase the injection time until you see the meniscus slightly move within the needle. 16. Inject the full volume using this setting. Typically 10– 20 nl is suitable for tracing from deep brain structures (see Note 3). 17. Leave the injection needle in for at least 5 min and then withdraw it very slowly. This step is necessary to avoid the backflow of MnCl2 through the injection canal. 18. Place a few drops of analgesic such as bupivicaine just underneath the scalp and away from the drill hole and then suture the wound with 6-0 nylon suture. 3.3.3.4. Imaging Parameters

Different imaging paradigms are possible for tracing from deep brain structures following setereotaxic Mn2+ injections. Both 2D and 3D protocols can be used (3D recommended). Following are optimal imaging parameters for the acquisition of T1 -weighted images of the guinea pig brain at 3T. These parameters are adapted from the work of Lee et al. for tracing the auditory pathway in guinea pigs (63) and are meant to be a starting guide. Further optimization is required depending on the organism used as well as on the magnetic field strength. Two-dimensional spin echo sequence TR = 450 ms TE = 13 ms

168

Massaad and Pautler

Matrix size = 256 × 256 FOV = 50 × 50 mm ST = 1.5 mm (with a slice gap of 0.1 mm) NEX = 10 Three-dimensional gradient echo sequence TR = 10.2 ms TE = 2.5 ms Flip angle = 30◦ Matrix size = 256 × 256 × 128 FOV = 50 × 50 × 50 mm NEX = 7 3.3.3.5. Expected Results

The expected results are a multi-synaptic tract tracing to all structures involved in the system that is peripherally injected. For example, following injection of Mn2+ to the cochlea (63), signal enhancement can be observed in the entire auditory pathway, including the cochlear nucleus, the lateral lemniscus, the inferior colliculus, the medial geniculate nucleus, and the trigeminal tract (Fig. 7.7).

Fig. 7.7. T1 -weighted, 2D spin-echo MR image (A) before MnCl2 administration and T1 -weighted, 3D gradient-echo image (B) after 12 h of MnCl2 administration at the left cochlea in the guinea pig. The images’ orientation was obtained at the coronal section, and the voxel resolution was 195×195×200 μm (3). The post-injection image shows signal enhancement of the auditory pathway. Enhanced structures are as follows: (a) cochlear nucleus (CN), (b) lateral lemniscus (LL), (c) inferior colliculus (IC), (d) medial geniculate nucleus (MGN), and (e) trigeminal tract (TT). Reprinted from Lee et al. (63), copyright 2007, with permission from Elsevier.

3.4. Concluding Remarks

As delineated by the multitude of techniques described in this chapter, MEMRI is undoubtedly a very useful technique for the study of the brain anatomy and activity. Perhaps the most

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI)

169

important aspect of MEMRI is that it is minimally invasive and offers the possibility of longitudinal studies. This is of utmost importance for efficient diagnosis and understanding of disease states. Although MEMRI has been quite developed and refined in several organ systems (37, 53) of various animal models such as rodents (3, 44, 81, 82), song-birds (56–58), and non-human primates (55, 83), its use in humans is still almost non-existent. To date, Mn2+ has been approved for clinical imaging only in its chelated form (84). This is primarily due to the toxicity associated with a high concentration of free Mn2+ ions. High concentrations of Mn2+ have been shown to cause acute cardiovascular depression (85) as well as neurodegenerative damage to the nervous system (4). Many efforts are currently focused on developing Mn2+ contrast agents lacking the traditional side effects of Mn2+ . One such agent, available from Eagle Vision Pharmaceuticals, consists of free Mn2+ ions formulated with Ca2+ to override the transient effect of Mn2+ as a Ca2+ competitive inhibitor. This agent is currently used in dogs and pigs for cardiac (86, 87) as well as for vascular imaging (88). The development of such agent shows promise for the imminent use of Mn2+ as a clinical contrast agent for cardiac and brain imaging.

4. Notes 1. For fur trimming in the rat, a conventional rodent hair clipper can be used. For the mouse, it is recommended to use small sharp scissors. Slightly pull on the skin in the opposite direction of the fur growth and then cut the hair as close as possible to the skin, taking care not to injure the skin in the process. 2. It is necessary to use a “thinned out” PE50 tube as opposed to a smaller tube such as PE10. Use of PE10 tubing may not allow successful contrast agent injection because of high back pressure. 3. To construct a calibrated volume gauge, use Photoshop (or an equivalent drawing software) to draw a vertical line and add horizontal graduations to it that are separated by 1 pixel. Print the pattern on clear plastic (such as transparencies), cut it down to its proper size, and attach it to the injection syringe by means of a small piece of modeling clay. On such a scale and using 1-mm quartz capillary tubes, each graduation will correspond to 10 nl of fluid. 4. Several different concentrations of Mn2+ have been used in systemic administration studies. Doses ranged from

170

Massaad and Pautler

6.6 mg/kg to 175 mg/kg. With the higher doses, temporary side effects may occur; however, those effects resolve completely within an hour of MnCl2 administration. Many experimental factors such as the concentration of MnCl2 , the rate of infusion, the route of administration as well as the type, and level of anesthesia play a critical role in a successful MnCl2 administration. 5. An alternative to the direct oral intubation for artificial ventilation is the performance of a tracheotomy. This is however not recommended for most experimental paradigms as it adds more surgical trauma to the animal. 6. α-Chloralose/urethane combination dose: 25 mg/kg αchloralose with 450 mg/kg urethane intraperitoneal injection. 7. Once the tube is properly inserted into the tracheal opening, it is relatively stable. To avoid any possible dislodging of the tube, it can be secured with 6-0 surgical nylon sutures to the front incisors of the animal 8. When the catheter is introduced in the external carotid artery, care should be taken to introduce it in the direction of the common carotid artery, so that blood flow to the inferior carotid artery remains undisturbed. 9. When administering MnCl2 systemically through the venous system, an alternative to the tail vein catheterization is femoral vein catheterization. Both techniques have been shown to be equally effective. 10. Mannitol can recrystallize in solution and cause microinfarcts to the animal. To minimize the occurrence of re-crystallization, use pre-warmed tools such as syringes, saline, tubing etc. at 45◦ C. Flush the catheter with warm saline prior to mannitol administration and keep the room temperature above 25◦ C. Additionally, a 0.22-μm filter can be used and should be connected as close as possible to the external carotid artery. 11. Sodium pentobarbital dose: 50 mg/kg intraperitoneal injection. Ketamine/xylazine combination dose: 7.5 mg/kg ketamine with 5 mg/kg xylazine.

References 1. London, R. E., Toney, G., Gabel, S. A., Funk, A. Magnetic resonance imaging studies of the brains of anesthetized rats treated with manganese chloride. Brain Res Bull 1989;23:229–235.

2. Lin, Y. J., Koretsky, A. P. Manganese ion enhances T1-weighted MRI during brain activation: An approach to direct imaging of brain function. Magn Reson Med 1997;38:378–388.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) 3. Pautler, R. G., Silva, A. C., Koretsky, A. P. In vivo neuronal tract tracing using manganese-enhanced magnetic resonance imaging. Magn Reson Med 1998;40(5): 740–748. 4. Santamaria, A. B. Manganese exposure, essentiality & toxicity. Indian J Med Res 2008;128:484–500. 5. Hurley, L. S., Keen, C. L. Manganese. In: Trace Elements in Human Health and Animal Nutrition, Underwood, E., Mertz, W. (eds.) New York, NY: Academic Press; 1987, pp 185–225. 6. Zwingmann, C., Leibfritz, D., Hazell, A. S. Brain energy metabolism in a subacute rat model of manganese neurotoxicity: An ex vivo nuclear magnetic resonance study using [1–13 c]glucose. Neurotoxicology 2004;25:573–587. 7. Patchett, M. L., Daniel, R. M., Morgan, H. W. Characterisation of arginase from the extreme thermophile ‘Bacillus caldovelox’. Biochim Biophys Acta 1991;1077: 291–298. 8. Wedler, F. C., Denman, R. B. Glutamine synthetase: The major Mn(II) enzyme in mammalian brain. Curr Top Cell Regul 1984;24:153–169. 9. Keen, C. Nutritional and toxicological aspects of manganese intake: an overview. In: Risk Assessment of Essential Elements, Mertz W., Abernathy C., Olin S. (eds.) Washington, DC: ILSI Press; 1994. 10. Strause, L. G., Hegenauer, J., Saltman, P., Cone, R., Resnick, D. Effects of longterm dietary manganese and copper deficiency on rat skeleton. J Nutr 1986;116: 135–141. 11. Friedman, B. J., Freeland-Graves, J. H., Bales, C. W., Behmardi, F., Shorey-Kutschke, R. L., Willis, R. A., Crosby, J. B., Trickett, P. C., Houston, S. D. Manganese balance and clinical observations in young men fed a manganese-deficient diet. J Nutr 1987;117:133–143. 12. Venugopal, B., Luckey, T. Metal Toxicity in Mammals. Chemical Toxicity of Metals and Metalloids. New York, NY: Plenum Press; 1978. 13. Aschner, M., Guilarte, T. R., Schneider, J. S., Zheng, W. Manganese: Recent advances in understanding its transport and neurotoxicity. Toxicol Appl Pharmacol 2007;221: 131–147. 14. Crossgrove, J., Zheng, W. Manganese toxicity upon overexposure. NMR Biomed 2004;17:544–553. 15. Sloot, W. N., Gramsbergen, J. B. Axonal transport of manganese and its relevance to

16.

17.

18. 19.

20.

21.

22.

23.

24.

25. 26. 27.

28.

29.

171

selective neurotoxicity in the rat basal ganglia. Brain Res 1994;657:124–132. Takeda, A., Kodama, Y., Ishiwatari, S., Okada, S. Manganese transport in the neural circuit of rat CNS. Brain Res Bull 1998;45:149–152. Tjälve, H., Mejàre, C., Borg-Neczak, K. Uptake and transport of manganese in primary and secondary olfactory neurones in pike. Pharmacol Toxicol 1995;77:23–31. Takeda, A., Ishiwatari, S., Okada, S. In vivo stimulation-induced release of manganese in rat amygdala. Brain Res 1998;811:147–151. Barbeau, A. Manganese and extrapyramidal disorders (a critical review and tribute to dr. George C. Cotzias). Neurotoxicology 1984;5:13–35. Mena, I., Marin, O., Fuenzalida, S., Cotzias, G. C. Chronic manganese poisoning. Clinical picture and manganese turnover. Neurology 1967;17:128–136. Finley, J. W. Manganese uptake and release by cultured human hepato-carcinoma (hepG2) cells. Biol Trace Elem Res 1998;64: 101–118. Kerper, L. E., Hinkle, P. M. Cellular uptake of lead is activated by depletion of intracellular calcium stores. J Biol Chem 1997;272:8346–8352. Mason, M. J., Mayer, B., Hymel, L. J. Inhibition of Ca2+ transport pathways in thymic lymphocytes by econazole, miconazole, and SKF 96365. Am J Physiol 1993;264: C654–C662. Murphy, V. A., Smith, Q. R., Rapoport, S. I. Saturable transport of ca into CSF in chronic hypo- and hypercalcemia. J Neurosci Res 1991;30:421–426. Takeda, A. Manganese action in brain function. Brain Res Brain Res Rev 2003;41:79–87. Chance, B. The energy-linked reaction of calcium with mitochondria. J Biol Chem 1965;240:2729–2748. Huang, C., Cheng, H., Lin, K., Cheng, J., Tsai, J., Liao, W., Fang, Y., Jan, C. Tamoxifen-induced [Ca2+ ]i rise and apoptosis in corneal epithelial cells. Toxicology 2009;255:58–64. Tas, P. W. L., Stössel, C., Roewer, N. Inhibition of the histamine-induced Ca2+ influx in primary human endothelial cells (HUVEC) by volatile anaesthetics. Eur J Anaesthesiol 2008;25:976–985. Merritt, J. E., Jacob, R., Hallam, T. J. Use of manganese to discriminate between calcium influx and mobilization from internal stores in stimulated human neutrophils. J Biol Chem 1989;264:1522–1527.

172

Massaad and Pautler

30. Narita, K., Kawasaki, F., Kita, H. Mn and Mg influxes through Ca channels of motor nerve terminals are prevented by verapamil in frogs. Brain Res 1990;510:289–295. 31. Simpson, P. B., Challiss, R. A., Nahorski, S. R. Divalent cation entry in cultured rat cerebellar granule cells measured using Mn2+ quench of fura 2 fluorescence. Eur J Neurosci 1995;7:831–840. 32. Cory, D. A., Schwartzentruber, D. J., Mock, B. H. Ingested manganese chloride as a contrast agent for magnetic resonance imaging. Magn Reson Imaging 1987;5:65–70. 33. Geraldes, C. F., Sherry, A. D., Brown, R. D., Koenig, S. H. Magnetic field dependence of solvent proton relaxation rates induced by Gd3+ and Mn2+ complexes of various polyaza macrocyclic ligands: implications for NMR imaging. Magn Reson Med 1986;3:242–250. 34. Fornasiero, D., Bellen, J. C., Baker, R. J., Chatterton, B. E. Paramagnetic complexes of manganese(II), iron(III), and gadolinium(III) as contrast agents for magnetic resonance imaging. The influence of stability constants on the biodistribution of radioactive aminopolycarboxylate complexes. Invest Radiol 1987;22:322–327. 35. Mendonça-Dias, M. H., Gaggelli, E., Lauterbur, P. C. Paramagnetic contrast agents in nuclear magnetic resonance medical imaging. Semin Nucl Med 1983;13:364–376. 36. Pautler, R. G. Biological applications of manganese-enhanced magnetic resonance imaging. Methods Mol Med 2006;124: 365–386. 37. Silva, A. C., Lee, J. H., Aoki, I., Koretsky, A. P. Manganese-enhanced magnetic resonance imaging (MEMRI): Methodological and practical considerations. NMR Biomed 2004;17:532–543. 38. Natt, O., Watanabe, T., Boretius, S., Radulovic, J., Frahm, J., Michaelis, T. Highresolution 3D MRI of mouse brain reveals small cerebral structures in vivo. J. Neurosci Methods 2002;120:203–209. 39. Watanabe, T., Natt, O., Boretius, S., Frahm, J., Michaelis, T. In vivo 3D MRI staining of mouse brain after subcutaneous application of MnCl2. Magn Reson Med 2002;48: 852–859. 40. Watanabe, T., Radulovic, J., Spiess, J., Natt, O., Boretius, S., Frahm, J., Michaelis, T. In vivo 3D MRI staining of the mouse hippocampal system using intracerebral injection of MnCl2. Neuroimage 2004;22:860–867. 41. Watanabe, T., Radulovic, J., Boretius, S., Frahm, J., Michaelis, T. Mapping of the habenulo-interpeduncular pathway in living

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

mice using manganese-enhanced 3D MRI. Magn Reson Imaging 2006;24:209–215. Bock, N. A., Paiva, F. F., Nascimento, G. C., Newman, J. D., Silva, A. C. Cerebrospinal fluid to brain transport of manganese in a non-human primate revealed by MRI. Brain Res 2008;1198:160–170. Silva, A. C., Lee, J. H., Wu, C. W., Tucciarone, J., Pelled, G., Aoki, I., Koretsky, A. P. Detection of cortical laminar architecture using manganese-enhanced MRI. J Neurosci Methods 2008;167:246–257. Aoki, I., Wu, Y. L., Silva, A. C., Lynch, R. M., Koretsky, A. P. In vivo detection of neuroarchitecture in the rodent brain using manganese-enhanced MRI. Neuroimage 2004;22:1046–1059. Deans, A. E., Wadghiri, Y. Z., Berrios-Otero, C. A., Turnbull, D. H. Mn enhancement and respiratory gating for in utero MRI of the embryonic mouse central nervous system. Magn Reson Med 2008;59:1320–1328. Duong, T. Q., Silva, A. C., Lee, S. P., Kim, S. G. Functional MRI of calcium-dependent synaptic activity: Cross correlation with CBF and BOLD measurements. Magn Reson Med 2000;43:383–392. Aoki, I., Tanaka, C., Takegami, T., Ebisu, T., Umeda, M., Fukunaga, M., Fukuda, K., Silva, A. C., Koretsky, A. P., Naruse, S. Dynamic activity-induced manganesedependent contrast magnetic resonance imaging (DAIM MRI). Magn Reson Med 2002;48:927–933. Yu, X., Wadghiri, Y. Z., Sanes, D. H., Turnbull, D. H. In vivo auditory brain mapping in mice with Mn-enhanced MRI. Nat Neurosci 2005;8(7):961–968. Parkinson, J. R. C., Chaudhri, O. B., Bell, J. D. Imaging appetite-regulating pathways in the central nervous system using manganeseenhanced magnetic resonance imaging. Neuroendocrinology 2009;89:121–130. Weng, J., Chen, J., Yang, P., Tseng, W. I. Functional mapping of rat barrel activation following whisker stimulation using activity-induced manganese-dependent contrast. Neuroimage 2007;36:1179–1188. Lu, H., Xi, Z., Gitajn, L., Rea, W., Yang, Y., Stein, E. A. Cocaine-induced brain activation detected by dynamic manganese-enhanced magnetic resonance imaging (MEMRI). Proc Natl Acad Sci USA 2007;104: 2489–2494. Chuang, K., Lee, J. H., Silva, A. C., Belluscio, L., Koretsky, A. P. Manganese enhanced MRI reveals functional circuitry in response to odorant stimuli. Neuroimage 2009;44:363–372.

Manganese-Enhanced Magnetic Resonance Imaging (MEMRI) 53. Wendland, M. F. Applications of manganeseenhanced magnetic resonance imaging (MEMRI) to imaging of the heart. NMR Biomed 2004;17:581–594. 54. Murayama, Y., Weber, B., Saleem, K. S., Augath, M., Logothetis, N. K. Tracing neural circuits in vivo with Mn-enhanced MRI. Magn Reson Imaging 2006;24: 349–358. 55. Saleem, K. S., Pauls, J. M., Augath, M., Trinath, T., Prause, B. A., Hashikawa, T., Logothetis, N. K. Magnetic resonance imaging of neuronal connections in the macaque monkey. Neuron 2002;34:685–700. 56. Tindemans, I., Verhoye, M., Balthazart, J., Van Der Linden, A. In vivo dynamic MEMRI reveals differential functional responses of RA- and area X-projecting neurons in the HVC of canaries exposed to conspecific song. Eur J Neurosci 2003;18: 3352–3360. 57. Van der Linden, A., Verhoye, M., Van Meir, V., Tindemans, I., Eens, M., Absil, P., Balthazart, J. In vivo manganese-enhanced magnetic resonance imaging reveals connections and functional properties of the songbird vocal control system. Neuroscience 2002;112:467–474. 58. Van der Linden, A., Van Meir, V., Tindemans, I., Verhoye, M., Balthazart, J. Applications of manganese-enhanced magnetic resonance imaging (MEMRI) to image brain plasticity in song birds. NMR Biomed 2004;17:602–612. 59. Van Meir, V., Verhoye, M., Absil, P., Eens, M., Balthazart, J., Van der Linden, A. Differential effects of testosterone on neuronal populations and their connections in a sensorimotor brain nucleus controlling song production in songbirds: A manganese enhanced-magnetic resonance imaging study. Neuroimage 2004;21:914–923. 60. Pautler, R. G. In vivo, trans-synaptic tracttracing utilizing manganese-enhanced magnetic resonance imaging (MEMRI). NMR Biomed 2004;17:595–601. 61. Pautler, R. G., Koretsky, A. P. Tracing odor-induced activation in the olfactory bulbs of mice using manganese-enhanced magnetic resonance imaging. Neuroimage 2002;16:441–448. 62. Watanabe, T., Michaelis, T., Frahm, J. Mapping of retinal projections in the living rat using high-resolution 3D gradient-echo MRI with Mn2+ -induced contrast. Magn Reson Med 2001;46(3):424–429. 63. Lee, J., Park, J., Lee, J., Bae, S., Lee, S., Jung, J., Kim, M., Lee, J., Woo, S., Chang, Y. Manganese-enhanced auditory tract-tracing

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

173

MRI with cochlear injection. Magn Reson Imaging 2007;25:652–656. Serrano, F., Deshazer, M., Smith, K. D. B., Ananta, J. S., Wilson, L. J., Pautler, R. G. Assessing transneuronal dysfunction utilizing manganese-enhanced MRI (MEMRI). Magn Reson Med 2008;60:169–175. Smith, K. D. B., Kallhoff, V., Zheng, H., Pautler, R. G. In vivo axonal transport rates decrease in a mouse model of alzheimer’s disease. Neuroimage 2007;35:1401–1408. Aoki, I., Naruse, S., Tanaka, C. Manganeseenhanced magnetic resonance imaging (MEMRI) of brain activity and applications to early detection of brain ischemia. NMR Biomed 2004;17:569–580. Kidoguchi, K., Tamaki, M., Mizobe, T., Koyama, J., Kondoh, T., Kohmura, E., Sakurai, T., Yokono, K., Umetani, K. In vivo X-ray angiography in the mouse brain using synchrotron radiation. Stroke 2006;37: 1856–1861. Zhao, X., Wu, N., Deng, M., Yin, Y., Zhou, J., Fang, Y., Huang, L. An improved method of left ventricular catheterization in rats. Physiol Meas 2006;27:N27–N33. Brown, R. H., Walters, D. M., Greenberg, R. S., Mitzner, W. A method of endotracheal intubation and pulmonary functional assessment for repeated studies in mice. J Appl Physiol 1999;87:2362–2365. Bissig, D., Berkowitz, B. A. Manganeseenhanced MRI of layer-specific activity in the visual cortex from awake and free-moving rats. Neuroimage 2009;44:627–635. Pautler, R. G., Mongeau, R., Jacobs, R. E. In vivo trans-synaptic tract tracing from the murine striatum and amygdala utilizing manganese enhanced MRI (MEMRI). Magn Reson Med 2003;50:33–39. Watanabe, T., Frahm, J., Michaelis, T. Functional mapping of neural pathways in rodent brain in vivo using manganese-enhanced three-dimensional magnetic resonance imaging. NMR Biomed 2004;17:554–568. Burnett, K. R., Goldstein, E. J., Wolf, G. L., Sen, S., Mamourian, A. C. The oral administration of MnCl2: A potential alternative to IV injection for tissue contrast enhancement in magnetic resonance imaging. Magn Reson Imaging 1984;2:307–314. Kita, H., Narita, K., Van der Kloot, W. Tetanic stimulation increases the frequency of miniature end-plate potentials at the frog neuromuscular junction in Mn2+ -, CO2+ -, and Ni2+ -saline solutions. Brain Res 1981;205:111–121. Drapeau, P., Nachshen, D. A. Manganese fluxes and manganese-dependent neurotrans-

174

76.

77.

78.

79.

80.

81.

82.

83.

Massaad and Pautler mitter release in presynaptic nerve endings isolated from rat brain. J Physiol (Lond) 1984;348:493–510. Rabin, O., Hegedus, L., Bourre, J. M., Smith, Q. R. Rapid brain uptake of manganese(II) across the blood-brain barrier. J Neurochem 1993;61:509–517. Murphy, V. A., Rosenberg, J. M., Smith, Q. R., Rapoport, S. I. Elevation of brain manganese in calcium-deficient rats. Neurotoxicology 1991;12:255–263. Yu, X., Sanes, D. H., Aristizabal, O., Wadghiri, Y. Z., Turnbull, D. H. Largescale reorganization of the tonotopic map in mouse auditory midbrain revealed by MRI. Proc Natl Acad Sci USA 2007;104: 12193–12198. Yu, X., Zou, J., Babb, J. S., Johnson, G., Sanes, D. H., Turnbull, D. H. Statistical mapping of sound-evoked activity in the mouse auditory midbrain using Mn-enhanced MRI. Neuroimage 2008;39: 223–230. Berkowitz, B. A., Gradianu, M., Bissig, D., Kern, T. S., Roberts, R. Retinal ion regulation in a mouse model of diabetic retinopathy: Natural history and the effect of Cu/Zn superoxide dismutase overexpression. Invest Ophthalmol Vis Sci 2009;50:2351–2358. Hu, T. C., Pautler, R. G., MacGowan, G. A., Koretsky, A. P. Manganese-enhanced MRI of mouse heart during changes in inotropy. Magn Reson Med 2001;46:884–890. Lee, J. H., Silva, A. C., Merkle, H., Koretsky, A. P. Manganese-enhanced magnetic resonance imaging of mouse brain after systemic administration of MnCl2: Dose-dependent and temporal evolution of T1 contrast. Magn Reson Med 2005;53:640–648. Bock, N. A., Kocharyan, A., Silva, A. C. Manganese-enhanced MRI visualizes V1 in the non-human primate visual cortex. NMR Biomed 2009, Available at:

84.

85.

86.

87.

88.

http://www.ncbi.nlm.nih.gov/pubmed/ 19322808 [Accessed July 21, 2009]. Federle, M. P., Chezmar, J. L., Rubin, D. L., Weinreb, J. C., Freeny, P. C., Semelka, R. C., Brown, J. J., Borello, J. A., Lee, J. K., Mattrey, R., Dachman, A. H., Saini, S., Harmon, B., Fenstermacher, M., Pelsang, R. E., Harms, S. E., Mitchell, D. G., Halford, H. H., Anderson, M. W., Johnson, C. D., Francis, I. R., Bova, J. G., Kenney, P. J., Klippenstein, D. L., Foster, G. S., Turner, D. A. Safety and efficacy of mangafodipir trisodium (MnDPDP) injection for hepatic MRI in adults: Results of the U.S. multicenter phase III clinical trials (safety). J Magn Reson Imaging 2000;12: 186–197. Wolf, G. L., Baum, L. Cardiovascular toxicity and tissue proton T1 response to manganese injection in the dog and rabbit. AJR Am J Roentgenol 1983;141:193–197. Storey, P., Chen, Q., Li, W., Seoane, P. R., Harnish, P. P., Fogelson, L., Harris, K. R., Prasad, P. V. Magnetic resonance imaging of myocardial infarction using a manganesebased contrast agent (EVP 1001-1): Preliminary results in a dog model. J Magn Reson Imaging 2006;23:228–234. Storey, P., Danias, P. G., Post, M., Li, W., Seoane, P. R., Harnish, P. P., Edelman, R. R., Prasad, P. V. Preliminary evaluation of EVP 1001-1: A new cardiac-specific magnetic resonance contrast agent with kinetics suitable for steady-state imaging of the ischemic heart. Invest Radiol 2003;38: 642–652. Zuo, C. S., Seoane, P., Lanigan, T., Harnish, P., Prasad, P. V., Storey, P., Li, W., Rofsky, N. M. T1 efficacy of EVP-ABD: A potential manganese-based MR contrast agent for hepatic vascular and tissue phase imaging. J Magn Reson Imaging 2002;16: 668–675.

Chapter 8 Sodium MRI Ronald Ouwerkerk Abstract Sodium (23 Na) imaging has a place somewhere between 1 H-MRI and MR spectroscopy (MRS). Like MRS it potentially provides information on metabolic processes, but only one single resonance of ionic 23 Na is observed. Therefore pulse sequences do not need to code for a chemical shift dimension, allowing 23 Na images to be obtained at high resolutions as compared to MRS. In this chapter the biological significance of sodium in the brain will be discussed, as well as methods for observing it with 23 NaMRI. Many vital cellular processes and interactions in excitable tissues depend on the maintenance of a low intracellular and high extracellular sodium concentration. Healthy cells maintain this concentration gradient at the cost of energy. Leaky cell membranes or an impaired energy metabolism immediately leads to an increase in cytosolic total tissue sodium. This makes sodium a biomarker for ischemia, cancer, excessive tissue activation, or tissue damage as might be caused by ablation therapy. Special techniques allow quantification of tissue sodium for the monitoring of disease or therapy in longitudinal studies or preferential observation of the intracellular component of the tissue sodium. New methods and high-field magnet technology provide new opportunities for 23 Na-MRI in clinical and biomedical research. Key words: MRI, sodium, 23 Na, other nuclei, cancer, stroke, brain.

1. Introduction The most exciting aspect of 23 Na-MRI is that the tissue sodium concentrations are very sensitive to changes in the metabolic state of tissues and the integrity of the cell membrane. Cells in healthy tissue actively maintain a large Na concentration gradient across the cell membrane, and almost any impairment of energy metabolism or insult to the cell membrane integrity leads to an increase in intracellular sodium (1). This leads to very significant 23 Na signal intensity changes in cancer, stroke, or myocardial M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_8, © Springer Science+Business Media, LLC 2011

175

176

Ouwerkerk

infarction. In excitable tissues, such as cardiac (1, 2), skeletal muscle (3), and brain, even normal activity can lead to transient changes in sodium (4).

2. The Physiology of Brain Tissue Sodium Content

One limitation of 23 Na-MR is that the sodium MR signal stems from both intracellular and extracellular compartments. There are some techniques that, at the cost of signal to noise or spatial resolution, do provide a weighting of the signal that favors the metabolically more interesting intracellular signal. The combined tissue sodium signal, however, can also provide a lot of information due to the fact that, through tissue perfusion, the extracellular sodium concentration [Na+ ]ex is constant and even in disease rarely deviates from the plasma sodium concentration of about 140 mmol/l. The intracellular sodium concentration [Na+ ]in is much lower, typically 10–15 mM, and this is maintained by active pumping of the Na+ /K+ ATPase, which is powered by ATP. The high difference in concentration of sodium (and potassium) between the intracellular and the extracellular compartments creates a potential that is used for transmitting nerve impulses and for pumping protons and small molecules across the cell membrane. Potassium leaves and sodium enters the cell both during depolarization as a result of a nerve impulse and when for instance protons are pumped out of the cell by the Na+ /H+ exchanger. Normally these ion fluxes are compensated by the exchange of intracellular sodium with extracellular potassium by Na+ /K+ ATPase using hydrolysis of high-energy ATP. In the brain, sodium is not only transported across membranes for the recovery of resting membrane potential but is even more solidly linked to the energetics of brain function through the pivotal role that the transmembrane sodium concentration gradient plays in the uptake of the neurotransmitter glutamate. Astrocytes, wrapped around synaptic contact sites, possess sodium glutamate co-transporters (5, 6), which clear glutamate from the extracellular space. These transporters are powered by the electrochemical gradient of Na+ . The sodium that enters the astrocyte with the glutamate is again pumped out by Na+ /K+ ATPase. Although the mechanisms also involve many other factors, such as Ca2+ (which is also linked to intracellular and extracellular sodium levels), the role of sodium as a link between glutamate and astrocytic energy metabolism is firmly established (7). In cultured fetal mouse astrocytes, it was found that under non-stimulated conditions, the Na+ /K+ ATPase consumes about 20% of astrocytic

Sodium MRI

177

ATP production (8), and sodium levels are maintained at around 10 mmol/l. Just 1 mmol/l extracellular glutamate can evoke a transient threefold increase in [Na+ ]in in primary cultured mouse astrocytes (9) lasting more than 2 min. Clearly, on continued or repeated activation the intracellular sodium level can rise considerably.

3. Measuring Sodium In Vivo The question is whether 23 Na-MRI can be used to observe these changes in vivo. Although the intracellular changes are substantial, the sodium concentration in the extracellular compartment is much higher and this may mask changes in the tissue sodium concentration due to an increase in intracellular sodium. Normally, however, the extracellular space is relatively small, and roughly half of the total observed tissue sodium is from intracellular sodium. Whereas, in spite of this, the normal transient increases in intracellular sodium may be too small and short lived to be observed with the current MR technology, abnormal circumstances or prolonged activation may lead to clearly observable changes in tissue sodium. When the demand for ATP exceeds the production of ATP, the ATP supply for the Na+ /K+ ATPase will be insufficient to maintain or quickly restore the low intracellular sodium concentration and thus a sustained increase of tissue sodium concentration (TSC) can be observed. This phenomenon has been observed in sustained exercise in muscle (3, 10, 11) and when energy supply is interrupted as in an ischemic heart muscle in animal models (12, 13). Due to the large sodium concentration a gradient is maintained across the cell membrane of resting healthy tissue and the influx of sodium can very rapidly more than double the intracellular sodium content. Even if that intracellular change is only half reflected in the tissue sodium content, it is still readily observable by 23 Na-MRI (13). Additionally, various factors, such as the concomitant potassium efflux and acidification, can trigger vasodilation in some tissues and thus lead to an increase in extracellular space and consequently in observed TSC. An increase in intracellular sodium paired with an increase in extracellular partial volume precludes a detailed physiological interpretation of changes in TSC without additional measurements. A 23 Na-MR method that is sensitive to either only intracellular sodium or any MR method that can be used to quantify the intracellular versus extracellular volumes would resolve this issue. Because neither goal can be achieved without introducing new

178

Ouwerkerk

ambiguities, or large signal to noise losses, or agents that cannot be used on human subjects, we still do not have a method to reliably measure [Na+ ]in in human disease. Fortunately, the intracellular and extracellular changes generally do not cancel each other, and thus, the TSC generally turns out to be a good indicator of tissue health or cancer malignancy.

4. 23 Na-MRI from Past to Present The feasibility of 23 Na-MRI on human subjects has been demonstrated as far back as the late 1980s (14), and the image contrast obtainable in cancer (15), edema, and stroke (16, 17) has also been demonstrated for quite some time. Despite this, 23 Na-MRI has yet to find wider acceptance as a possible clinical MR tool. At present, the increased availability of high field MRI systems appears to lead to a revival of 23 Na-MRI evidenced by a recent surge not only in methodological papers on 23 Na-MRI of the brain (18–21) but also in new clinical MR research papers, where 23 Na-MRI is used to study disease in humans (17, 22–26). Reduced scan times now allow the addition of a 23 Na-MRI to existing 1 H-MRI protocols. It is through the merging with a comprehensive 1 H-MRI protocol that the potential of 23 Na-MRI as a diagnostic tool can best be developed. A scanner with higher field strength will offer a better signal to noise, which can be traded for higher resolution or shorter experiment time. A few recent and early results are listed in Table 8.1. The technical advances since the early 1990s have paid off in an increase in signal-to-noise ratio (SNR) and reduction of voxel size and total scan time. From this table, it is also clear that 23 Na-MRI results are better in terms of SNR when scan methods with a short echo time (TE) are used. Brain 23 Na at 3 T with TPI has been reported to yield the expected twofold SNR increase over the same method at 1.5 T. Thus, it should be possible to acquire <100 μl voxel size 3D images of the brain in about 10 min. The very high field of 9.4 T with a 23 Na frequency of 108 MHz yielded very impressive brain sodium images in a scan time of less than 6 min (21). 1 H-MRI at such high fields is very challenging, because of the short wavelength of the 1 H resonance radio frequency (RF) compared to the dimensions of the human body. On the other hand, the RF resonance frequency for 23 NaMRI is comparable to that of 1 H at only 2.5 T. Thus, one motivation for pursuing 23 Na-MRI on a high field scanner might be that you can use relatively simple RF coil systems with proven designs. More importantly, even at 3 T a threshold appears to have been

Sodium MRI

179

Table 8.1 A selection of 23 Na studies of the human brain Year

Field

Scan time

Voxel (mm)

Voxel (μl)

Method

TE

Brain SNR

Reference

2008

4.7 T

11

3×3×3

27

Densityweighted TPI

>450 μ

~22

(18)

2007

1.5 T

10

4×4×4

64

Projection imaging

200 μs

3

(19)

2007

1.5 T

10

4×8×10

320

FLASH

2.7 ms

19

(19)

2003

4T

14:20

3×6×11

194

3D-GRE

1.6 ms

>20 WM

(20)

2007

9.4 T

5:56

3×3×3

27

TPI

260 μs

Unknown (21)

2006

4T

30

7.5×7.5×7.5 423

SPRITE

<0.5 ms

17

(29)

2003

1.5 T

14:52

6×6×6

TPI

400 μs

>20

(36)

∼1–2

(74)

1.5

(75)

220

1987

1.5 T

16:20

4×4×20

280

Multi-echo GRE

13 ms∗

1985

1.3 T

10:30

5×5×10

273

3D FT (3D-GRE)

14 ms

TPI, Twisted projection imaging; GRE, Gradient-recalled-echo; FLASH, Fast low-angle shot; ∗ TE series 13, 26 and 39 ms SPRITE, Single-point ramped imaging with T1 enhancement.

reached, where 23 Na-MRI scan times have been reduced to the point where the addition of a 23 Na-MRI to an existing 1 H-MRI protocol has become a viable option for studies on patients rather than just extremely tolerant volunteers.

5. MR Properties of 23 Na In Vivo The SNR of 23 Na-MRI is much lower than that of 1 H-MRI, mainly because of the difference in abundance between water protons and sodium ions in the body and also because the gyromagnetic ratio is a factor four lower. This lower SNR is partially offset by the short longitudinal relaxation time, T1 , on the order of 25–40 ms, which allows signal averaging with fairly rapid repetition rates, TR. Even when the TR is relatively long to allow quantification of TSC without T1 relaxation corrections acquisition rates (TR > three times T1 ) of about 10 excitations per second are still possible. However, the transverse relaxation time, T2 , of sodium is not helping at all. The sodium T2 is very fast and is bi-exponential in many biological tissues and in gels. This dual exponential decay is a consequence of the spin number of 23 Na, which is 3/2. Protons have a spin number 1/2 and can have two distinct energy

180

Ouwerkerk

levels. Nuclei with spin number 3/2 can have four distinct energy levels with spin number m = −3/2, −1/2, 1/2, and 3/2. There are therefore more possible transitions between the energy levels. Without going into quantum physics too much, we can understand that the energy level in a nuclear spin can change between two adjacent levels, called a single quantum transition, but it can also be transferred to other levels with double or triple quantum transitions. We normally see only single quantum transitions, and in a watery solution all the possible single quantum transitions have the same energy, so we see a simple single line spectrum. In highly ordered restricting environments, such as cartilage or in gels, the gaps between the energy levels for 23 Na start to differ. This is due to anisotropy. When the sodium comes into close contact with the ordered molecules, and the system has a distinct directionality on the molecular level, this can lead to a spectrum with a central peak and two satellites separated from the central peak by the quadrupole frequency. Also the relaxation times associated with the different transitions will differ. The possibility of distinct transitions will lead to a bi-exponential transverse signal decay with a short T2 , on the order of 1–2 ms for 60% of the signal and a longer T2 on the order of 20–30 ms for 40% of the signal. The relaxation time of the satellite transitions is shorter, and therefore these lines in the MR spectrum will be much broader, sometimes so broad that they disappear into the baseline. In the past, this has led to speculations that part of the MR signal in vivo is invisible. Unless the acquisition delay is short this is indeed the case, and even if observed, the short T2 is bound to broaden the points spread function in 23 Na imaging. The images will be a superposition of a sharp and a fuzzy image. Images recorded with TE that are long compared to this short TE may look sharper than short TE images with the same nominal resolution, but this could be remedied by time domain filtering of the short TE images. In most biological tissues, not all of the many different environments in which sodium is present (cytosol, endoplasmatic reticulum, mitochondria, extracellular matrix, blood plasma, cerebro-spinal fluid, etc.) will have the same density of ordered binding sites that lead to bi-exponential relaxation. Thus, except for a homogeneous, highly ordered tissue, such as cartilage, the fraction of the fast relaxing component is much less than 60%. Still, substantial signal losses can occur as a result of T2 relaxation. The T2 for 23 Na in the brain has been determined with a single exponential model and by some even with a bi-exponential model. The results, summarized in Table 8.2, give some idea of what sequence parameters to use for 23 Na-MRI of the brain, and it is clear that CSF and crystalline saline solutions have a long T2 , whereas white matter and gray matter have (slow) T2 s in the order of 17–25 ms. For quantification of TSC, short TE, ideally on the order of 0.2–0.4 ms, is required to reduce those T2 losses

Sodium MRI

Table 8.2 Literature values for transverse animal tumor models

23 Na-MR

Tumor

GM

WM

T2 fast ms (42%)

3.6a

-

-

T2 slow ms (58%)

18a

17–18

17–18

CSF

relaxation rates in human brain and in VH

Saline

57

59

Whole brain

2.1 ± 0.3

T2 slow ms (40%)

20 ± 2.3 4.6 ± 1.8b

T2 slow ms (48%)

23.3 ± 7.6b

T2 ms

Reference

Perman (76)

T2 ast ms (60%)

T2 ast ms (52%)

181

Constantinides (77)

Summers (78) 11

T2 ms

12 32

T2 ms

47

58

53

42

60

21

Winkler (79) Feinberg (75)

57

Perman (80)

GM, gray matter; WM, white matter; CSF, cerebro-spinal fluid; VH, Vitreous humor. a Rabbit VX2 carcinoma. b IMR-5 neuroblastoma in nude mice.

to less than 5–10% of the total signal. In the next section, a few techniques are discussed that will enable the acquisition of 23 Na with minimal signal losses due to T2 relaxation.

6. 23 Na-MR Techniques The gradient-recalled echo technique (GRE) is a good initial choice for 23 Na-MRI, because it is relatively simple to implement. The sequence can be set up by a few simple adaptations from a 1 H-MRI sequence that is available on any scanner platform. The requirement for a short TE, together with the need for small receiver bandwidths (RBW) for better SNR, makes it difficult to use gradient-recalled echo sequences with small T2 losses and good SNR. When using GRE as a method for 23 Na-MRI it is advisable to optimize the SNR by recording a series of images with different RBW, all at the minimum achievable TE for that RBW to find the best compromise in terms of SNR. Typically, for a 64-point readout, the optimum at 1.5 or 3 T will be around 4–16 kHz RBW with a TE of about 5 ms for the lowest RBW and 2 ms for 16 kHz RBW. The actual optimal numbers are very

182

Ouwerkerk

scanner specific, as faster gradients will allow a shorter TE for a given RBW. A sample 3D 23 Na GRE image of a healthy volunteer is shown in Fig. 8.1. This scan (after obtaining informed written consent to an IRB-approved protocol) was performed on a 1.5 T GE Signa scanner (GE Medical Systems, Milwaukee, WI) using a sequence that was only minimally modified from the product 1 Hfast GRE sequence. Modifications included the replacement of

Fig. 8.1. Fast GRE 23 Na image of the brain of a healthy volunteer at 1.5 T. Non-selective excitation was used with an otherwise standard 3D-GRE sequence on a GE Signa scanner (GE Medical Systems, Milwaukee, WI), recording 64 points at 8 kHz RBW with a TR/TE = 30.9/5 ms, 32 slices (not all shown), and 12 averages for 7.5×7.5×15 mm voxel size in a total scan time of <4 min. Two tubes with reference NaCl solutions of 50 and 100 mmol/l were placed next to the head. In the second image from the left in the second row from the bottom the mean SNR in a 5 cm circle in the center of the brain was 14.

Sodium MRI

183

the slice-selective pulse by a block pulse (an adiabatic half-passage pulse can also be used) and reducing the minimum number of readout and phase-encoding points to 64 and 32, respectively. This 7.5 × 7.5 × 15-mm voxel size data set was recorded at 8 kHz RBW with a 60◦ flip angle and TR/TE = 30.9/4.9 ms in less than 4 min and yielded a brain SNR of about 14.

7. Toward a Shorter TE If we assume that a substantial part of the brain sodium signal has bi-exponential relaxation, a TE in the order of 4–5 ms could mean a loss of up to 50% of the 23 Na signal to T2∗ relaxation. Therefore a method that allows a further reduction of the TE will lead to large gains in SNR. Recently, Ultra-short TE (UTE) chemical shift imaging has been used to acquire 23 Na and 31 P maps with short effective TE (27). In this method, the TE is minimized for the center of k-space by varying the length of the phase encode gradient pulse. By using CSI, a time domain signal is obtained for each spatial point. This yields more information as now the T2∗ can be determined for each pixel, but the imaging time for a given resolution is significantly increased. Robson et al. managed to acquire a cardiac 23 Na-MR image of both the fast and the slow relaxing components with a resolution of 5 × 5 × 15 mm or 0.4 ml voxel size. The total scan time was not given, but without cardiac gating losses, the four times averaging of total k-space would take close to an hour. This scan time can be reduced by 25–50% with weighted k-space scanning, but the method is definitely always slower than methods that use the time domain to code spatial information. The tactic of varying the phase-encoding pulse times, rather than the phase encode amplitudes, could also be used for 3D-GRE, but, in GRE, the minimal attainable TE will still be limited by the readout gradient-coding pulse and, thus, RBW, rather than by the phase encode pulses. Two distinct techniques for imaging solids using interleaved RF transmit and acquisition schemes may be used to obtain 23 NaMR images with negligible losses of the signal from the fast relaxing component. The first, single-point ramped imaging with T1 enhancement (SPRITE), was initially developed to image polymers (28), but has successfully been applied to 23 Na-MRI (29). The essence of the method is to use many very small flip angle RF pulses and only acquire one single data point immediately after each RF pulse (Fig. 8.2a). In practice, a few points can be acquired to compensate for time losses related to slow switching times of the MR scanner. Technically, this method is very challenging, because many software and hardware adaptations have to be made to a standard clinical scanner to achieve the required

184

Ouwerkerk

Fig. 8.2. Three methods for short TE imaging of 23 Na. a Single-point ramped imaging with T1 enhancement (SPRITE) (28). The gray shaded area is magnified to show the RF pulse (B1) with flip angle α and the short delay time (td ) between the pulse and the single point acquisition (acq). The Z-gradient GZ is ramped stepwise during the pulse train, whereas X and Y gradients are steppes between repetitions. b Sweep imaging with Fourier transformation (SWIFT) (30). The RF amplitude (B1) and offset frequency ( ωRF) are both modulated as a HS8 pulse (75) and Gz is kept constant. The gray shaded area shows a magnification of a part of the RF B1, offset, and acquisition. c 3D cone projection imaging (38, 39). Non-localized RF excitation with a tanh/tan-modulated adiabatic half-passage pulse (76) is followed with the shortest possible delay by spiral gradients. Two single projections are shown, and the k-space diagram shows the k-space trajectory of these gradients on two cones. Typically the lower cones need in the order of 48–32 projections whereas the inner cones less. A total of 1,000–2,000 projections with a 30–100 ms TR is common.

fast switching between RF transmit and receive. Ultimately, as can be seen in the literature results listed in Table 8.1, the SNR and resolution obtained with SPRITE at 4 T are not competitive with other 23 Na-MRI techniques. In its defense, it must be noted that it can deliver a quantitative sodium image that includes the fast decaying signal. Also, when a small number of points rather than

Sodium MRI

185

a single point are collected after each RF pulse, SPRITE, like the UTE-CSI method, can even yield some information on the fast T2∗ . Reconstruction methods for SPRITE imaging depends on the k-space trajectory, but the simplest implementation creates linear projections that can be recovered with FFT and then used in a 3D radon transform to create the image. The other MRI technique that has an effective TE short enough to image solids is SWIFT (sweep imaging with Fourier transformation) (30). The crucial difference with SPRITE is that in SWIFT the RF frequency is swept while a constant gradient encodes the readout (Fig. 8.2b). The swept RF pulse selects a slab, and at the same time one projection is encoded in the data that are recorded interleaved with the RF pulse. Thus, similar to SPRITE there is an interleaved pattern of small RF pulses and single point acquisitions, but unlike SPRITE, each pulse sequentially excites only a part of the sample due to the frequency sweep and the properties of the RF pulse. Also, the slab-selective properties of the SWIFT RF pulse train allows limitation of the field of view to a region smaller than the coil’s field of view. The method requires a relatively simple special reconstruction consisting of a deconvolution of the signal and the RF pulse transfer function to yield single line projections. The projections are designed to span enough of 3D k-space, and a regular (radon) projection reconstruction method can be used to reconstruct the 3D image. Unfortunately, like SPRITE, the data acquisition part of this method is hard to implement due to the required rapid switching between RF transmit and receive. Presently, no papers have been published with 23 Na-MR images acquired with SWIFT, but expectations are that this method could be more efficient than SPRITE because of the sequential excitation properties. One method to overcome the short T2 problem without the hardware and software complication of rapid interleaved RF transmission detection is to use 3D projection imaging (PI) methods (19, 31, 32). In 3D-PI sequences, the image is formed by line scans recorded with a constant readout gradient immediately after a non-selective RF excitation pulse (19, 31–33). There is no need for refocusing of a slice-selective gradient or a phase-encoding gradient pulse. Thus, the signal acquisition can begin very shortly after the excitation pulse with an effective TE that is only limited by the pulse duration and the switching time between transmit and receive. To reduce the large number of excitations needed for a complete 3D image with PI using constant projection gradients during readout, the projections can be twisted by using time-varying gradients (34, 35). The initial part of the gradient wave form still codes a linear projection in k-space, but the higher k-space frequencies that provide the detail in the image are covered more efficiently by a spiraling path (Fig. 8.2c). The so-called twisted

186

Ouwerkerk

spirals are prescribed on cones in k-space, and the cone angle is varied stepwise to cover a 3D half-sphere in k-space. This method has been successfully applied to quantification of TSC in humans (17, 36, 37). In practice, the gradient waveforms as calculated for the original TPI method may require very large gradient slew rates. Therefore, either the gradients can be low-pass filtered or some other gradient design techniques may be used (38, 39), provided that a sizeable initial part of k-space is covered more or less radially in a time of the order of the fast T2 , and the total readout time is kept short relative to the slow T2 . Whereas most of the spiral techniques use the spirals-oncones approach other strategies for efficient 3D imaging can be explored. An example is a UTE/TPI hybrid, where a stack of spirals is phase encoded with the phase-encoding pulse varied in length as in UTE-CSI (40). This approach allows processing as a 2D spiral technique, and the raw data are, thus, much easier to process. Undoubtedly more methods will become available and some may be generated by the recent interest in ultra-short TE 1 H-MRI imaging for imaging bone (41) ligaments (42), or paramagnetic contrast agents (43). Many of the RF pulse or gradient sequence advances developed for short T2 1 H imaging could be of use for 23 Na-MRI.

8. Data Processing The UTE-CSI and GRE techniques yield data sets that can be processed on the scanner by simple fast Fourier transform (FFT). Processing the data acquired using the non-Cartesian sampling techniques, such as TPI, takes a lot more effort. Although most scanners will have processing software for 2D spiral scan techniques, the available spiral scan methods for 1 H do not have best gradient waveforms for 23 Na, and processing of 3D data sets is usually not supported at all (only for stacks of 2D spirals, where the third dimension is phase encoded). Thus, some off-line processing is required. Whereas all 2D gridding techniques probably can be expanded to 3D, many of the algorithms turn out to be computationally intensive and demand a lot of computer memory. The 3D data set can be processed with non-uniform FFT or interpolated, or gridded (44), to a Cartesian grid. For this, the exact k-space trajectory is known, and extra lag times or gradient hardware limitations may cause deviations from the k-space trajectories predicted from numerical integration of the gradient waveforms. For most non-Cartesian reconstruction techniques, a sampling density function must also be calculated from the k-space trajectories (44).

Sodium MRI

9. Separating the Sodium Signal Components

187

The success in biomedical research of 23 Na-MRI in cells, perfused animal organs, or even whole animals depended heavily on the ability to separate the intracellular and extracellular components. Most of the events that are of biochemical or clinical interest involve a net transfer of sodium from the extracellular to the intracellular compartment. Neither the increase in intracellular sodium concentration nor the increase in extracellular space is entirely unique to cancer. In perfused ex vivo organs and in some animal models, intracellular and extracellular sodium can be separated with shift reagents that remain exclusively extracellular. The shift reagents typically consist of a rare earth metal ion, usually dysprosium (Dy) chelated with an organic (e.g., triethylenetraminehexaacetic acid, TTHA) or inorganic compound (triployphosphate, PPPi). An example of a shift reagent used on a suspension of red blood cells is shown in Fig. 8.3 (45). In perfused organ systems, the tissue survives and only competitive binding of Ca2+ by the chelating agent is a problem, particularly in hearts. For whole animals, however, to date, all suitable shift reagents are toxic and thus of no use for human studies. The triple quantum filtered 23 Na-MRI technique (46–48) has been suggested as an alternative way to separate intracellular and extracellular sodium. TQF certainly provides some suppression of extracellular signals, thus accentuating the tumor-specific changes in the intracellular sodium. This selectivity should not be overstated as in perfused animal hearts (49) and in red blood cell suspensions (46), and the extracellular sodium was found to significantly contribute to multiple quantum filtered signals. Also, the price in signal-to-noise ratio (SNR) of multiple quantum filters (MQF) is very high, leading to much longer scan times and reduced resolution. When a TQF technique is used for imaging, the preparation time, τ , of the TQF pre-sequence has to be optimized for a given set of fast and slow T2 (50, 51). Given this optimum τ , the amount of TQF signal observable with 23 NaMR spectroscopy can be estimated (Fig. 8.4a, b) (50), and this gives some idea of how much signal we can get in an image voxel (52). The optimum τ , and the received signal intensity for a given τ , can change rather steeply with changes in fast T2 and to a lesser extent also to changes in the slow T2 (Figs. 8.4a–d). Thus, signal changes in a diseased state could be the result of changes in T2 rather than an actual change in TSC. The optimization requires the recording of a time domain signal to measure the fast and slow T2 (53). When this is done, non-localized TQF parameters will be optimal for some average of the various relaxation rates found in the brain. A TQF version of the UTE-CSI on a suitable

188

Ouwerkerk

r

A

e i

B

i

10

0

e

-10

freq [ppm]

Fig. 8.3. Spectra of >80% hematocrit suspension of red blood cells with the extracellular component (e) shifted upfield by Dy(PPPi)2 and the sodium in the reference sample (r) shifted downfield by Dy(TTHA). a Control spectrum and b after 1 h exposure to 1 mM ouabain. The intracellular signal (i) increased and the potassium leaving the cells reduced the shift of the extracellular component by competitive binding to the Dy(PPPi)2 shift reagent. Figure modified from (77).

animal model may help to understand the changes observed with TQF better. Even so, in non-human primates some really dramatic changes in TQF sodium signals have been demonstrated occlusion and subsequent reperfusion. In the same study, a SQ 23 NaMRI only 12 min prior to the TQF images shows only small changes (relative to the >300% increase in TQF signals) (54). Clearly, in spite of ambiguities caused by the T2 dependency of the TQ filter the TQF appears to be far more sensitive to the changes in the brain after ischemia and reperfusion. Unfortunately, the visible changes observed with SQ 23 Na were not quantified in this study.

Sodium MRI

189

Fig. 8.4. Maximum signal intensities obtainable with single, double, or triple quantum-filtered (SQF, DQF, TQF, respectively) 23 Na-MR detected at an acquisition delay of 0.4 ms calculated according to Eqs. [2], [3], and [4] in (50). a The maximum signal intensities observable with 23 Na-MR at TE = 0.4 ms, as a function of the fast T2 with triple quantum filtering (solid line, slow T2 = 30 ms, and long dashed line, slow T2 = 20 ms) and without multiple quantum filter (short dashed line, slow T2 = 20 ms). b The same data as in (a) now expressed as ratio of the signal over the single SQF signal. The dot dash line is the signal ratio for a double quantum filter. All data in (a) and (b) assume an optimized multiple quantum filter. c The signal as a function of the preparation time τ as for three different fast T2 with the slow T2 fixed at 20 ms and a TE of 0.1 ms. d As in (c), but the fast T2 fixed at 1 ms and the slow T2 set at three values 10, 20, and 30 ms.

In light of all the shortcomings of MQF 23 Na-MR, the search for a method to distinguish intracellular and extracellular signals must go on. A relatively low-tech solution is inversion recovery (IR) T1 weighting (12). This approach has been used to study chemotherapy response in a mouse–human prostate cancer xenograft (55) to selectively observe cartilage tissue (56) and in the brain to suppress the CSF signal (57). The brain application looks promising even though the IR method reduces the SNR of normal brain tissues and the selective suppression of CSF appears to work as long as B0 field homogeneity is good. Also, the IR pulse works on differences in both T1 and T2 , because with the species with shorter T2 , the inversion pulses are less effective and the resulting incomplete inversion leads to a quicker signal

190

Ouwerkerk

recovery. The IR method precludes quantitative 23 Na-MRI, and it is unclear how T1 might change in disease. Perhaps where the MR properties of 23 Na fail to give us a handle on intracellular and extracellular volume, we should instead be looking at the much higher natural abundance and SNR available in 1 H spins to quantify intracellular and extracellular compartment volumes. At present, without a reliable means to exclusively quantify intracellular sodium, we can still use 23 NaMR imaging of the total tissue sodium to our advantage, but only in combination with an optimized, comprehensive 1 H-MRI protocol. To appreciate what type of information 23 Na-MRI adds to an MRI examination, we can be pragmatic and study TSC in disease and see whether 23 Na is a good biomarker. At the same time, we need to continue studies on perfused organs and whole-animal models in order to understand the origin of the disease-specific contrast delivered by measurement of TSC with 23 Na-MRI.

10. Brain Tissue Sodium and Cancer

Proliferating cells have an abnormally high sodium content (58), because the intracellular sodium concentration is elevated as a result of altered Na+ /H+ transport kinetics (59–62) and pH. Outside the cells, continuous perfusion of living tissue will ensure a constant sodium concentration of approximately 140 mmol/l. Thus, an increase in the extracellular partial tissue volume through the increased vascularization (angiogenesis) and the increased interstitial space in tumors (63, 64) will also lead to increases in tissue sodium concentration (TSC) in tumors. Again, as in active or energetically challenged excitable tissues, both intracellular sodium concentration and extracellular volume are contributing to an increase in tissue sodium. Using a quantitative 23 Na-MRI technique, a 50% increase in TSC, relative to non-involved contralateral tissues, was found in malignant brain tumors (36). An example study of a patient with an astrocytoma is shown in Fig. 8.5. Likewise, in malignant breast tumors, an increase of 50% in TSC was found relative to noninvolved glandular tissue and benign lesions. A small study of 13 subjects with suspected low-grade glioma 23 Na-MRI, combined with 1 H-MRS, revealed a doubling of the sodium signal compared with contralateral white matter (65). A negative correlation was found between sodium signal and N-acetylaspartate (NAA). In this study, the ratio of NAA over the 23 Na signal was shown to be the best metric for separating glioma from healthy tissue. The 23 Na-MRI data in this study were recorded with a gradient-recalled echo (GRE) sequence with a

Sodium MRI

191

Fig. 8.5. 1 H- and 23 Na images of the brain of a patient (36-year-old male) with an astocytoma grade III. a T1 -weighted post-contrast (Gd–). b T2 -weighted fluid-attenuated inversion recovery (FLAIR). c Quantitative 23 Na-twisted projection image with 100 ms TR and 0.4 ms TE. Two reference sample tubes with 2% agarose gel containing known [Na+ ] were placed on either side of the head for quantification of brain TSC and to facilitate registration of 1 H and 23 Na images. The 23 Na images were recorded at 1.5 T with a separate 23 Na birdcage coil that replaced the 1 H birdcage coil after the 1 H-MRI protocol without moving the patient’s head or reference phantoms. The 23 Na image (c) is an oblique slice from the isotropic 3D 23 Na data set, interpolated to register with the 1 H-MR images. Level contours drawn on the post-Gd T1 -weighted image (a) were copied to the 23 Na image (c) to show the location of the Gd enhancement.

relatively long TE of 3.8 ms. The results were corrected for T2∗ relaxation assuming an 18 ms (single exponential) T2∗ relaxation time in normal tissue versus 27 ms in the low-grade glioma. The latter value was measured in only one subject, but it is a substantial difference with the normal value. If this increase in T2∗ holds true for all brain tumors, it makes the detection of cancer-related increases in TSC with single quantum 23 Na-MR less dependent on the choice of MR sequence timing.

11. Sodium Changes in Stroke 23 Na-MRI

has been used to study stroke in humans as far back as 1993 (16). With the limited SNR no significant increase in sodium was seen in the first 13 h post-infarction. After that initial increase, 23 Na was seen to peak at 45–82 h. In animal models, the sequence of events could be studied more accurately. In a rat model of stroke induced by tandem occlusion of the right middle cerebral artery (MCAO) and common carotid arteries (CCAs), sodium as determined from punch samples with flame spectrometry appeared to increase linearly with time up to at least about 450 min after occlusion at a rate of about 1 mmol/kg dry weight/min against a staring concentration of 240–250 mmol/kg dry weight (66). In animal models, these findings have been confirmed with 23 Na-MRI (67, 68). Whether measuring TSC or using special techniques to be more sensitive to [Na+ ]in , the complex sequence of changes in the brain after transient ischemia cannot be fully understood by

192

Ouwerkerk

looking at the changes in sodium alone. It is important to know what happens to cell volume and the relative volume of the extracellular space in order to know what a change in observed sodium signal means. The observed change can be due to actual influx of sodium into the cell (a sign of compromised membrane integrity or a shortfall in energy metabolism) or shrinkage of the cell. In TQF 23 Na, this would lead to a more restricted environment, which would change relaxation parameters and probably increase the portion of intracellular sodium experiencing anisotropic environments that lead to a TQF signal. Although a study that combined 1 H-MRI and 23 Na-MRI (69) failed to show that any of these techniques can unambiguously predict the potential infarct size, non-human primate studies and human studies using combined 1 H- and 23 Na-MRI showed encouraging results (17).

12. 23 Na-MRI to Monitor Therapy Lack of substrate and oxygen can cause quite significant changes in tissue sodium content. As soon as the energy-dependent Na+ /K+ -ATPase stops pumping sodium out of the cell, passive sodium influx from the extracellular environment will rapidly raise the intracellular levels of sodium several fold. This effect is exacerbated when stress on the cells increases the permeability of the cell wall for sodium ions. The acute effect of a therapy that causes cell death or membrane rupture on an appreciable scale should, therefore, be easy to monitor with 23 Na-MRI. This was confirmed by the observation of an increase in sodium 24 h after administration of taxotere in a xenograft animal model of prostate cancer (55). In the long run, the elevated sodium found in cancer should again decline if the therapy is successful. In a limited number of patients with breast cancer, who were undergoing preoperative systemic chemotherapy, the effect of the therapy in responders was demonstrated as a decline in the tumor TSC along with a decline in lesion size (70, 71). In this application of 23 Na-MRI, it is important to have a reference or a quantification method by which to compare the results of baseline images and multiple post-chemo therapy images. Of course therapies other than chemotherapy can also be monitored. Surgery in the brain will leave a void that is filled with CSF. The high signal of CSF could obscure high TSC in the adjacent residual tissue. Thus, for postoperative evaluations the method of Stobbe et al. to suppress the CSF signal (57) might be interesting. Ablation methods should lead to an easily detectable increase in TSC. The effect of the highly focused ultrasound (HIFUS) ablation of human uterine fibroids could be

Sodium MRI

193

seen on 23 Na-MRI images obtained within 24 h of therapy (72). The long-term effects of such therapies may be more difficult to predict, because this depends on what type of tissue replaces the damaged tissue in the ablated areas. Whereas most of the aforementioned therapies typically are designed to induce quick cell death in a lesion, one study actually implies that 23 Na can also be used to detect cell death in a slow progressive disease, such as mild Alzheimer’s disease. A small, but significant, increase in hippocampal sodium signal as measured in five patients with a 23 Na GRE sequence at 3 T was seen to inversely correlate with hippocampal volume (73). It is encouraging to see that such subtle changes can also be detected, even with a relatively easily implemented sequence, such as GRE.

13. How to Perform a 23 Na-MRI Study of the Brain

All the preceding information on 23 Na methods is vital knowledge for gathering and publishing biomedical 3 Na research data. Even so, given the right hardware and research tools on the MR scanner, the task of measuring a 23 Na-MRI data set does not have to be daunting. A relatively simple method, such as 3D-GRE, straight projection imaging or UTE-CSI, can be implemented relatively easily. In the following recipe, the 3D-GRE method was selected for simplicity and universal availability. Stepwise, here is what to do: 1. Choose the best scanner for the protocol. The highest possible field is good for 23 Na, but the availability of a comprehensive 1 H-MRI protocol for the disease to be studied is just as important. The scanner must be equipped with the hardware required for measuring other nuclei. If not present, a 23 Na pre-amplifier must be acquired. These are relatively inexpensive, and it is advisable to have a spare. 2. Obtain a suitable MR sequence or get access to the scanner software and create one of the sequences above. Test the sequence in a simulator to see whether literature values for RBW, resolution, and minimum TE can be reproduced. For a GRE sequence, the selective RF pulse can be replaced by a hard (rectangular) RF pulse or by an adiabatic half-passage (AHP) pulse (Fig. 8.6). The latter ensures that even if the coil homogeneity is not perfect, the whole brain is excited with a 90◦ flip angle. Because of the short T1 of 23 Na, a 90◦ flip angle can yield optimum SNR per unit time, with a relatively short TR. Assuming a brain 23 Na T1 is about 30 ms the optimum TR in terms of optimum SNR

194

Ouwerkerk

Fig. 8.6. A 3D-GRE sequence for 23 Na-MRI. a The standard sequence as present on most scanners for 1 H-MRI with phase encoding in the Y and Z directions. The slice is relatively thick and subdivided by phase encoding in the Z direction superimposed on the slice-selective refocusing gradient lobe. b The modified sequence for 23 Na-MRI. The RF pulse has been replaced by a non-selective AHP. The dephasing (X) and phase-encoding (Y, Z) gradient lobes are made as short as possible, a partial echo readout is used, and the RF pulse is a non-selective tanh/tan-modulated AHP pulse. A relatively large field of view and low resolution keep all minimum gradient times short, but a (desirable) low RBW will increase the TE.

per unit time for the 90◦ flip angle is just under 40 ms, or 1.25 × T1 . 3. Figure 8.6 shows a comparison between a standard 3DGRE sequence and a version suitable for 23 Na-MRI. In the optimize sequence of Fig. 8.6b, the slice-selective pulse is replaced by a tan/tanh modulated AHP. Not only does this allow the use of coil with a less homogeneous B1 field (which could allow a better filling factor and thus SNR), but it also allows the elimination of the slice-selective

Sodium MRI

195

gradient and the refocusing lobe. Because the symmetry point of the AHP is at the end of the pulse, the nominal TE is not increased by an increase of the pulse duration, but there will be some T2 relaxation during the pulse for the fast T2 component of the signal (37). The limiting factor on achieving minimal TE for this sequence is still the length of the dephasing gradient and the RBW. Without resorting to the aforementioned UTE-CSI or projection imaging methods, this cannot be avoided. 4. Obtain a 23 Na head coil that can be swapped for a clinicalgrade 1 H-MR head coil without moving the subjects’s head (36) or get a dual-tuned 23 Na-1 H head coil that has a clinical-grade performance on the 1 H channels(s) without sacrificing too much of the 23 Na performance (74). For brain, the 23 Na coil must be transmit and receive in quadrature. This requires a splitter for 23 Na to separate transmit RF from the receiver input and at the other end split the RF signals into two orthogonal phase quadrature signals. This component should typically be included in the coil, and the pre-amplifier mentioned in point 1 could be integrated as well. The 1 H can be receive only and possibly multi-channel. 5. Create phantoms to test the coil. One phantom should be large and loaded with 100–150 mmol/l NaCl and doped with ca. 2 g/l CuSO4 . It is advisable to create gel phantoms containing known concentrations of sodium ranging from 20 to 150 mmol/l for calibration and for use as feducials and reference standards. The use of 2–4% (weight/volume) agarose and about 2 g/l CuSO4 will reduce the T2 and T1 of the samples. 6. Measure 23 Na-MRI images with different flip angles spanning at least a factor two variation. The actual local flip angles can be determined from a sine fit of the image pixel intensities as a function of nominal flip angle. If AHP pulses are used, the B1 must be varied to determine the threshold B1 for adiabaticity. Determine the minimum B1 at which the signal is B1 invariant for all locations. For actual experiments on the brain, use a B1 that is at least 50% higher than the threshold for as much as patient safety limitations of the specific absorption rate (SAR) will allow. Note that the patient’s weight entered in phantom studies is important in the SAR calculations. 7. Measure the SNR with different RBW and the minimum TE at each RBW. Determine the optimum settings. 8. Measure the 23 Na T1 and T2 (or T2∗ ) of the concentration and reference phantoms. If the 23 Na-MR sequence does

196

Ouwerkerk

not allow this, it can be done non-localized with a spectroscopy sequence one sample at a time. This approach is tedious, but precise. The agarose gel and CuSO4 should bring the relaxation rates of the sodium in the phantom close to physiological values, but this must be verified and recorded for corrections of reference phantom data collected with in vivo studies. Create a concentration calibration plot (Fig. 8.7). Repeat this with the concentration test phantoms in different positions.

Fig. 8.7. Example of a single concentration calibration series. Seven 50-ml tubes were filled with 2% w/vol aqueous agarose gel with 2 g/l CuSO4 and 30, 45, 60, or 75 mmol/l NaCl. A 1,240 projection TPI sequence with TR/TE = 100/0.4 ms was used to record an image with 6 mm isotropic actual resolution in 10 min using a 3-turn solenoid breast coil. Images were reconstructed by gridding to 64 × 64 × 64 points and FFT to a 22-cm isotropic field of view image. The SNR was determined from mean pixel intensities in circular regions in the tubes and regions devoid of meaningfull signal of 1–2 thousand pixels. SNR was calculated as the ratio of the signal minus the mean noise divided by the noise standard deviation. The SNR correlated with concentration with r2 > 0.98.

9. Test the sequence on a healthy volunteer and measure SNR in a relatively CSF-free part of the brain and if possible determine gray and white matter SNR from a segmented brain image created with a registered 1 H-MRI. Place two or more agarose gel samples with known sodium concentrations against the skull (Fig. 8.1). These can serve as concentration reference and as feducials. The agarose and CuSO4 drastically improves visibility on T2 -weighted 1 H images. From these tests, determine the minimum scan time that yields the desired resolution and estimated contrast to noise. 10. Determine the estimated TSC. If possible, measure T2∗ and T1 in the brain. Even when not aiming for quantitation of TSC, the sequence should be set up to minimize signal losses as a result of relaxation, because we cannot know how T2∗ and T1 might change in disease. Ideally, signal changes reflect changes in TSC.

Sodium MRI

197

11. Test the entire 1 H-23 Na protocol on healthy volunteers. 12. Scan patients, analyze the data, and publish! Of course, the more complex techniques, such as TPI or SWIFT, could possibly yield better SNR or more precise quantification of TSC. Whether this is worth the extra effort depends on your long-term research goals, technical prowess, and endurance. The above recipe is designed to provide a quick and solid introduction into 23 Na-MRI.

14. Quantification of TSC Apart from the reduced signal losses, the very short TE methods have the added advantage that quantification can be performed with minimal or no correction for T2 losses. By extending the TR to three times the T1 (or less of smaller flip angles are used) the correction for signal saturation can also be bypassed. This slower TR reduces the efficiency in terms of SNR per unit time, but it removes the ambiguity associated with the use of relaxation corrections when we cannot be sure that these remain the same in disease. Unless the SNR for brain tissues is really high (>100) sodium concentrations are best calculated from mean intensities in regions of interest (ROI). A convenient way to do is to trace ROI on co-registered with 1 H images for guidance and copy the regions to the 3D 23 Na images. The two or three sodium concentration reference tubes placed close to the head provide a feducial for co-registering images. For 23 Na imaging of the brain with a quadrature head coil we can assume homogeneous receive sensitivity and use the external reference samples or use a B1 map, as measured in point 5 of the list above, to correct for differences in receive sensitivity and flip angles (unless AHP pulse were used). Alternatively, the signal of CSF or even ocular fluids can be used as an internal standard. These fluids have a long T2 > 40 ms and also an even longer T1 . The latter needs to be determined for necessary T1 corrections; the T2 corrections are less of a problem.

15. Conclusion To get the most out of 23 Na-MRI, it is advisable to complement the TSC-related information gained from 23 Na-MR with 1 H-MR techniques that provide additional information of the

198

Ouwerkerk

physiological environment of sodium in the tissue. Although the water distribution between intracellular and extracellular is different and no technique gives direct information about the relative compartment volumes some well-established 1 H-MR techniques, such as contrast-enhanced (CE)-MRI or diffusionweighted (DW)-MRI, are sensitive to changes in cell volume, extracellular space, and membrane permeability. The best protocol will be different for each particular application, and in each application, we have to examine whether 23 NaMRI has enough added value to justify the extra scan time and effort. But in doing so, we must not forget that practically all MRI exams consist of a series of scans and only the whole set will yield the desired diagnostic sensitivity and specificity. We cannot expect that 23 Na-MRI can compete with clinical 1 H-MRI protocols when implemented in an exam consisting of only a scout image and a 23 Na-MRI, but a well-designed protocol including 23 Na-MRI can yield exiting results. References 1. Murphy, E., Eisner, D. A. Regulation of intracellular and mitochondrial sodium in health and disease. Circ Res 2009;104: 292–303. 2. Hilgemann, D. W., Yaradanakul, A., Wang, Y., Fuster, D. Molecular control of cardiac sodium homeostasis in health and disease. J Cardiovasc Electrophysiol 2006;17(Suppl 1):S47–S56. 3. Ouwerkerk, R., Lee, R. F., Bottomley, P. A. Dynamic changes in sodium levels in human exercising muscle measured with 23 Na-MRI. Proc Lntl Soc Mag Reson Med 1999;7: p 1530. 4. Clausen, T. The sodium pump keeps us going. Ann N Y Acad Sci 2003;986: 595–602. 5. Jakovcevic, D., Harder, D. R. Role of astrocytes in matching blood flow to neuronal activity. Curr Top Dev iol 2007;79:75–97. 6. Magistretti, P. J., Pellerin, L. Cellular mechanisms of brain energy metabolism and their relevance to functional brain imaging. Philos Trans R Soc Lond iol Sci 1999;354: 1155–1163. 7. Magistretti, P. J. Neuron-glia metabolic coupling and plasticity. J Exp iol 2006;209: 2304–2311. 8. Silver, I. A., Erecinska, M. Energetic demands of the Na+ /K+ ATPase in mammalian astrocytes. Glia 1997;21:35–45. 9. Chatton, J. Y., Marquet, P., Magistretti, P. J. A quantitative analysis of L-glutamateregulated Na+ dynamics in mouse cor-

10.

11.

12.

13.

14.

15.

tical astrocytes: Implications for cellular bioenergetics. Eur J Neurosci 2000;12: 3843–3853. Constantinides, C. D., Gillen, J. S., Boada, F. E., Pomper, M. G., Bottomley, P. A. Human skeletal muscle: Sodium MR imaging and quantification-potential applications in exercise and disease. Radiology 2000;216: 559–568. Weber, M. A., Nielles-Vallespin, S., Huttner, H. B. et al. Evaluation of patients with paramyotonia at 23 Na MR imaging during cold-induced weakness. Radiology 2006;240:489–500. Hotta, Y., Ando, H., Takeya, K., Sakakibara, J. Direct measurement of increased myocardial cellular 23 Na NMR signals in perfused guinea-pig heart induced by dihydroouabain and grayanotoxin-I. Mol Cell iochem 1994;139:59–70. Constantinides, C. D., Kraitchman, D. L., O‘Brien, K. O., Boada, F. E., Gillen, J., Bottomley, P. A. Noninvasive quantification of total sodium concentrations in acute reperfused myocardial infarction using Na23 MRI. Magn Reson Med Sci 2001;46: 1144–1151. Igarashi, H. Studies of metabolic changes during and following cerebral ischemia in gerbils by in vivo multi nuclear magnetic resonance spectroscopy. Nippon Ika Daigaku Zasshi 1989;56:339–348. Perman, W. H., Heiberg, E. V., Hayes, C. E., Dunphy, T. 3D sodium-23 imaging of

Sodium MRI

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

breast lesions. Proc Int Soc Magn Reson Med Sci 1995;3:p 1196. Shimizu, T., Naritomi, H., Sawada, T. 23 Na-MRI Sequential changes on after cerebral infarction. Neuroradiology 1993;35:416–419. Thulborn, K. R., Davis, D., Snyder, J., Yonas, H., Kassam, A. Sodium MR imaging of acute and subacute stroke for assessment of tissue viability. Neuroimaging Clin N Am 2005;15:639–653, xi–xii. Stobbe, R., Beaulieu, C. Advantage of sampling density weighted apodization over postacquisition filtering apodization for sodium MRI of the human brain. Magn Reson Med 2008;60:981–986. Nielles-Vallespin, S., Weber, M. A., Bock, M. et al. 3D radial projection technique with ultrashort echo times for sodium MRI: Clinical applications in human brain and skeletal muscle. Magn Reson Med 2007;57:74–81. Clayton, D. B., Lenkinski, R. E. MR imaging of sodium in the human brain with a fast three-dimensional gradient-recalled-echo sequence at 4 T. Acad Radiol 2003;10: 358–365. Atkinson, I. C., Renteria, L., Burd, H., Pliskin, N. H., Thulborn, K. R. Safety of human MRI at static fields above the FDA ST guideline: sodium imaging at 9.4T does not affect vital signs or cognitive ability. J Magn Reson Imaging 2007;26:1222–1227. Sandstede, J. J. W., Hillenbrand, H., Beer, M. et al. Time course of Na-23 signal intensity after myocardial infarction in humans. Magn Reson Med 2004;52:545–551. Jacobs, M. A., Ouwerkerk, R., Wolff, A. C. et al. Multiparametric and multinuclear magnetic resonance imaging of human breast cancer: Current applications. Technol Cancer Res Treat 2004;3:543–550. Jacobs, M. A., Ouwerkerk, R., Kamel, I., Bottomley, P. A., Bluemke, D. A., Kim, H. S. Proton, diffusion-weighted imaging, and sodium (Na-23) MRI of uterine leiomyomata after MR-guided highintensity focused ultrasound: A preliminary study. J Magn Reson Imaging 2009;29: 649–656. Ouwerkerk, R., Bleich, K. B., Gillen, J. S., Pomper, M. G., Bottomley, P. A. Tissue sodium concentration in human brain tumors as measured with 23Na MR imaging. Radiology 2003; 227(2):529–537. Ouwerkerk, R., Jacobs, M. A., Macura, K. J. et al. Elevated tissue sodium concentration in malignant breast lesions detected with noninvasive Na-23 MRI. Breast Cancer Res Treat 2007;106:151–160.

199

27. Robson, M. D., Tyler, D. J., Neubauer, S. Ultrashort TE chemical shift imaging (UTECSI). Magn Reson Med 2005;53:267–274. 28. Kennedy, C. B., Balcom, B. J., Mastikhin, I. V. Three-dimensional magnetic resonance imaging of rigid polymeric materials using single-point ramped imaging with T-1 enhancement (SPRITE). Can J ChemRevue Canadienne De Chimie 1998;76: 1753–1765. 29. Romanzetti, S., Halse, M., Kaffanke, J., Zilles, K., Balcom, B. J., Shah, N. J. A comparison of three SPRITE techniques for the quantitative 3D imaging of the 23 Na spin density on a 4T whole-body machine. J Magn Reson 2006;179:64–72. 30. Idiyatullin, D., Corum, C., Park, J. Y., Garwood, M. Fast and quiet MRI using a swept radiofrequency. J Magn Reson 2006;181:342–349. 31. Ra, J. B., Hilal, S. K., Oh, C. H. An algorithm for MR imaging of the short T2 fraction of sodium using the FID signal. J Comput Assist Tomogr 1989;13: 302–309. 32. Jerecic, R., Bock, M., Wacker, C., Bauer, W., Schad, L. R. 23 Na-MRI of the human heart using a 3D radial projection technique. Biomed Tech (erl) 2002;47(Suppl 1 Pt 1):458–459. 33. Jerecic, R., Bock, M., Nielles-Vallespin, S., Wacker, C., Bauer, W., Schad, L. R. ECG-gated 23 Na-MRI of the human heart using a 3D-radial projection technique with ultra-short echo times. MAGMA 2004;16: 297–302. 34. Boada, F. E., Gillen, J. S., Shen, G. X., Chang, S. Y., Thulborn, K. R. Fast three dimensional sodium imaging. Magn Reson Med 1997;37:706–715. 35. Thulborn, K. R., Davis, D., Adams, H., Gindin, T., Zhou, J. Quantitative tissue sodium concentration mapping of the growth of focal cerebral tumors with sodium magnetic resonance imaging. Magn Reson Med 1999;41:351–359. 36. Ouwerkerk, R., Bleich, K. B., Gillen, J. S., Pomper, M. G., Bottomley, P. A. Tissue sodium concentration in human brain tumors as measured with 23 Na MR imaging. Radiology 2003;227:529–537. 37. Ouwerkerk, R., Weiss, R. G., Bottomley, P. A. Measuring human cardiac tissue sodium concentrations using surface coils, adiabatic excitation, and twisted projection imaging with minimal T2 losses. J Magn Reson Imaging 2005;21:546–555. 38. Gurney, P. T., Hargreaves, B. A., Nishimura, D. G. Design and analysis of a practi-

200

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

Ouwerkerk cal 3D cones trajectory. Magn Reson Med 2006;55:575–582. Ouwerkerk, R.. Using Heading and Speed to Design Gradient Waveforms with Lower Slew Rates. In: ISMRM workshop on nonCartesian MRI; 2007 25–28 Feb 2007; Sedona, AZ; 2007. Qian, Y. X., Boada, F. E. Acquisitionweighted stack of spirals for fast highresolution three-dimensional ultra-short echo time MR imaging. Magn Reson Med 2008;60:135–145. Du, J., Hamilton, G., Takahashi, A., Bydder, M., Chung, C. B. Ultrashort echo time spectroscopic imaging (UTESI) of cortical bone. Magn Reson Med 2007;58:1001–1009. Rahmer, J., Bornert, P., Dries, S. P. Assessment of anterior cruciate ligament reconstruction using 3D ultrashort echotime MR imaging. J Magn Reson Imaging 2009;29:443–448. Liu, W., Dahnke, H., Rahmer, J., Jordan, E. K., Frank, J. A. Ultrashort T2 ∗ relaxometry for quantitation of highly concentrated superparamagnetic iron oxide (SPIO) nanoparticle labeled cells. Magn Reson Med 2009;61:761–766. Jackson, J. I., Meyer, C. H., Nishimura, D. G., Macovski, A. Selection of a convolution function for Fourier inversion using gridding. IEEE Trans Med Imaging 1991;10: 473–478. Ouwerkerk, R., Vanechteld, C. J. A., Staal, G. E. J., Rijksen, G. Erythrocyte Na+ /K+ ATPase activity measured with Na-23 NMR. Magn Reson Medicine 1989;12:164–171. Knubovets, T., Shinar, H., Navon, G. Quantification of the contribution of extracellular sodium to 23 Na multiple-quantum-filtered NMR spectra of suspensions of human red blood cells. J Magn Reson 1998;131: 92–96. Winter, P. M., Poptani, H., Bansal, N. Effects of chemotherapy by 1,3-bis(2-chloroethyl)1-nitrosourea on single-quantum- and triple-quantum-filtered 23 Na and 31P nuclear magnetic resonance of the subcutaneously implanted 9L glioma. Cancer Res 2001;61:2002–2007. Dizon, J. M., Tauskela, J. S., Wise, D., Burkhoff, D., Cannon, P. J., Katz, J. Evaluation of triple-quantum-filtered 23 Na NMR in monitoring of intracellular Na content in the perfused rat heart: Comparison of intra- and extracellular transverse relaxation and spectral amplitudes. Magn Reson Med 1996;35: 336–345. Jelicks, L. A., Gupta, R. K. On the extracellular contribution to multiple quantum fil-

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60. 61.

tered 23 Na NMR of perfused rat heart. Magn Reson Med 1993;29:130–133. Navon, G. Complete elimination of the extracellular 23 Na NMR signal in triple quantum filtered spectra of rat hearts in the presence of shift reagents. Magn Reson Med 1993;30:503–506. Boada, F. E., Tanase, C., Davis, D. et al. Non-invasive assessment of tumor proliferation using triple quantum filtered 23/Na-MRI: Technical challenges and solutions. Conf Proc IEEE Eng Med iol Soc 2004;7:5238–5241. Ooms, K. J., Cannella, M., Vega, A. J., Marcolongo, M., Polenova, T. Na23 TQF NMR imaging for the study of spinal disc tissue. J Magn Reson 2008;195: 112–115. Borthakur, A., Hancu, I., Boada, F. E., Shen, G. X., Shapiro, E. M., Reddy, R. In vivo triple quantum filtered twisted projection sodium MRI of human articular cartilage. J Magn Reson 1999;141:286–290. LaVerde, G., Nemoto, E., Jungreis, C. A., Tanase, C., Boada, F. E. Serial triple quantum sodium MRI during non-human primate focal brain ischemia. Magn Reson Med 2007;57:201–205. Kline, R. P., Wu, E. X., Petrylak, D. P. et al. Rapid in vivo monitoring of chemotherapeutic response using weighted sodium magnetic resonance imaging. Clin Cancer Res 2000;6:2146–2156. Rong, P., Regatte, R. R., Jerschow, A. Clean demarcation of cartilage tissue 23 Na by inversion recovery. J Magn Reson 2008;193: 207–209. Stobbe, R., Beaulieu, C. In vivo sodium magnetic resonance imaging of the human brain using soft inversion recovery fluid attenuation. Magn Reson Med 2005;54: 1305–1310. Cameron, I. L., Smith, N. K., Pool, T. B., Sparks, R. L. Intracellular concentration of sodium and other elements as related to mitogenesis and oncogenesis in vivo. Cancer Res 1980;40:1493–1500. Rotin, D., Steele-Norwood, D., Grinstein, S., Tannock, I. Requirement of the Na+ /H+ exchanger for tumor growth. Cancer Res 1989;49:205–211. Lagarde, A. E., Pouyssegur, J. M. The Na+ :H+ antiport in cancer. Cancer iochem iophys 1986;9:1–14. Nagy, I., Lustyik, G., Lukacs, G., Nagy, V., Balazs, G. Correlation of malignancy with the intracellular Na+ :K+ ratio in human thyroid tumors. Cancer Res 1983;43: 5395–5402.

Sodium MRI 62. Reshkin, S. J., Bellizzi, A., Caldeira, S. et al. Na+ /H+ exchanger-dependent intracellular alkalinization is an early event in malignant transformation and plays an essential role in the development of subsequent transformation-associated phenotypes. FASEB J 2000;14:2185–2197. 63. Patchett, A. A., Cordes, E. H. The design and properties of N-carboxyalkyldipeptide inhibitors of angiotensin-converting enzyme. Adv Enzymol Relat Areas Mol iol 1985;57: 1–84. 64. Hayes, C., Padhani, A. R., Leach, M. O. Assessing changes in tumour vascular function using dynamic contrast-enhanced magnetic resonance imaging. NMR iomed 2002;15:154–163. 65. Bartha, R., Megyesi, J. F., Watling, C. J. Low-grade glioma: correlation of short echo time 1H-MR spectroscopy with 23 Na MR imaging. AJNR Am J Neuroradiol 2008;29:464–470. 66. Wang, Y., Hu, W. X., Perez-Trepichio, A. D. et al. Brain tissue sodium is a ticking clock telling time after arterial occlusion in rat focal cerebral ischemia. Stroke 2000;31:1386–1391. 67. Jones, S. C., Kharlamov, A., Yanovski, B. et al. Stroke onset time using sodium MRI in rat focal cerebral ischemia. Stroke 2006;37:883–888. 68. Yushmanov, V. E., Yanovski, B., Kharlamov, A., LaVerde, G., Boada, F. E., Jones, S. C. Sodium mapping in focal cerebral ischemia in the rat by quantitative Na23 MRI. J Magn Reson Imaging 2009;29: 962–966. 69. Lin, S. P., Song, S. K., Miller, J. P., Ackerman, J. J. H., Neil, J. J. Direct, longitudinal comparison of H-1 and Na-23 MRI after transient focal cerebral ischemia. Stroke 2001;32:925–932. 70. Jacobs, M. A., Ouwekerk, R., Wolff, A. C. et al. Diffusion Weighted Imaging, ADC Mapping, and Sodium MR Imaging of Oper-

71.

72.

73.

74.

75.

76.

77.

201

able Breast Cancer After Neoadjuvant Therapy: Preliminary Results. In: Proceedings of the Seventeenth Meeting of the International Society of Magnetic Resonance in Medicine; 2009 May; Honolulu, HI, USA; 2009, p 2117. Jacobs, M. A., Ouwekerk, R., Wolff, A. C. et al. Multinuclear and Multiparametric MR imaging as an early treatment response biomarker for preoperative systemic therapy in breast cancer: Preliminary Results. In: Proceedings of the Seventeenth Meeting of the International Society of Magnetic Resonance in Medicine; 2009 May; Honolulu, HI, USA; 2009, p 2220. Jacobs, M. A., Ouwerkerk, R., Kamel, I., Bottomley, P. A., Bluemke, D. A., Kim, H. S. Proton, diffusion-weighted imaging, and sodium (23 Na) MRI of uterine leiomyomata after MR-guided high-intensity focused ultrasound: A preliminary study. J Magn Reson Imaging 2009;29:649–656. Mellon, E. A., Pilkinton, D. T., Clark, C. M. et al. Sodium MR imaging detection of mild Alzheimer disease: Preliminary study. Am J Neuroradiol 2009;30:978–984. Shen, G. X., Boada, F. E., Thulborn, K. R. Dual-frequency, dual-quadrature, birdcage RF coil design with identical B1 pattern for sodium and proton imaging of the human brain at 1.5 T. Magn Reson Med 1997;38:717–725. Tannus, A., Garwood, M. Improved performance of frequency-swept pulses using offset-independent adiabaticity. J Magn Reson Ser A 1996;120:133–137. Garwood, M., Yong, K. Symmetrical pulses to induce arbitrary flip angles with compensation for RF inhomogeneity and resonance offsets. J Magn Reson 1991;94: 511–525. Ouwerkerk, R., van Echteld, C. J., Staal, G. E., Rijksen, G. Erythrocyte Na+ /K+ ATPase activity measured with 23 Na NMR. Magn Reson Med 1989;12:164–171.

wwwwwww

Chapter 9 MR Spectroscopy and Spectroscopic Imaging of the Brain He Zhu and Peter B. Barker Abstract Magnetic resonance spectroscopy (MRS) and the related technique of magnetic resonance spectroscopic imaging (MRSI) are widely used in both clinical and preclinical research for the non-invasive evaluation of brain metabolism. They are also used in medical practice, although their ultimate clinical value continues to be a source of discussion. This chapter reviews the general information content of brain spectra and commonly used protocols for both MRS and MRSI and also touches on data analysis methods and quantitation. The main focus is on proton MRS for application in humans, but many of the methods are also applicable to other nuclei and studies of animal models as well. Key words: Brain, magnetic resonance spectroscopy, spectroscopic imaging, spatial localization, metabolites.

1. Introduction In vivo magnetic resonance spectroscopy (MRS) of the human brain has developed rapidly since its first observation in the 1980s (1, 2). Early studies in both humans and animals focused on the 31 P nucleus which allowed the measurement of energy metabolites such as phosphocreatine and ATP, as well as inorganic phosphate and phosphoesters (1). With the development of improved techniques for spatial localization and water suppression, proton MRS became more prevalent in the 1990s because of its higher sensitivity and greater convenience (since it can be performed without hardware modification on most MRI machines, unlike MRS of other nuclei) (3). While interest remains, particularly at high magnetic field strengths, in nuclei such as 31 P, 23 Na, and M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_9, © Springer Science+Business Media, LLC 2011

203

204

Zhu and Barker 13 C

(particularly for isotopically labeled and/or hyper-polarized molecules (4)), the vast majority of brain MRS studies in vivo use the proton. The remainder of this article therefore focuses on protocols for 1 H-MRS.

2. Information Content of Proton MR Spectra of the Brain

Because of its relatively low sensitivity, only small, mobile molecules which are present in millimolar quantities are generally detectable in an in vivo MR spectrum. At commonly used field strengths such as 1.5 or 3.0 T, only signals from choline (Cho), creatine (Cr), and N-acetylaspartate (NAA) are observed in normal brain at long echo times (e.g., 140 or 280 ms) (Fig. 9.1a), while compounds such as lactate, alanine, or others may be detectable in pathological conditions which increase their concentration (5–7). At short echo times (e.g., 35 ms or less) other compounds such as glutamate, glutamine, myo-inositol, as well as lipids and macromolecular resonances (Fig. 9.1b), are detectable. A summary of all compounds that have been detected in the human brain by proton MRS is given in Table 9.1, and a complete list of metabolite structures and their spectra can be found in (8). The biological significance of the major compounds is discussed below.

(a)

(b)

Fig. 9.1. 3 T PRESS brain spectra recorded from a 2-year-old boy with TE 135 ms (a) and TE 30 ms (b) at the level of the centrum semiovale with a nominal voxel size of 1.5 cm3 . In the long TE spectrum, signals are present from choline (Cho), creatine (Cr), and N-acetylaspartate (NAA), while in the short TE spectrum additional signals from myo-inositol (mI), glutamate and glutamine (Glx), and lipids (Lip) are present.

MR Spectroscopy and Spectroscopic Imaging of the Brain

205

Table 9.1 Compounds detected by proton MRS in the human brain

Compounds normally present

Compounds which may be detected under pathological or other abnormal conditions

Large signals at long TE

Long TE

N-Acetylaspartate (NAA) Creatine (Cr) and phosphocreatine (PCr) Cholines (Cho): Glycerophosphocholine (GPC) Phosphocholine (PC), free choline (Cho)

Lactate (Lac) β-Hydroxy-butyrate, acetone

Large signals at short TE

Short TE

Glutamate (Glu) Glutamine (Gln) myo-Inositol (mI)

Lipids Macromolecules Phenylalanine Galactitol

Small signals (short or long TE)

Exogenous compounds (short or long TE)

N-Acetylaspartylglutamate (NAAG), aspartate Taurine, betaine, scyllo-inositol, ethanolamine Threonine Glucose, glycogen Purine nucleotides Histidine

Propan-1,2-diol Mannitol Ethanol Methylsulfonylmethane (MSM)

Succinate, pyruvate Alanine Glycine

Small signals that can be detected with the use of spectral editing techniques γ-Amino-butyric acid (GABA) Homocarnosine Glutathione Threonine Vitamin C (ascorbic acid)

2.1. N-Acetylaspartate

NAA is the largest signal in the normal adult brain spectrum, resonating at 2.01 ppm, with a small and usually unresolved contribution from N-acetylaspartylglutamate (NAAG) at 2.04 ppm (9, 10). NAA is one of the most abundant amino acids in the central nervous system. It has been speculated to be a source of acetyl groups for lipid synthesis, a regulator of protein synthesis, a storage form of acetyl-CoA or aspartate, a breakdown product of NAAG (which, unlike NAA, is a neurotransmitter), or an osmolyte (11). NAA is synthesized in neuronal mitochondria, from aspartate and acetyl-coA. NAA is often referred to as a “neuronal marker,” since immunocytochemical studies have suggested that NAA is predominantly restricted to neurons, axons,

206

Zhu and Barker

and dendrites within the central nervous system (12). However, other studies have suggested that NAA may be found in nonneuronal cells, such as mast cells or isolated oligodendrocyte preparations (13–15). Overall, NAA does appear to be a good surrogate marker of neuronal health, but (as with all surrogate markers) it may sometimes change independent of neuron cell density or function. 2.2. Choline

The “choline” signal (“Cho,” 3.20 ppm) is a composite peak consisting of contributions from the trimethylamine groups of glycerophosphocholine (GPC), phosphocholine (PC), and a small amount of free choline itself (16). These compounds are involved in membrane synthesis and degradation, and they are often elevated in disease states where increased membrane turnover is involved (e.g., tumors). Glial cells have also been reported to have high levels of Cho (17, 18). Other pathological processes which lead to Cho elevation include active demyelination (19), either resulting from the degradation of myelin phospholipids primarily to GPC or perhaps due to inflammation (20). Low brain Cho has been observed in hepatic encephalopathy (21), and there is also some evidence to suggest that dietary intake of choline can modulate cerebral Cho levels (22).

2.3. Creatine

The “creatine” methyl resonance (“Cr,” 3.03 ppm) is a composite peak consisting of both creatine and phosphocreatine, compounds that are involved in energy metabolism via the creatine kinase reaction, generating ATP. A resonance from the CH2 of creatine can also be observed at 3.91 ppm. In vitro, glial cells contain a two- to fourfold higher concentration of creatine than do neurons (23). Creatine also shows quite large regional variations, with lower levels in white matter than gray matter in normal brain, as well as very high levels of Cr in the cerebellum compared to supratentorial regions (24).

2.4. Lactate

The lactate resonance (a doublet with a 7 Hz coupling constant centered at 1.31 ppm) is usually not detectable in the brain under normal conditions. However, lactate is often detected by MRS in pathological conditions such as acute hypoxic (25) or ischemic (5, 26) injury, or in brain tumors (27) or mitochondrial diseases (7, 28).

2.5. myo-Inositol

One of the larger signals in short echo time spectra occurs from myo-inositol (mI) at 3.5–3.6 ppm. mI is a pentose sugar, which is part of the inositol triphosphate intracellular second messenger system. Glial cells in vitro have been shown to contain higher levels of mI than neurons (29, 30). mI has been reported to be reduced in hepatic encephalopathy (31), and increased in Alzheimer’s dementia (32) and demyelinating diseases (33).

MR Spectroscopy and Spectroscopic Imaging of the Brain

207

2.6. Glutamate and Glutamine

Glutamate (Glu) is the most abundant amino acid in the brain and is the dominant neurotransmitter (34). At 1.5 T, there is almost complete overlap of Glu and glutamine (Gln), and they are detected as a composite “Glx” peak (21). At higher fields (3.0 T and above), Glu and Gln become better resolved and can be quantified individually with good accuracy using appropriate spectral analysis techniques (35). Glu has been found to be elevated in MS plaques (36), and elevated cerebral Gln is commonly observed in patients with liver failure (for example, hepatic encephalopathy (31) and Reye’s syndrome (37)).

2.7. Less Commonly Detected Compounds

Approximately 25 additional compounds have been detected in proton spectra of the human brain (Table 9.1). Some of these compounds are present in the normal human brain, but are difficult to detect routinely because they are very small and/or have overlapping peaks. Some examples of these compounds include NAAG, aspartate, taurine, scyllo-inositol, betaine, ethanolamine, purine nucleotides, histidine, glucose, and glycogen (38). Other compounds are yet more difficult to detect and require the use of “spectral editing” techniques (see later), because in conventional spectra they overlap and are obscured by much larger signals. Examples of compounds requiring spectral editing to be measured include γ-amino-butyric acid (GABA) and glutathione (GSH) (39, 40). Some compounds are only detected under disease or other abnormal conditions. Examples include the ketone bodies βhydroxy-butyrate and acetone (41, 42) in patients who are ketotic and other compounds such as phenylalanine (in phenylketonurea (43)), galactitol, ribitol, arabitol in “polyol disease” (44), and succinate, pyruvate, alanine, glycine, and threonine in various disorders. Exogenous compounds which are able to cross the blood– brain barrier may be detected by proton MRS; examples include the drug delivery vehicle propan-1,2-diol (45), ethanol (46), and methylsulfonylmethane (MSM) (47). Histidine, homocarnosine, and the amide resonance of NAA are low signal intensity compounds downfield from water which can be detected by the use of short echo times, appropriate water suppression methods, and high magnetic field strengths. Using oral loading of histidine, Vermathen et al. were able to estimate brain pH from the chemical shift difference of the C2 and C4 resonances of the imidazole side chain of histidine (48); similarly, Rothman et al. were able to use the imidazole resonances of homocarnosine to estimate brain pH in epilepsy patients who were receiving vigabatrin (49). The rate of exchange of the NAA amide protons with water is also pH sensitive and can be used to estimate brain pH (50).

208

Zhu and Barker

3. Spatial Localization Techniques 3.1. Single-Voxel Techniques

Nearly all single-voxel localization techniques use three orthogonal slice-selective pulses to select a signal from the region (“voxel”) where they intersect (Fig. 9.2a). Signals from outside the voxel are removed by the use of “crusher” field gradient

Fig. 9.2. Single-voxel localization techniques: (a) spatial localization is achieved by collecting signals from the intersection of three slice-selective RF pulses applied in orthogonal directions; (b) the STEAM sequence, consisting of three 90◦ slice-selective pulses; (c) the PRESS sequence, consisting of a slice-selective 90◦ excitation pulse and two 180◦ sliceselective refocusing pulses; (d) the LASER sequence, which uses a non-slice-selective adiabatic half-passage excitation pulse, followed by three pairs of hyperbolic secant 180◦ refocusing pulses; (e) the semi-LASER sequence, which uses a slice-selective 90◦ excitation pulse and two pairs of hyperbolic secant 180◦ refocusing pulses; (f) the “SPECIAL” pulse sequence, which uses an alternating slice-selective 180◦ inversion pulse (every second average) in combination with a 90◦ –180◦ bar-selective spin echo.

MR Spectroscopy and Spectroscopic Imaging of the Brain

209

pulses, alternating the phases of the slice-selective pulses and receiver (‘phase-cycling’), and the use of outer-volume suppression pulses (51, 52) (53, 54). Typical voxel sizes for human brain spectroscopy are 4–8 cm3 . The “STEAM” sequence (52, 55) (Fig. 9.2b) uses three 90◦ pulses to form a “stimulated echo,” while the “PRESS” sequence (Fig. 9.2c) uses one 90◦ and two 180◦ refocusing pulses to create a spin echo. STEAM and PRESS have been compared in detail (56); perhaps the biggest difference is that the spin-echobased PRESS sequence has twice the signal compared to STEAM and is therefore often preferred. However, advantages of STEAM include better slice profiles and higher bandwidth of the 90◦ pulses, lower RF power requirements, and the ability to obtain shorter echo times. In this regard, STEAM may be particularly advantageous for brain MRS at high field strengths (e.g., above 3 T (35)). Short TE STEAM may be preferable for observing resonances with shorter T2 s (57), while long TE PRESS (with its superior SNR) should generally be used for resonances with longer T2 s (such as Cho, Cr, NAA, and lactate). In vivo MRS performed at high field strengths (e.g., 3 T or higher) is associated with additional technical challenges. For instance, uniform RF transmit (B1 ) fields become difficult to achieve because of wavelength effects in volume RF coils (58) or when using inhomogeneous surface coils for excitation. In either case, it may be difficult to achieve the desired flip angles in PRESS or STEAM, and the flip angles may vary inside the voxel, resulting in signal loss. To address these problems, adiabatic excitation or refocusing pulses have been implemented in techniques such as “LASER” (localization by adiabatic selective refocusing; 59, 60) (Fig. 9.2d) or its simplified version “semi-LASER” (61, 62) (Fig. 9.2e). The LASER sequence consists of a non-slice-selective adiabatic half-passage 90◦ pulse for excitation and three pairs of hyperbolic secant (HS) refocusing pulses in three directions for localization. Since a single HS 180◦ pulse with a slice-selection gradient produces a large first-order phase variation across the spectrum, two consecutive HS pulses are needed to cancel it out (63, 64). The LASER sequence produces a more uniform excitation profile and takes advantage of the large bandwidths of the adiabatic HS pulses to reduce chemical shift displacement errors. However, the large number of RF pulses used in LASER results in higher RF power requirement and longer TE compared to conventional localization sequences. The semi-LASER sequence consists of a non-adiabatic 90◦ slice-selective pulse and two pairs of adiabatic HS pulses for refocusing as in LASER; while some insensitivity to B1 inhomogeneity is lost, this sequence does have reduced RF power and can achieve shorter TE than LASER. At very high fields for human MRS such as 7 T, apparent metabolite T2 relaxation times are significantly shorter than at

210

Zhu and Barker

lower field strengths, and it is therefore desirable to minimize the TE of the localization sequence as much as possible. A localization technique dubbed “SPECIAL” (spin-echo full-intensity acquired localized spectroscopy; 65) () combines desirable features such as the full signal intensity of PRESS and the shorter TE of STEAM. The sequence consists of a slice-selective inversion pulse followed by a spin-echo sequence, with each pulse applied in a different direction (Fig. 9.2f). The sequence collects full-intensity signal from a 1D strip-like volume defined by the intersection of the selected slices of the 90◦ and 180◦ pulses. The slice-selective inversion pulse is applied to every other TR (similar to what is used in the “ISIS” experiment (66)) so that a minimum of two scans are required to achieve full spatial localization. This sequence is a promising technique to investigate compounds with short T2 s in vivo. 3.2. Multiple-Voxel (MRSI) Techniques

While single-voxel MRS can be performed quickly and easily in most parts of the human brain, it provides no information on the spatial variations of metabolites and is generally limited to one or two brain regions in most clinical studies. In contrast, MRSI is usually more time-consuming but can be used to measure multiple-voxel locations simultaneously. Most often, MRSI is based on signal excitation of a restricted region using the PRESS sequence in combination with phaseencoding in two directions (Fig. 9.3) (55). This allows B0 field homogeneity to be optimized on the desired region of interest, limits the number of phase-encoding steps needed for a given spatial resolution, and avoids exciting lipid signals from the scalp. Other methods for lipid suppression are discussed later.

Fig. 9.3. 2D-PRESS-MRSI pulse sequence. A PRESS sequence is used to excite a large volume of brain tissue while excluding signal from lipid in the scalp and/or regions of poor field homogeneity, and then phase-encoding gradients (in blue, GY and GZ ) are used to localize spectra from regions within the excited region. A CHESS prepulse and crusher gradient are applied for water suppression. Crusher gradients applied around the 180◦ refocusing pulses are also shown.

MR Spectroscopy and Spectroscopic Imaging of the Brain

211

Fig. 9.4. 2D-PRESS-MRSI scan (3 T, TR/TE 1,700/135 ms, nominal voxel size 1.5 cm3 ) in the axial plane of a 3-year-old girl with developmental delay. The central 6×6 spectra are shown (indicated in red on the localizer T2 -weighted MRI).

Figure 9.4 shows the results of this sequence from a 3-year-old girl with an idiopathic developmental delay. However, PRESS-MRSI also has shortcomings, including unreliable spectra at the edges of the PRESS box due to imperfect slice profiles of the 180◦ pulses, the inability to perform multislice acquisitions (although 3D is possible), and the difficulty in covering to the edges of the brain because of the rectangular shape of the PRESS excitation. An alternative approach is a sliceselective spin-echo sequence that excites a whole transverse slice (Fig. 9.5) and which can be used in a multi-slice mode (53). Preceding the spin-echo sequence there are usually multiple, carefully placed OVS pulses to suppress the lipid signals from the scalp (53), as well as the usual water suppression pulses. Figure 9.6 shows an example of one slice from a multi-slice 2D MRSI data set from a normal volunteer. To minimize artifacts due to residual water and lipid and also field inhomogeneity, MRSI with large spatial coverage is typically performed at long echo time (e.g., 140 or 280 ms). The multi-slice technique can allow a sufficient number of slices to cover the whole brain but the resulting scan time can be too long with conventional phase-encoding techniques. Specifically, the length of the pulse sequence for each slice is in the range of 0.5–1.0 s including all RF pulses and data acquisition window, which needs to be long enough to gain enough spectral resolution. Four or five such slices interleaved can result in a TR prohibitively long that causes long scan times (67). 3D-PRESSMRSI, on the other hand, can also lead to very long scan times if large brain coverage is prescribed. The number of phase-encoding steps (N) is equal to the field of view (FOV) divided by the desired

212

Zhu and Barker

Fig. 9.5. A multi-slice 2D-MRSI pulse sequence. A slice-selective spin echo is preceded by an optimized water and lipid suppression scheme (“HGDB”), including outer-volume suppression (OVS) pulses (8, indicated in red) for spatial suppression of scalp signals (103). 2D phase-encoding gradients (blue) are applied on GY and GZ . In this example, three slices are collected within one repetition time (TR). Gradients associated with the water and lipid suppression are omitted for clarity.

spatial resolution (N = FOV/ ). Therefore, in order to minimize the scan time (i.e., minimize N) without reducing a desired spatial resolution, it is important to prescribe a FOV as small as possible constrained only by the dimensions of the object to be imaged. In the case of brain imaging, the left–right FOV should be smaller than the anterior–posterior, since the brain (usually of an oval shape) is smaller in this dimension (68). In general, scan time can be reduced by an additional 25–30% with a reduced FOV in the left–right direction. Generally, if large FOVs and high resolutions are sought in all three dimensions, fast MRSI techniques are required to maintain clinically reasonable scan times. 3.3. Fast MRSI Techniques

A number of different approaches for fast MRSI have been developed and reviewed previously (69). Some of the more frequently used methods are discussed here.

3.3.1. “Turbo” or Fast Spin-Echo MRSI

One of the earliest approaches to fast MRSI was to use a multipleecho acquisition, with each echo having its own phase-encoding gradient (70) (Fig. 9.7a). Typically 3 or 4 echoes are used, reducing scan time by a factor of 3 or 4 if the same repetition time (TR) is used as in a single-echo experiment. This approach does suffer from several limitations, however: the spectral resolution is limited by having to keep each echo readout short, while the later echoes suffer from reduced signal due to T2 relaxation. Compounds with short T2 relaxation times cannot be observed with this technique. For multiple echoes, the minimum TR is generally

MR Spectroscopy and Spectroscopic Imaging of the Brain

213

Fig. 9.6. Example data from one slice (at the level of the lateral ventricles) of a multi-slice 2D-MRSI data from a normal human subject recorded at 3.0 T using the pulse sequence of Fig. 9.5. TR 2.5 s, TE 140 ms, nominal voxel size 0.65 cm3 . In addition to the T1 -weighted MRI scan, spectroscopic images of choline, creatine, and N-acetylaspartate and selected spectra from the left and right hemispheres are shown.

longer than for a single-echo acquisition, so that time-savings may be less than expected. For these reasons, multi-echo MRSI has only seen limited adoption in practice, although it has been applied to studies of brain tumors (71). 3.3.2. Echo-Planar (EPSI) and Spiral-MRSI

In EPSI, an oscillating read gradient is applied during data acquisition so that both spectral and spatial information are collected simultaneously (72) (73). The oscillating read gradient can be viewed as repeatedly collecting one line of k-space at different time points. Conventional phase-encoding is then applied in the other one or two directions to extend the experiment to either two or three spatial dimensions, respectively (Fig. 9.7b). The EPSI readout reduces the number of phase-encoding steps by an order of magnitude compared to conventional MRSI, thereby achieving a large scan time reduction. EPSI-encoding can be

214

Zhu and Barker

Fig. 9.7. Different readout strategies for fast MRSI pulse sequences. In the examples shown here, all three sequences use spin-echo excitation preceded by CHESS water suppression and OVS lipid suppression, although other excitation and suppression sequences can be used. (a) In fast spin-echo (“turbo”) MRSI, multiple spin echoes are acquired, each one with its own phase-encoding gradient; (b) in echo-planar spectroscopic imaging (EPSI), an oscillating read gradient is applied during data acquisition; and (c) in spiral-MRSI, two oscillating read gradients are applied during data acquisition. For full 3D-encoding, conventional phase-encoding gradients can be applied in the remaining directions.

implemented as 2D multi-slice with spin-echo excitation, or 3D with a PRESS or spin-echo thick slab excitation (74, 75). EPSI is one of the fastest acquisition techniques for MRSI, but does have some limitations. High-performance gradient systems are required, and any imbalances between positive and negative gradient lobes can lead to “ghost” artifacts in the metabolic images. This problem can be addressed by processing the readout lines of two opposite directions separately, with subsequent combination of the two data sets after spatial transformation. This solution, however, reduces the available spectral bandwidth, which may already be quite low in EPSI (76). SNR may also be slightly lower than conventional MRSI recorded in the same scan time depending on the readout gradient waveform and whether ramp sampling is used or not. Because of its speed and much improved post-processing methods, EPSI (also called “PEPSI” (proton EPSI)) has been gaining popularity in whole-brain 3D MRSI (75) and has also been used to study various brain pathologies (77, 78). Spiral-MRSI is similar to EPSI in that read gradients are applied during data acquisition (Fig. 9.7c). In spiral-MRSI, however, gradient waveforms in 2D are applied so that k-space data

MR Spectroscopy and Spectroscopic Imaging of the Brain

215

are sampled from the center to the edge along a spiral trajectory. These gradient waveforms are repeated several times concurrent with the evolution of the readout time in one TR. The k,t-space can be filled using multiple shots to satisfy the desired FOV and spatial and spectral resolutions (79). Post-processing of spiral-MRSI data usually starts with a “regridding” procedure (80) where raw data are resampled onto a Cartesian k-space grid by interpolation, after which conventional MRSI processing by Fourier transformation can be done. Spiral-MRSI has some unique advantages over EPSI, including the ability to manipulate the point-spread function, scan time, and SNR by varying the sample spacing in k-space with a variable density trajectory (81, 82). In addition, if the center of k-space is collected at the beginning of every spiral readout, it is possible to correct errors in phase or frequency associated with motion or other processes (83). Like EPSI, spiral-MRSI applies a read gradient during data acquisition, so it shares a similar level of dependence on the gradient system performance. EPSI is being applied to clinical applications, but is partly hampered due to lack of commercial availability and the need for dedicated reconstruction software. However, spiral-MRSI has been used to map metabolic abnormalities in patients with multiple sclerosis (84). 3.3.3. Parallel Encoded-MRSI

Parallel imaging techniques originally developed for speeding up MRI can also be adopted for MRSI (85, 86). The basic principle is to use the different sensitivity profiles of multiple, phasedarray receiver coils to encode spatial information, so that fewer phase-encoding steps are required, thereby reducing scan time. In the “SENSE” approach, Fourier transformation of undersampled k-space data leads to “aliased” spectroscopic images from each channel, which can then be unfolded and reconstructed using each coil’s sensitivity profile to produce a single spectroscopic image with uniform sensitivity. Alternatively, algorithms such as “SMASH” and “GRAPPA” (87–89) can be implemented to interpolate the missing k-space data points, which are then Fourier transformed as in conventional MRSI. Both “SENSE” (90) and “GRAPPA” (91) MRSI have been successfully implemented in humans. Reducing scan time using parallel encoding is an attractive option since it can be performed with any existing MRSI pulse sequence. 2D- and 3D-MRSI involve phase-encoding in multiple directions, so SENSE-encoding can also be performed in 2D or 3D (provided that receive arrays with appropriate geometry are available) leading to large scan time reductions (90). An example of a multi-slice MRSI scan with a SENSE factor of 6=3×2, in AP and RL directions, applied using a 32-channel head coil, is shown in a patient with a brain tumor in Fig. 9.8. This scan took 5:05 min; with conventional phase-encoding, scan

216

Zhu and Barker

Fig. 9.8. An example of 2D-SENSE-MRSI in a patient with a high-grade left frontal glioma recorded post-treatment at 3.0 T. TR 2.5 s, TE 140 ms, nominal voxel size 0.65 cm3 , SENSE acceleration factor = 6, scan time 5:05 min. FLAIR and noncontrast T1 -weighted (“MP-RAGE”) MR images, spectroscopic images (choline, creatine, NAA), and selected spectra are shown. The core of the lesion is characterized by reduced levels of NAA and other metabolites, consistent with necrotic tissue, while the T2 hyperintense perilesional areas demonstrate elevated Cho compared to the contralateral hemisphere, consistent with residual or recurrent tumor.

time would have been more than 30 min. Scan times on the order of 1–2 min can be achieved by combining SENSE-MRSI with other fast MRSI techniques, such as Turbo MRSI (92) or EPSI (93), as long as sufficient SNR is available. Parallel MRSI also has some potential problems. Errors in the coil sensitivity profiles and/or use of too high SENSE factors will lead to incomplete unfolding of MRSI data and the presence of artifacts. Unfolding of strong peri-cranial lipid signals is particularly challenging; successful SENSE-MRSI requires the application of efficient lipid suppression techniques (94) (see later). 3.4. Other Approaches to MRSI of the Brain

Other approaches to MRSI of the brain exist that are based on a t1 evolution period (as in high-resolution 2D NMR spectroscopy) to encode spectral information, with a fast imaging readout to determine spatial information. Scan times for these methods depend on the number of t1 values required to obtain sufficient spatial resolution and may be relatively short compared to conventional MRSI. These methods show promise in small animal studies, but for the most part have not been applied in humans. The sensitivity may be somewhat lower than in conventional MRSI because of the T2 signal decay that occurs during t1 time period. However, they do offer some unique advantages, such as using a

MR Spectroscopy and Spectroscopic Imaging of the Brain

217

constant evolution time with a sliding refocusing pulse, to produce “homonuclear decoupled” proton brain spectra (95).

4. Water and Lipid Suppression

Brain metabolites observed by MRS are in the millimolar concentration range, while brain water is approximately 80 M. Lipids in scalp tissue are also present in very high concentrations. Therefore, efficient water suppression (and lipid suppression for MRSI) is vital for the reliable observation and measurement of brain metabolites. The most common method for water suppression is to presaturate the water signal using frequency-selective saturation pulses applied prior to the localization sequence (“CHESS”) (96). Multiple CHESS pulses with optimized flip angles and delays can be used to give good suppression factors over a range of transmit B1 values and water T1 relaxation times (which is important for the suppression of both brain water and CSF) (97, 98). For example, the “WET” scheme employs up to four Gaussian pulses (98), while the “VAPOR” scheme consists of seven pulses (57). Occasionally, it may be desirable to perform water suppression during the localization sequence, either opposed, or in addition, to presaturation. For instance, a water suppression pulse can be placed in the STEAM sequence between the second and third 90◦ pulses because the magnetization of the stimulated echo pathway is stored along the Z-axis during this time period. This can be used to improve suppression compared to presaturation only. It is also possible to include frequency-selective refocusing pulses inside a localization sequence to improve water suppression (e.g., the “MEGA” or “BASING” sequences) (99, 100). Lipid suppression is commonly performed in three different ways. One method, to suppress lipid signals in the scalp, is to use spatial outer-volume suppression (OVS) pulses (Fig. 9.5) (53). Alternatively, an inversion recovery scheme can be used, taking advantage of the difference in T1 values between lipid (typically 300 ms at 1.5 T) and metabolites (1,000–2,000 ms) (101). At 1.5 T, an inversion time of 200 ms (= T1 ∗ ln [2]) will selectively null the lipid signal, while most of the metabolite magnetization remains inverted. This method can suppress lipid signals anywhere in the brain because no assumption is made about the spatial distribution of the lipid, but may somewhat reduce metabolite SNR. Finally, lipid suppression may also be performed using frequencyselective saturation pulses, similar to frequency-selective water suppression techniques such as CHESS (94). Recently, methods

218

Zhu and Barker

have been designed which combine both water and lipid suppression into a single, dual band approach (94, 102, 103).

5. Spectral Editing Techniques

As mentioned above, certain molecules, such as GABA or glutathione, are almost totally obscured in conventional brain MRS by signals from other compounds which are present at much higher concentrations. “Spectral editing” techniques are required in order to detect these molecules while suppressing the signal from the unwanted compounds. A commonly used sequence for this purpose is the so-called “MEGA-PRESS” sequence (Fig. 9.9). Spectral editing makes use of molecules which contain “coupled” spin systems – the presence of coupling (J, measured in Hz) between functional groups allows the signal on one group to be modulated by applying a selective radiofrequency pulse on the other. For instance, for GABA, setting the frequency of the selective editing pulse to 1.9 ppm resonance of GABA will refocus the outer two peaks of the 3.02 ppm GABA pseudo-triplet. A second scan is performed without the selective pulse, and with TE=1/J (68 ms), the unaffected modulation results in two inverted peak (the outer two lines of the triplet) at 3.02 ppm. By subtracting the first scan from the second, the 3.02 GABA resonance can be selected (Fig. 9.10a) (99). For glutathione, the resonance at 2.95 ppm can be observed with the selective editing pulse set to 4.56 ppm and TE of 130 ms (Fig. 9.10b). A similar approach

Fig. 9.9. Pulse sequence for spectral editing (“MEGA-PRESS”). Frequency-selective editing pulses are added (blue) to the conventional PRESS sequence on alternating scans. The TE (= TE1 +TE2 ) is set equal to 1/J for doublets. Subtraction of alternating scans causes cancellation of all signals not effected by the editing pulses, leaving only the target edited molecules.

MR Spectroscopy and Spectroscopic Imaging of the Brain

219

Fig. 9.10. Examples of (a) GABA- and (b) glutathione-edited spectra using the MEGAPRESS pulse sequence. TR 2 s, voxel size 3.5×3.5×3.5 = ∼43 cm3 , centered on the anterior cingulate gyrus, with TE 68 ms for GABA and 130 ms for GSH. Scan time was 8 min 32 s for GABA and 17 min 4 s for GSH. Note that in (a) glutamate/glutamine (Glx) coedit with GABA, and in (b) the aspartyl resonances of NAA coedit with GSH.

(called the “BASING” sequence) can be used to selectively detect brain lactate without contamination from lipids, making use of the coupled lactate resonances at 1.3 and 4.1 ppm (104).

6. Data Analysis and Quantification The concentration of a metabolite is linearly proportional to its spectral peak area. However, peak area measurements in in vivo spectroscopy are complicated by resonance overlap, baseline distortions, and non-ideal lineshapes and will also depend on factors such as relaxation times, pulse sequence used, and scanner hardware (e.g., receiver gain, coil loading). Various methods have been used to measure peak areas, ranging from simple integration, to fitting algorithms in the time- or frequency domains

220

Zhu and Barker

Fig. 9.11. An example of the “LCModel” analysis method. The experimental data are fit as a linear combination of spectra of pure compounds recorded under the same experimental conditions as the in vivo spectrum. Automated baseline and phase correction is performed, and an estimate of metabolite concentrations provided relative to the brain water signal. In this example of a 2×2×2 cm PRESS spectrum recorded at 3.0 T from a normal control subject (TR/TE/number of averages = 2,000/35/128), the difference between the original data and the curve-fit (red) is shown in the top trace. Metabolite concentrations in blue correspond to those with an estimated uncertainty of less than 20%.

(105, 106). One of the more widely used methods for spectral quantitation in recent years is the linear combination model (“LCModel”) method developed by Provencher et al. (107) (Fig. 9.11). The LCModel fits the in vivo spectrum as a combination of pure, model spectra from each of the expected compounds in the brain (107). The model also includes automatic phase correction and baseline correction, or the baseline may also be modeled as a combination of macromolecular resonances. Provided that each scanner is properly calibrated with the appropriate model solutions, the program returns metabolite concentrations (relative to an unsuppressed water signal) as well as estimates of uncertainty (e.g., Cramer–Rao lower bounds). Quantification methods based on internal or external standards have been extensively developed and tested for single-voxel spectroscopy (108). With care, it is also possible to quantify spectroscopic images (109). Many studies, however, do not attempt to quantify metabolite concentrations, but rather report relative amounts (ratios) of each metabolite, often using Cr as a reference. While ratios have some inherent advantages, for instance to account for partial volume effects or to enhance spectroscopic “contrast” in conditions where metabolites may change in opposite directions (e.g., Cho increases, NAA decreases), they also may be misleading if all metabolites are changing simultaneously. In

MR Spectroscopy and Spectroscopic Imaging of the Brain

221

particular, Cr shows quite marked regional variations (24), and often changes in pathology, so caution should be used when interpreting ratios of metabolites to Cr.

7. Conclusions MRS and MRSI are mature techniques that are very commonly used for research studies in both humans and animal models. The protocols described in this chapter represent the most widely used and validated techniques currently used, but are by no means comprehensive. MRS protocol development, particularly for high-field applications, hyper-polarization, and fast MRSI techniques continue to be an active area of research investigation.

Acknowledgments Supported in part by National Institutes of Health grant P41 RR015241. References 1. Cady, E. B., Costello, A. M., Dawson, M. J., Delpy, D. T., Hope, P. L., Reynolds, E. O., Tofts, P. S., Wilkie, D. R. Noninvasive investigation of cerebral metabolism in newborn infants by phosphorus nuclear magnetic resonance spectroscopy. Lancet 1983;1(8333):1059–1062. 2. Bottomley, P. A., Edelstein, W. A., Foster, T. H., Adams, W. A. In vivo solventsuppressed localized hydrogen nuclear magnetic resonance spectroscopy: A window to metabolism? Proc Natl Acad Sci USA 1985;82(7):2148–2152. 3. Barker, P. B., Lin, D. D. In vivo proton MR spectroscopy of the human brain. Prog NMR Spect 2006;49:99–128. 4. Goldman, M., Johannesson, H., Axelsson, O., Karlsson, M. Hyperpolarization of 13C through order transfer from parahydrogen: A new contrast agent for MRI. Magn Reson Imaging 2005;23(2):153–157. 5. Barker, P. B., Gillard, J. H., van Zijl, P. C., Soher, B. J., Hanley, D. F., Agildere, A. M., Oppenheimer, S. M., Bryan, R. N.

6.

7.

8.

9.

Acute stroke: Evaluation with serial proton MR spectroscopic imaging. Radiology 1994;192(3):723–732. Remy, C., Grand, S., Lai, E. S., Belle, V., Hoffmann, D., Berger, F., Esteve, F., Ziegler, A., Le Bas, J. F., Benabid, A. L., Decorps, M., Segebarth, C. M. 1H MRS of human brain abscesses in vivo and in vitro. Magn Reson Med 1995;34(4):508–514. Lin, D. D., Crawford, T. O., Barker, P. B. Proton MR spectroscopy in the diagnostic evaluation of suspected mitochondrial disease. AJNR Am J Neuroradiol 2003;24(1):33–41. Govindaraju, V., Young, K., Maudsley, A. A. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 2000;13(3):129–153. Frahm, J., Michaelis, T., Merboldt, K. D., Hanicke, W., Gyngell, M. L., Bruhn, H. On the N-acetyl methyl resonance in localized 1H NMR spectra of human brain in vivo. NMR Biomed 1991;4(4): 201–204.

222

Zhu and Barker

10. Pouwels, P. J., Frahm, J. Differential distribution of NAA and NAAG in human brain as determined by quantitative localized proton MRS. NMR Biomed 1997;10(2):73–78. 11. Barker, P. B. N-acetyl aspartate – a neuronal marker? Ann Neurol 2001;49(4):423–424. 12. Simmons, M. L., Frondoza, C. G., Coyle, J. T. Immunocytochemical localization of Nacetyl-aspartate with monoclonal antibodies. Neuroscience 1991;45(1):37–45. 13. Bhakoo, K. K., Pearce, D. In vitro expression of N-acetyl aspartate by oligodendrocytes: Implications for proton magnetic resonance spectroscopy signal in vivo. J Neurochem 2000;74(1):254–262. 14. Burlina, A. P., Ferrari, V., Facci, L., Skaper, S. D., Burlina, A. B. Mast cells contain large quantities of secretagogue-sensitive Nacetylaspartate. J Neurochem 1997;69(3): 1314–1317. 15. Urenjak, J., Williams, S. R., Gadian, D. G., Noble, M. Specific expression of Nacetylaspartate in neurons, oligodendrocytetype-2 astrocyte progenitors, and immature oligodendrocytes in vitro. J Neurochem 1992;59(1):55–61. 16. Barker, P., Breiter, S., Soher, B., Chatham, J., Forder, J., Samphilipo, M., Magee, C., Anderson, J. Quantitative proton spectroscopy of canine brain: In vivo and in vitro correlations. Magn Reson Med 1994;32: 157–163. 17. Gill, S. S., Small, R. K., Thomas, D. G., Patel, P., Porteous, R., van Bruggen, N., Gadian, D. G., Kauppinen, R. A., Williams, S. R. Brain metabolites as 1H NMR markers of neuronal and glial disorders. NMR Biomed 1989;2(5–6):196–200. 18. Gill, S. S., Thomas, D. G., Van, B. N., Gadian, D. G., Peden, C. J., Bell, J. D., Cox, I. J., Menon, D. K., Iles, R. A., Bryant, D. J. Proton MR spectroscopy of intracranial tumours: In vivo and in vitro studies. J Comput Assist Tomogr 1990;14(4):497–504. 19. Davie, C. A., Hawkins, C. P., Barker, G. J., Brennan, A., Tofts, P. S., Miller, D. H., McDonald, W. I. Detection of myelin breakdown products by proton magnetic resonance spectroscopy. Lancet 1993;341(8845):630–631. 20. Brenner, R. E., Munro, P. M., Williams, S. C., Bell, J. D., Barker, G. J., Hawkins, C. P., Landon, D. N., McDonald, W. I. The proton NMR spectrum in acute EAE: The significance of the change in the Cho:Cr ratio. Magn Reson Med 1993;29(6):737–745. 21. Kreis, R., Ross, B. D., Farrow, N. A., Ackerman, Z. Metabolic disorders of the brain in chronic hepatic encephalopathy

22.

23.

24.

25.

26.

27.

28.

29.

30.

31. 32.

detected with H-1 MR spectroscopy. Radiology 1992;182(1):19–27. Stoll, A. L., Renshaw, P. F., De Micheli, E., Wurtman, R., Pillay, S. S., Cohen, B. M. Choline ingestion increases the resonance of choline-containing compounds in human brain: An in vivo proton magnetic resonance study. Biol Psychiatry 1995;37(3):170–174. Urenjak, J., Williams, S. R., Gadian, D. G., Noble, M. Proton nuclear magnetic resonance spectroscopy unambiguously identifies different neural cell types. J Neurosci 1993;13:981–989. Jacobs, M. A., Horska, A., van Zijl, P. C., Barker, P. B. Quantitative proton MR spectroscopic imaging of normal human cerebellum and brain stem. Magn Reson Med 2001;46(4):699–705. Penrice, J., Cady, E. B., Lorek, A., Wylezinska, M., Amess, P. N., Aldridge, R. F., Stewart, A., Wyatt, J. S., Reynolds, E. O. Proton magnetic resonance spectroscopy of the brain in normal preterm and term infants, and early changes after perinatal hypoxiaischemia. Pediatr Res 1996;40(1):6–14. Petroff, O. A., Graham, G. D., Blamire, A. M., al-Rayess, M., Rothman, D. L., Fayad, P. B., Brass, L. M., Shulman, R. G., Prichard, J. W. Spectroscopic imaging of stroke in humans: Histopathology correlates of spectral changes. Neurology 1992;42(7): 1349–1354. Alger, J. R., Frank, J. A., Bizzi, A., Fulham, M. J., DeSouza, B. X., Duhaney, M. O., Inscoe, S. W., Black, J. L., van Zijl, P. C., Moonen, C. T. Metabolism of human gliomas: Assessment with H-1 MR spectroscopy and F-18 fluorodeoxyglucose PET. Radiology 1990;177(3):633–641. Mathews, P. M., Andermann, F., Silver, K., Karpati, G., Arnold, D. L. Proton MR spectroscopic characterization of differences in regional brain metabolic abnormalities in mitochondrial encephalomyopathies. Neurology 1993;43(12):2484–2490. Brand, A., Richter-Landsberg, C., Leibfritz, D. Multinuclear NMR studies on the energy metabolism of glial and neuronal cells. Dev Neurosci 1993;15(3–5):289–298. Flogel, U., Willker, W., Leibfritz, D. Regulation of intracellular pH in neuronal and glial tumour cells, studied by multinuclear NMR spectroscopy. NMR Biomed 1994;7(4): 157–166. Ross, B. D., Michaelis, T. Clinical applications of magnetic resonance spectroscopy. Magn Reson Q 1994;10:191–247. Shonk, T. K., Moats, R. A., Gifford, P., Michaelis, T., Mandigo, J. C., Izumi, J.,

MR Spectroscopy and Spectroscopic Imaging of the Brain

33.

34. 35.

36.

37.

38.

39.

40.

41.

42.

43.

Ross, B. D. Probable Alzheimer disease: Diagnosis with proton MR spectroscopy. Radiology 1995;195(1):65–72. Kruse, B., Hanefeld, F., Christen, H. J., Bruhn, H., Michaelis, T., Hanicke, W., Frahm, J. Alterations of brain metabolites in metachromatic leukodystrophy as detected by localized proton magnetic resonance spectroscopy in vivo. J Neurol 1993;241(2): 68–74. Magistretti, P. J., Pellerin, L., Rothman, D. L., Shulman, R. G. Energy on demand. Science 1999;283(5401):496–497. Tkac, I., Andersen, P., Adriany, G., Merkle, H., Ugurbil, K., Gruetter, R. In vivo 1H NMR spectroscopy of the human brain at 7 T. Magn Reson Med 2001;46(3):451–456. Srinivasan, R., Sailasuta, N., Hurd, R., Nelson, S., Pelletier, D. Evidence of elevated glutamate in multiple sclerosis using magnetic resonance spectroscopy at 3 T. Brain 2005;128(Pt 5):1016–1025. Kreis, R., Pfenninger, J., Herschkowitz, N., Boesch, C. In vivo proton magnetic resonance spectroscopy in a case of Reye’s syndrome. Intensive Care Med 1995;21(3): 266–269. van Zijl, P. C. M., Barker, P. B. Magnetic Resonance Spectroscopy and Spectroscopic Imaging for the Study of Brain Metabolism. Imaging Brain Structure and Function, Vol. 820. New York, NY: Annals of the New York Academy of Sciences; 1997, pp 75–96. Rothman, D. L., Petroff, O. A., Behar, K. L., Mattson, R. H. Localized 1H NMR measurements of gamma-aminobutyric acid in human brain in vivo. Proc Natl Acad Sci USA 1993;90(12):5662–5666. Terpstra, M., Henry, P. G., Gruetter, R. Measurement of reduced glutathione (GSH) in human brain using LCModel analysis of difference-edited spectra. Magn Reson Med 2003;50(1):19–23. Pan, J. W., Telang, F. W., Lee, J. H., de Graaf, R. A., Rothman, D. L., Stein, D. T., Hetherington, H. P. Measurement of betahydroxybutyrate in acute hyperketonemia in human brain. J Neurochem 2001;79(3): 539–544. Seymour, K. J., Bluml, S., Sutherling, J., Sutherling, W., Ross, B. D. Identification of cerebral acetone by 1H-MRS in patients with epilepsy controlled by ketogenic diet. Magma 1999;8(1):33–42. Kreis, R., Pietz, J., Penzien, J., Herschkowitz, N., Boesch, C. Identification and quantitation of phenylalanine in the brain of patients with phenylketonuria by means of localized in vivo 1H magnetic-

44.

45.

46.

47.

48.

49.

50.

51. 52.

53.

54. 55.

223

resonance spectroscopy. J Magn Reson B 1995;107(3):242–251. van der Knaap, M. S., Wevers, R. A., Struys, E. A., Verhoeven, N. M., Pouwels, P. J., Engelke, U. F., Feikema, W., Valk, J., Jakobs, C. Leukoencephalopathy associated with a disturbance in the metabolism of polyols. Ann Neurol 1999;46(6):925–928. Cady, E. B., Lorek, A., Penrice, J., Reynolds, E. O., Iles, R. A., Burns, S. P., Coutts, G. A., Cowan, F. M. Detection of propan-1,2diol in neonatal brain by in vivo proton magnetic resonance spectroscopy. Magn Reson Med 1994;32(6):764–767. Hanstock, C. C., Rothman, D. L., Shulman, R. G., Novotny, E. J., Jr., Petroff, O. A., Prichard, J. W. Measurement of ethanol in the human brain using NMR spectroscopy. J Stud Alcohol 1990;51(2):104–107. Rose, S. E., Chalk, J. B., Galloway, G. J., Doddrell, D. M. Detection of dimethyl sulfone in the human brain by in vivo proton magnetic resonance spectroscopy. Magn Reson Imaging 2000;18(1):95–98. Vermathen, P., Capizzano, A. A., Maudsley, A. A. Administration and (1)H MRS detection of histidine in human brain: Application to in vivo pH measurement. Magn Reson Med 2000;43(5):665–675. Rothman, D. L., Behar, K. L., Prichard, J. W., Petroff, O. A. Homocarnosine and the measurement of neuronal pH in patients with epilepsy. Magn Reson Med 1997;38(6): 924–929. Mori, S., Eleff, S. M., Pilatus, U., Mori, N., van Zijl, P. C. Proton NMR spectroscopy of solvent-saturable resonances: A new approach to study pH effects in situ. Magn Reson Med 1998;40(1):36–42. Frahm, J. Localized proton spectroscopy using stimulated echoes. J Magn Reson 1987;72(3):502–508. van Zijl, P. C., Moonen, C. T., Alger, J. R., Cohen, J. S., Chesnick, S. A. High field localized proton spectroscopy in small volumes: Greatly improved localization and shimming using shielded strong gradients. Magn Reson Med 1989;10(2):256–265. Duyn, J. H., Gillen, J., Sobering, G., van Zijl, P. C., Moonen, C. T. Multisection proton MR spectroscopic imaging of the brain. Radiology 1993;188(1):277–282. Ordidge, R. J. Random noise selective excitation pulses. Magn Reson Med 1987;5(1): 93–98. Moonen, C. T. W., Sobering, G., van Zijl, P. C. M., Gillen, J., von Kienlin, M., Bizzi, A. Proton spectroscopic imaging of human brain. J Magn Reson 1992;98(3):556–575.

224

Zhu and Barker

56. Moonen, C. T., von Kienlin, M., van Zijl, P. C., Cohen, J., Gillen, J., Daly, P., Wolf, G. Comparison of single-shot localization methods (STEAM and PRESS) for in vivo proton NMR spectroscopy. NMR Biomed 1989;2(5–6):201–208. 57. Tkac, I., Starcuk, Z., Choi, I. Y., Gruetter, R. In vivo 1H NMR spectroscopy of rat brain at 1 ms echo time. Magn Reson Med 1999;41(4):649–656. 58. Collins, C. M., Liu, W., Schreiber, W., Yang, Q. X., Smith, M. B. Central brightening due to constructive interference with, without, and despite dielectric resonance. J Magn Reson Imaging 2005;21(2):192–196. 59. Slotboom, J., Mehlkopf, A. F., Bovee, W. M. M. J. A single-shot localization pulse sequence suited for coils with inhomogeneous RF fields using adiabatic slice-selective RF pulses. J Magn Reson 1991;95:396–404. 60. Garwood, M., DelaBarre, L. The return of the frequency sweep: Designing adiabatic pulses for contemporary NMR. J Magn Reson 2001;153(2):155–177. 61. Scheenen, T. W., Heerschap, A., Klomp, D. W. Towards 1H-MRSI of the human brain at 7T with slice-selective adiabatic refocusing pulses. Magma 2008;21(1–2):95–101. 62. Scheenen, T. W., Klomp, D. W., Wijnen, J. P., Heerschap, A. Short echo time 1HMRSI of the human brain at 3T with minimal chemical shift displacement errors using adiabatic refocusing pulses. Magn Reson Med 2008;59(1):1–6. 63. Hwang, T. -L., Shaka, A. J. Water suppression that works. Excitation sculpting using arbitrary waveforms and pulsed field gradients. J Magn Reson Ser A 1995;112: 275–279. 64. Levitt, M. H., Freeman, R. Compensation for pulse imperfections in NMR spin echo experiments. J Magn Reson 1981;43:65–80. 65. Mekle, R., Mlynarik, V., Gambarota, G., Hergt, M., Krueger, G., Gruetter, R. MR spectroscopy of the human brain with enhanced signal intensity at ultrashort echo times on a clinical platform at 3T and 7T. Magn Reson Med 2009;61(6):1279–1285. 66. Ordidge, R. J., Connelly, A., Lohman, J. A. B. Image-selected in vivo spectroscopy (ISIS) – a new technique for spatially selective NMR-spectroscopy. J Magn Reson 1986;66(2):283–294. 67. Ernst, R. R., Bodenhausen, G., Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions. New York, NY: Oxford University Press; 1990, 640 p. 68. Golay, X., Gillen, J., van Zijl, P. C., Barker, P. B. Scan time reduction in proton

69.

70.

71.

72. 73.

74.

75.

76.

77.

78.

79.

80. 81.

magnetic resonance spectroscopic imaging of the human brain. Magn Reson Med 2002;47(2):384–387. Pohmann, R., von Kienlin, M., Haase, A. Theoretical evaluation and comparison of fast chemical shift imaging methods. J Magn Reson 1997;129(2):145–160. Duyn, J. H., Moonen, C. T. Fast proton spectroscopic imaging of human brain using multiple spin-echoes. Magn Reson Med 1993;30(4):409–414. Martin, A. J., Liu, H., Hall, W. A., Truwit, C. L. Preliminary assessment of turbo spectroscopic imaging for targeting in brain biopsy. AJNR Am J Neuroradiol 2001;22(5): 959–968. Mansfield, P. Spatial mapping of chemical shift in NMR. Magn Reson Med 1984;1: 370–386. Posse, S., Tedeschi, G., Risinger, R., Ogg, R., Le Bihan, D. High speed 1H spectroscopic imaging in human brain by echo planar spatial-spectral encoding. Magn Reson Med 1995;33(1):34–40. Posse, S., DeCarli, C., Le Bihan, D. Threedimensional echo-planar MR spectroscopic imaging at short echo times in the human brain. Radiology 1994;192(3):733–738. Ebel, A., Soher, B. J., Maudsley, A. A. Assessment of 3D proton MR echo-planar spectroscopic imaging using automated spectral analysis. Magn Reson Med 2001;46(6): 1072–1078. Ebel, A., Maudsley, A. A., Weiner, M. W., Schuff, N. Achieving sufficient spectral bandwidth for volumetric 1H echo-planar spectroscopic imaging at 4 Tesla. Magn Reson Med 2005;54(3):697–701. Pelletier, D., Nelson, S. J., Grenier, D., Lu, Y., Genain, C., Goodkin, D. E. 3-D echo planar (1)HMRS imaging in MS: Metabolite comparison from supratentorial vs. central brain. Magn Reson Imaging 2002;20(8):599–606. Govindaraju, V., Gauger, G. E., Manley, G. T., Ebel, A., Meeker, M., Maudsley, A. A. Volumetric proton spectroscopic imaging of mild traumatic brain injury. AJNR Am J Neuroradiol 2004;25(5):730–737. Adalsteinsson, E., Irarrazabal, P., Topp, S., Meyer, C., Macovski, A., Spielman, D. M. Volumetric spectroscopic imaging with spiral-based K-space trajectories. Magn Reson Med 1998;39(6):889–898. Block, K. T., Frahm, J. Spiral imaging: A critical appraisal. J Magn Reson Imaging 2005;21(6):657–668. Adalsteinsson, E., Star-Lack, J., Meyer, C. H., Spielman, D. M. Reduced spatial side

MR Spectroscopy and Spectroscopic Imaging of the Brain

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

93.

lobes in chemical-shift imaging. Magn Reson Med 1999;42(2):314–323. Sarkar, S., Heberlein, K., Hu, X. Truncation artifact reduction in spectroscopic imaging using a dual-density spiral K-space trajectory. Magn Reson Imaging 2002;20(10): 743–757. Kim, D. H., Adalsteinsson, E., Spielman, D. M. Spiral readout gradients for the reduction of motion artifacts in chemical shift imaging. Magn Reson Med 2004;51(3):458–463. Adalsteinsson, E., Langer-Gould, A., Homer, R. J., Rao, A., Sullivan, E. V., Lima, C. A., Pfefferbaum, A., Atlas, S. W. Gray matter N-acetyl aspartate deficits in secondary progressive but not relapsing-remitting multiple sclerosis. AJNR Am J Neuroradiol 2003;24(10):1941–1945. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962. Sodickson, D. K., Manning, W. J. Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38(4):591–603. Jakob, P. M., Griswold, M. A., Edelman, R. R., Sodickson, D. K. AUTO-SMASH: A selfcalibrating technique for SMASH imaging. Simultaneous acquisition of spatial harmonics. Magma 1998;7(1):42–54. McKenzie, C. A., Yeh, E. N., Ohliger, M. A., Price, M. D., Sodickson, D. K. Selfcalibrating parallel imaging with automatic coil sensitivity extraction. Magn Reson Med 2002;47(3):529–538. Griswold, M. A., Jakob, P. M., Heidemann, R. M., Nittka, M., Jellus, V., Wang, J., Kiefer, B., Haase, A. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47(6):1202–1210. Dydak, U., Weiger, M., Pruessmann, K. P., Meier, D., Boesiger, P. Sensitivity-encoded spectroscopic imaging. Magn Reson Med 2001;46(4):713–722. Tsai, S. Y., Otazo, R., Posse, S., Lin, Y. R., Chung, H. W., Wald, L. L., Wiggins, G. C., Lin, F. H. Accelerated proton echo planar spectroscopic imaging (PEPSI) using GRAPPA with a 32-channel phased-array coil. Magn Reson Med 2008;59(5):989–998. Dydak, U., Pruessmann, K. P., Weiger, M., Tsao, J., Meier, D., Boesiger, P. Parallel spectroscopic imaging with spin-echo trains. Magn Reson Med 2003;50(1):196–200. Lin, F. H., Tsai, S. Y., Otazo, R., Caprihan, A., Wald, L. L., Belliveau, J. W., Posse, S. Sensitivity-encoded (SENSE)

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

106.

225

proton echo-planar spectroscopic imaging (PEPSI) in the human brain. Magn Reson Med 2007;57(2):249–257. Smith, M. A., Gillen, J., McMahon, M. T., Barker, P. B., Golay, X. Simultaneous water and lipid suppression for in vivo brain spectroscopy in humans. Magn Reson Med 2005;54(3):691–696. Dreher, W., Leibfritz, D. Detection of homonuclear decoupled in vivo proton NMR spectra using constant time chemical shift encoding: CT-PRESS. Magn Reson Imaging 1999;17(1):141–150. Haase, A., Frahm, J., Hanicke, W., Matthei, D. 1 H NMR chemical shift selective imaging. Phys Med Biol 1985;30(4):341–344. Moonen, C. T. W., van Zijl, P. C. M. Highly efficient water suppression for in vivo proton NMR spectroscopy. J Magn Reson 1990;88:28–41. Ogg, R. J. WET a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR spectroscopy. J Magn Reson B 1994;104:1–10. Mescher, M., Merkle, H., Kirsch, J., Garwood, M., Gruetter, R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 1998;11(6):266–272. Star-Lack, J., Nelson, S. J., Kurhanewicz, J., Huang, L. R., Vigneron, D. B. Improved water and lipid suppression for 3D PRESS CSI using RF band selective inversion with gradient dephasing (BASING). Magn Reson Med 1997;38(2):311–321. Spielman, D. M., Pauly, J. M., Macovski, A., Glover, G. H., Enzmann, D. R. Lipidsuppressed single- and multisection proton spectroscopic imaging of the human brain. J Magn Reson Imaging 1992;2(3):253–262. Gu, M., Spielman, D. M. B1 and T1 insensitive water and lipid suppression using optimized multiple frequency-selective preparation pulses for whole-brain 1H spectroscopic imaging at 3T. Magn Reson Med 2009;61(2):462–466. Zhu, H., Ouwerkerk, R., Barker, P. B. Dual band water and lipid suppression for MR spectroscopic imaging at 3 Tesla. Magn Reson Med 2009;(accepted for publication). Kelley, D. A., Wald, L. L., Star-Lack, J. M. Lactate detection at 3T: Compensating J coupling effects with BASING. J Magn Reson Imaging 1999;9(5):732–737. Raphael, C. In vivo NMR spectral parameter estimation: A comparison between time and frequency domain methods. Magn Reson Med 1991;18:358–370. de Beer, R., van den Boogaart, A., van Ormondt, D., Pijnappel, W. W., den

226

Zhu and Barker

Hollander, J. A., Marien, A. J., Luyten, P. 108. Henriksen, O. In vivo quantitation of metabolite concentrations in the brain by R. Application of time-domain fitting in the means of proton MRS. NMR Biomed quantification of in vivo 1H spectroscopic 1995;8(4):139–148. imaging data sets. NMR Biomed 1992;5(4): 109. Soher, B. J., van Zijl, P. C., Duyn, 171–178. J. H., Barker, P. B. Quantitative pro107. Provencher, S. W. Estimation of metaboton MR spectroscopic imaging of the lite concentrations from localized in vivo human brain. Magn Reson Med 1996;35(3): proton NMR spectra. Magn Reson Med 356–363. 1993;30(6):672–679.

Chapter 10 Amide Proton Transfer Imaging of the Human Brain Jinyuan Zhou Abstract Amide proton transfer (APT) imaging is a new MRI technique that detects endogenous mobile proteins and peptides in tissue via saturation of the amide protons in the peptide bonds. Initial studies have shown promise in detecting tumor and stroke, but this technique was hampered by magnetic field inhomogeneity and a low signal-to-noise ratio. Several important prerequisites for performing APT imaging experiments include designing an effective APT imaging pulse sequence based on the hardware capability, optimizing the experimental protocol for the best clinical imaging quality, and developing data-processing approaches for effective image assessment. In this chapter, technical issues, such as pulse sequence design and optimization, magnetic field inhomogeneity correction, specific absorption rate minimization, and scan duration, are addressed. Key words: APT, magnetization transfer, protein, brain tumor, field inhomogeneity, MRI.

1. Introduction Proteins constitute 18% of the total mass of a typical mammalian cell. From an MRI point of view, these cellular proteins can be divided into two broad types: bound proteins, which possess solid-like properties and have protons with short T2 (∼ μs), and mobile proteins and peptides, which rotate rapidly and whose protons have relatively long T2 (∼ tens of ms). Solid-like macromolecules can be detected by conventional magnetization transfer (MT) (1, 2). It was recently demonstrated that it is possible to produce endogenous mobile protein- and peptide-based MRI contrast (3) using a chemical exchange saturation transfer (CEST) enhancement scheme (4). This approach, called amide proton transfer (APT) imaging (3), was shown to be sensitive to M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_10, © Springer Science+Business Media, LLC 2011

227

228

Zhou

pH changes in stroke (3, 5, 6) due to the effect of pH on proton exchange, and to be able to provide brain tumor contrast (7–9), based on the increased cellular content of proteins and peptides in malignant tumors (10, 11). APT imaging has the potential to expand the range of molecular MRI techniques to the endogenous protein and peptide level. It is a safe, non-invasive technology that can be easily implemented using existing hardware for clinical neuroimaging of brain tumors, stroke, and other neurologic disorders. The APT approach is in early development, and the technique is far from optimal. In this chapter, the initial experience and the prerequisites for performing an effective APT imaging experiment on the human brain at 3T are provided, using imaging of a human glioma as an example.

2. Materials 2.1. Hardware and Software

1. A Philips 3T MRI scanner (Philips Medical Systems, Best, The Netherlands). 2. An eight-channel phased-array head coil. 3. Two earplugs, a foam head holder, and a headband. 4. A personal computer loaded with the Philips pulseprogramming environment (PPE) and interactive data language (IDL, Research Systems, Inc., Boulder, CO, USA).

2.2. Phantoms, Human Subjects, and Other Materials

1. A bottle of liquid egg whites from the supermarket (see Note 1). 2. Several large bottles of water (3–5 L). 3. Healthy human subjects. 4. Patients with brain tumors. 5. Gadolinium contrast agents (ProHance, Bracco Diagnostic Inc., Princeton, NJ, USA).

3. Methods The effect of APT is associated with a low-concentration proton pool of mobile proteins and peptides. Therefore, the first step in conducting a successful APT imaging experiment is to design an effective imaging pulse sequence for the maximal APT enhancement. Because APT occurs in a small offset range around the water resonance frequency, a weak RF field should be applied to

Amide Proton Transfer Imaging of the Human Brain

229

avoid too much water signal intensity attenuation due to direct water saturation and conventional MT. After the pulse sequence is programmed, it must be tested and optimized for the best image and z-spectrum (see Note 2) quality. APT imaging is confounded by local magnetic field (B0 ) inhomogeneity (see Note 3). The initial human studies used a two-offset approach, symmetric around the global water center frequency, causing significant field-inhomogeneity-based image artifacts in many regions, especially in the frontal lobe and near the skull (8, 9). To solve this problem, a practical, six-offset, multi-acquisition method, combined with a single-acquisition zspectrum, may be used (see Fig. 10.1). As a compromise between optimum signal-to-noise ratio (SNR) and the ability for some correction, this method can acquire high-SNR APT images with B0 inhomogeneity correction within an acceptable scanning time (a few minutes).

Fig. 10.1. Schemes for APT-image acquisition. a Standard two-offset APT scan (+3.5 ppm for label, –3.5 ppm for reference). b Six-offset APT scan (±3, ±3.5, ±4 ppm). The effects of conventional MT and direct water saturation reduce the water signal intensities at all offsets (±3, ±3.5, ±4 ppm), and the existence of APT causes an extra reduction around the offset of 3.5 ppm (Reproduced from Ref. 9 with permission from Wiley-Liss, Inc.).

Finally, it is important to develop an APT data-processing approach for effective image assessment. This includes the following two steps: determination of the water-center-frequency shifts for each voxel and B0 inhomogeneity correction for z-spectra or APT images. The flow chart for APT-image data processing is shown in Fig. 10.2. 3.1. Design of an APT Imaging Pulse Sequence

1. These instructions assume the use of a Philips 3T MRI scanner, together with a body coil for RF transmission and an eight-channel phased-array coil for reception. Adjustments should be made according to the capability and limitations of the scanner hardware, particularly the duty cycle of radiofrequency (RF) power amplifiers (see Note 4). 2. Add a low-power long block pulse for proton saturation (up to 4 μT and 500 ms). The weak continuous-wave (CW) RF

230

Zhou

Fig. 10.2. Flow chart of APT-image data processing. The procedure is divided into two steps: (1) the generation of a water-frequency shift map and (2) the correction of APTimage data using the obtained shift map. Both steps are performed on a voxel-by-voxel basis.

saturation scheme has been widely used for various CEST imaging experiments. The RF saturation power and time are the most important pulse sequence parameters that must be optimized. 3. Use the turbo spin-echo (TSE) for imaging readout. A single slice is acquired (see Note 5). Single-shot acquisition should be used. Sensitivity encoding (SENSE) is used to reduce the TSE factor and specific absorption rate (SAR). 4. Add lipid suppression, such as selective partial inversion recovery (SPIR). 5. Adjust all pulse sequence parameters. This should also include the repetition time (TR), echo time (TE), field of view (FOV), image matrix, and slice thickness. The SAR should be kept below the U.S. Food and Drug Administration (FDA) limit for head (3.0 W/kg). 3.2. Test on Phantoms

1. Buy a bottle of egg whites from the supermarket. Shake the phantom for 5 min and put the phantom into the center of the magnet. To increase the loading, add several large bottles of water around the phantom. 2. Acquire localizer images and a SENSE reference scan. 3. Perform a z-spectrum experiment using an offset range of 8 to –8 ppm with an interval of 0.5 ppm (one signal average). One image without RF saturation is acquired for

Amide Proton Transfer Imaging of the Human Brain

231

normalization. To protect the scanner, conservative MRI parameters should be used for the first test: saturation time 100 ms, saturation power 1 μT, and TR 10 s. Higher order of (up to second order) volume shimming should be used. 4. Improve image quality and remove any image artifacts by modifying the imaging parameters, including the SENSE factor. 5. Repeat the z-spectrum experiments using the other pulse sequence parameters: saturation time 500 ms, saturation power 2–4 μT, and TR 3 s. 6. Examine the z-spectrum characteristics from a small region of interest (ROI). An ideal z-spectrum should be very smooth across all frequency offsets, with the lowest signal intensity at the water frequency, but the z-spectrum is generally shifted by the B0 inhomogeneity. 7. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for each z-spectrum experiment (see Section 3.6). 8. Plot the z-spectrum and MTRasym -spectrum (see Note 2) for a small ROI, and compare the curves at the different power levels. The maximal APT effect at 3.5 ppm is about 5–8%. 3.3. Optimization on Healthy Normal Subjects

1. Complete the initial screening and the consent form. 2. Put the subject into the center of the magnet. The head of the subject should be restrained to avoid motion artifacts. This is accomplished using a very comfortable foam head holder and a headband. 3. Acquire localizer images and a SENSE reference scan. 4. Perform a z-spectrum experiment using an offset range of 8 to –8 ppm with an interval of 0.5 ppm (saturation time 500 ms, power 1 μT, FOV 212×212 mm2 , matrix 128×64, slice thickness 5 mm, TR 3 s, TE 11 ms, one signal average). One image without RF saturation is acquired for normalization. Higher order of (up to second order) volume shimming should be used. 5. Improve image quality and remove any image artifacts by modifying the imaging parameters, including the SENSE factor. 6. Repeat the z-spectrum experiments with higher saturation power levels at 2–4 μT. 7. Move the subject out of the magnet. 8. Examine the z-spectrum characteristics from a small ROI. An ideal z-spectrum should be very smooth across all

232

Zhou

frequency offsets with the lowest signal intensity at the water frequency, but the z-spectrum may be shifted by the B0 inhomogeneity. 9. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for each z-spectrum experiment (see Section 3.6). 10. Plot the z-spectra and MTRasym -spectra (see Note 2) from gray matter, white matter, and cerebrospinal fluid (CSF). Compare the curves at the different power levels. Determine a characteristic RF saturation power at which MTRasym (3.5 ppm) is approximately zero for the whole brain (see Note 6). 3.4. Two- and Six-Offset Acquisition Scheme for APT Images

1. These instructions assume that the APT pulse sequence has been optimized with the z-spectrum experiments (see Section 3.3). The characteristic RF saturation power is 3 μT. The other imaging parameters are as follows: saturation time 500 ms, TR 3 s, FOV 212×212 mm2 , matrix 128×64, slice thickness 5 mm, TR 3 s, and TE 11 ms. 2. Acquire the saturation images at +3.5 ppm twice (eight signal averages). 3. Determine the signal intensity from a small ROI in one saturation image. Subtract the two saturation images and determine the noise from the same ROI in the difference image. Calculate the SNR. The SNR should be 100:1 or better to see the APT effect that is a few percent of the water intensity. 4. Perform an APT-image experiment using a standard twooffset acquisition scheme (+3.5 ppm for label and –3.5 ppm for reference) and eight signal averages. One image without RF saturation is acquired for normalization. 5. Calculate the APT image (see Note 2) and examine the effect of the B0 inhomogeneity on the APT images. 6. Acquire APT-image data using six frequency offsets (±3, ±3.5, ±4 ppm) and eight signal averages. One image without RF saturation is acquired for normalization. In this practical acquisition scheme, four extra offsets around ±3.5 ppm are acquired in the high SNR scan, and it is possible to correct for the artifacts on the APT image caused by B0 inhomogeneity. 7. Acquire a z-spectrum (33 offsets from 8 to –8 ppm with intervals of 0.5 ppm, one average) as an extra scan (see Note 7). One image without RF saturation is acquired for normalization.

Amide Proton Transfer Imaging of the Human Brain

233

8. Calculate the corrected APT image (see Section 3.6). The APT image is shown in color. For the healthy subject, a typical APT image is quite homogenous over the whole slice. 3.5. Scanning on Patients

1. The optimized z-spectrum and APT-image scans are added to the conventional MRI protocol. The total scan time for each subject should be limited to 1 h to maximize patient comfort. 2. Complete the initial screening and the consent form. 3. Put the subject into the center of the magnet. The head of the subject should be restrained to avoid motion artifacts. This is accomplished using standard configurations with a very comfortable foam head holder and a headband. 4. Acquire localizer images and a SENSE reference scan. 5. Acquire T2 -weighted images using the dual-echo TSE sequence. 6. Acquire fluid attenuated inversion recovery (FLAIR) images. 7. Identify the location of the lesion from the T2 -weighted and FLAIR images, and put the APT-image slice(s) on the lesion. 8. Acquire the six-offset APT-image scan. 9. Turn off the prescan to avoid changes in shim and frequency offset settings. 10. Acquire the z-spectrum scan. The same slice localization as that used for the APT-image scan should be used. 11. Acquire the T1 -weighted images using the magnetizationprepared rapid gradient-echo (MPRAGE) sequence. 12. Inject the gadolinium contrast agent into the patient (0.2 mL/kg, i.v.). 13. Acquire the gadolinium-enhanced T1 -weighted images using the MPRAGE sequence. 14. Move the subject out of the magnet. 15. Perform the B0 inhomogeneity correction on a voxel-byvoxel basis for each z-spectrum experiment (see Section 3.6). Plot the z-spectra and MTRasym -spectra (see Note 2) from the lesion and other ROIs. 16. Calculate the corrected APT image (see Section 3.6). The APT image is displayed in color. An example of the results produced is shown in Fig. 10.3.

3.6. Data Processing

1. These instructions assume the use of IDL. They are easily adaptable to other languages, such as MATLAB.

234

Zhou

Fig. 10.3. MR images of a patient with an anaplastic astrocytoma. Elevated APT signal can be seen in Gd-enhanced tumor core (red arrow), potentially providing unique information about the presence and grade of brain tumors, without the injection of exogenous contrast agents. The hyperintense APT area is comparable in size to the lesion identified on FLAIR but larger than that on the Gd-T1 w image (Reproduced from Ref. 9 with permission from Wiley-Liss, Inc.).

2. Fit the full z-spectrum through all 33 offsets using a 12thorder polynomial (the maximum order available with IDL) on a voxel-by-voxel basis (see Fig. 10.2). 3. Interpolate the fitted curve using an offset resolution of 1 Hz (2,049 points). 4. The actual water resonance should be at the frequency with the lowest signal intensity. The deviation of the water frequency in Hertz forms a map of water-center-frequency shifts. 5. To correct for the field inhomogeneity effects on zspectra, the measured z-spectrum for each voxel is interpolated to 2,049 points and shifted along the direction of the offset axis to correspond to 0 ppm at its lowest intensity. 6. The realigned z-spectra are interpolated back to 33 points for visual purposes. 7. Plot the z-spectrum and MTRasym -spectrum for a small ROI. The outermost points of ±7.5 and ±8 ppm are excluded in the display. 8. To correct for the field inhomogeneity effects on APT images (see Fig. 10.2), the acquired APT data for offsets +4, +3.5, and +3 ppm (or +512, +448, +384 Hz) for each voxel are interpolated to 257 points over the range from +4.5 to +2.5 ppm (or +576, +575, . . ., +320 Hz). 9. Realign the APT-image data using the fitted z-spectrum central frequency shift for the same voxel.

Amide Proton Transfer Imaging of the Human Brain

235

10. Perform the B0 inhomogeneity correction on a voxelby-voxel basis for the negative-offset data (–3, –3.5, –4 ppm). 11. Calculate the corrected APT image using the shiftcorrected data at the two offsets ±3.5 ppm. 12. The calculated APT image is thresholded based on the signal intensity of the S0 image to remove voxels outside the brain and displayed in color.

4. Notes 1. Many polymers, such as poly-L-lysine, dendrimers, and histone (12, 13), have a strong APT effect. However, these chemicals are very expensive; thus, they are not cost-effective for making large phantoms for use in human scanners. Egg whites in a bottle (protein 10%) from the supermarket are a straightforward protein phantom for APT imaging studies. In addition, some fruits, such as cantaloupe, are very useful phantoms with which to test imaging pulse sequences. Using a cantaloupe, up to about 25% of the CEST effect can be observed around a 1–2 ppm offset, which is due to various sugars in fruits. 2. In MT-type imaging (1, 2), water saturation is often measured as a function of transmitter frequency, producing the “z-spectra (14).” Such spectra are dominated by large direct water saturation around the water frequency at about 4.7 ppm and other saturation effects, such as conventional MT based on semi-solid tissue structures. The CEST effect is generally identified by asymmetry analysis with respect to this water signal (3), which is generally assigned to a reference frequency of 0 ppm. Quantitatively, the MT ratio (MTR) is defined as follows: MTR = 1 − Ssat /S0 , where Ssat and S0 are the signal intensities with and without RF irradiation, respectively. The MTR asymmetry (MTRasym ) parameter with respect to the water frequency is defined as follows: MTRasym = MTR( + offset) − MTR(−offset) = Ssat ( − offset)/S0 − Ssat ( + offset)/S0 .

236

Zhou

The APT image is quantified by the MTRasym parameter at ±3.5 ppm as follows: MTRasym (3.5 ppm) = Ssat ( − 3.5 ppm)/S0 − Ssat ( + 3.5ppm)/S0 

= MTRasym (3.5ppm) + APTR .

3. Under higher order of slice shimming, as seen in the shift in the center of the z-spectrum, the B0 inhomogeneity is typically less than 20 Hz over most of the slice, but it could be as large as 60–80 Hz in the sinus and ear areas. A small B0 inhomogeneity can easily cause a few percentage point change in the MT asymmetry data, thus resulting in large artifacts on APT images. 4. The magnitude of the APT effect increases exponentially with the RF saturation time, and several seconds of saturation time are required to maximize the measurements. The use of weak CW saturation pulses is feasible on animal MRI scanners. It may also be feasible on human MRI systems if a transmit/receive (T/R) head coil is used. When body coil excitation with a phased-array coil receive is used, the RF saturation pulse is restricted. In the Philips 3T MRI scanner with the tube amplifier, as used in this study, the saturation pulse duration is limited to 500 ms. 5. The clinical application of APT imaging to data remains limited to single-slice. Multi-slice or whole-brain imaging would be feasible, but many technical issues, such as long scan times, APT contrast loss between slices, magnetic field inhomogeneity, and SAR must be resolved properly. 6. As the applied RF saturation power increases, MTRasym (3.5 ppm) for gray matter and white matter increases initially and then decreases back to zero, while that for CSF remains at around zero. At the characteristic RF saturation power, MTRasym (3.5 ppm) is approximately zero for both normal brain tissue and CSF. In contrast, MTRasym (3.5 ppm) would be hyperintensive in tumors and hypointensive in stroke regions. 7. If the z-spectrum measurement has the same saturation parameters as the APT-image scan, the z-spectrum data can be used for both identification of the APT effect and fitting of the B0 inhomogeneity map. If the z-spectrum is used to fit only the B0 inhomogeneity map, then lower saturation power and shorter saturation time may be used. Such a scan, called water saturation shift referencing (WASSR) (15), would provide a narrower z-spectrum; thus, the B0 inhomogeneity map can be fitted more accurately.

Amide Proton Transfer Imaging of the Human Brain

237

Acknowledgments The author thanks the team at the F.M. Kirby Research Center for Functional Brain Imaging for helpful discussions and technical assistance. This work was supported in part by grants from NIH (EB002634, EB005252, EB009112, EB009731, and RR015241) and the Dana Foundation. References 1. Wolff, S. D., Balaban, R. S. Magnetization transfer contrast (MTC) and tissue water proton relaxation in vivo. Magn Reson Med 1989;10:135–144. 2. Henkelman, R. M., Stanisz, G. J., Graham, S. J. Magnetization transfer in MRI: A review. NMR Biomed 2001;14:57–64. 3. Zhou, J., Payen, J., Wilson, D. A., Traystman, R. J., van Zijl, P. C. M. Using the amide proton signals of intracellular proteins and peptides to detect pH effects in MRI. Nat Med 2003;9:1085–1090. 4. Ward, K. M., Aletras, A. H., Balaban, R. S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST). J Magn Reson 2000;143:79–87. 5. Sun, P. Z., Zhou, J., Sun, W., Huang, J., van Zijl, P. C. M. Delineating the boundary between the ischemic penumbra and regions of oligaemia using pHweighted magnetic resonance imaging (pHWI). J Cereb Blood Flow Metab 2007;27: 1129–1136. 6. Jokivarsi, K. T., Grohn, H. I., Grohn, O. H., Kauppinen, R. A. Proton transfer ratio, lactate, and intracellular pH in acute cerebral ischemia. Magn Reson Med 2007;57:647–653. 7. Zhou, J., Lal, B., Wilson, D. A., Laterra, J., van Zijl, P. C. M. Amide proton transfer (APT) contrast for imaging of brain tumors. Magn Reson Med 2003;50: 1120–1126. 8. Jones, C. K., Schlosser, M. J., van Zijl, P. C., Pomper, M. G., Golay, X., Zhou, J. Amide proton transfer imaging of human brain tumors at 3T. Magn Reson Med 2006;56:585–592.

9. Zhou, J., Blakeley, J. O., Hua, J. et al. Practical data acquisition method for human brain tumor amide proton transfer (APT) imaging. Magn Reson Med 2008;60: 842–849. 10. Hobbs, S. K., Shi, G., Homer, R., Harsh, G., Altlas, S. W., Bednarski, M. D. Magnetic resonance imaging-guided proteomics of human glioblastoma multiforme. J Magn Reson Imaging 2003;18:530–536. 11. Howe, F. A., Barton, S. J., Cudlip, S. A. et al. Metabolic profiles of human brain tumors using quantitative in vivo 1 H magnetic resonance spectroscopy. Magn Reson Med 2003;49:223–232. 12. Goffeney, N., Bulte, J. W. M., Duyn, J., Bryant, L. H., van Zijl, P. C. M. Sensitive NMR detection of cationic-polymer-based gene delivery systems using saturation transfer via proton exchange. J Am Chem Soc 2001;123:8628–8629. 13. McMahon, M. T., Gilad, A. A., Zhou, J., Sun, P. Z., Bulte, J. W. M., van Zijl, P. C. M. Quantifying exchange rates in chemical exchange saturation transfer agents using the saturation time and saturation power dependencies of the magnetization transfer effect on the magnetic resonance imaging signal (QUEST and QUESP): pH calibration for poly-L-lysine and a starburst dendrimer. Magn Reson Med 2006;55:836–847. 14. Bryant, R. G. The dynamics of water-protein interactions. Annu Rev Biophys Biomol Struct 1996;25:29–53. 15. Kim, M., Gillen, J., Landman, B. A., Zhou, J., van Zijl, P. C. M. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Magn Reson Med 2009;61:1441–1450.

wwwwwww

Chapter 11 High-Field MRI of Brain Iron Jozef H. Duyn Abstract Recent developments in high-field MRI have provided opportunities to detect iron in human brain with much improved sensitivity. The combination of increased magnetic field strength with multi-channel detectors has made it possible to routinely obtain images at about 300 μm resolution. These images can be sensitized to tissue iron by exploiting the improved magnetic susceptibility contrast at high field. Together, these techniques have the potential to map the fine scale distribution of iron in human brain at the level of fiber bundles and cortical laminae, and may aid in the understanding of the role and transport of iron in normal brain and in disease. In this chapter, we will look at these techniques in detail and present some examples of high-field MRI data of human brain. Key words: Iron, ferritin, MRI, brain, magnetic susceptibility.

1. Introduction Cellular iron has important roles in brain development and function, and abnormal concentrations may lead to pathological conditions (1–3). Since the beginning of MRI, attempts have been made to map in vivo brain iron distributions under normal and pathological conditions (4–6). MRI is a versatile technique that is able to generate a variety of contrasts, a number of which reflect tissue iron content. For example, iron can affect the MRI signal through several of its major contrast parameters including T1 , T2 , and T2 ∗ (6). All of these have been used to study brain iron, and each has its advantages and disadvantages. The recent proliferation of high-field scanners of 7 T and above has reinvigorated the study of brain iron with MRI. Major factors in this have been the increased sensitivity (signal-to-noise M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_11, © Springer Science+Business Media, LLC 2011

239

240

Duyn

ratio or SNR) available with the new scanners, in particular when employing magnetic susceptibility contrast. This increased sensitivity is due to the paramagnetic properties of iron in most of its forms that are present in brain. Magnetic susceptibility contrast is reflected in T2 ∗ and resonance frequency, both of which may convey information about tissue iron content (6–8). In the following, we will discuss some of the methodological aspects of high-field MRI sensitized to brain iron, including the equipment and acquisition techniques involved, as well as the analysis and interpretation of the data.

2. Materials 2.1. MRI Hardware 2.1.1. Magnets

Advances in magnet technology have led to the availability of stronger magnets and higher sensitivity for human MRI. While early clinical MRI systems employed magnets of 1 T field strength or below, today’s scanners are at 1.5 T and 3.0 T, and at even 7.0 T. Field strengths continue to increase, and currently experimental 9.4 T and 11.7 T systems are in use or being developed. The increased sensitivity of these systems can be exploited to improve spatial resolution. In addition, magnetic susceptibility contrast is increased, providing particular advantages for the detection of iron. These increases are dependent on the echo time (see Methods) and for typical conditions are around threefold for both T2 ∗ and resonance frequency contrast (Fig. 11.1).

2.1.2. RF Detectors

An additional increase in sensitivity has come from the development of multi-channel detectors that has played out over the last decade. Multi-channel detectors improve signal detection by allowing the individual detector elements to be placed closer to the object under study and, at the same time, limit the noise received from the sample. Together, these advantages result in sensitivity improvements that can average around two- to threefold over the brain for 32 channel detectors that are currently widely available. A second advantage of multi-channel detectors is the fact that they enable image acceleration through parallel imaging techniques (9, 10). This is a particularly useful feature to achieve high resolution, which requires more data to be acquired and therefore longer scan times.

2.1.3. Respiratory Compensation

The SNR and CNR increases at high field are accompanied by an increased sensitivity to physiological variations and motion that can compromise image quality. In order to fully exploit the

High-Field MRI of Brain Iron

241

Fig. 11.1. Simulated contrast-to-noise ratio (CNR) for susceptibility weighted MRI at 3 T and 7 T. The CNR gain at high field is dependent of echo time and ranges from 2 to 4 for echo time in the commonly used range of 10–60 ms. The simulation ignored T1 effects and assumed identical acquisition bandwidth, a linear sensitivity increase with field strength, and tissue R2 ∗ values of 20–1 and 30 s–1 at 3 T and 7 T, respectively.

potential of high-field MRI, these unwanted confounds need to be dealt with. For example, the respiratory cycle can induce subtle magnetic field fluctuations in the brain that lead to ghosting artifacts in susceptibility weighted MRI (11). This is particularly noticeable at high field as the amplitude of these fluctuations increases linearly with field strength. Currently, prototype hardware exists to compensate for these field fluctuations and improve image quality (Fig. 11.2). Major components are a pressure belt to register chest motion, a computer to calculate field (shim)

Fig. 11.2. Compensation of respiration-induced magnetic field fluctuations. The patient’s chest position, registered with a stretch sensor placed around the chest, is used to estimate adjustments to RF frequency, magnetic field gradients, and magnet field shims. This is done with a personal computer, which sends the adjustment values to the MRI system electronics.

242

Duyn

corrections, and shim coils driven with fast current sources to allow rapid field adjustments. 2.1.4. Head Motion Correction

Sudden or even slow head motion during MRI scanning can compromise image quality, in particular when scanning at high resolution. A number of techniques have been proposed to compensate for this motion (12–15). Recent implementations employ video cameras to track head motion and feed this information back to the scanner to make adjustments to the image acquisition process (14, 15). An alternative approach measures the local, position-dependent magnetic field induced by the MRI gradient system through the use of small coils placed around the head (16). Although these techniques are quite effective, they have not been yet fully developed for widespread use.

3. Methods 3.1. Acquisition Methods

Magnetic susceptibility inclusions in brain tissues, for example local areas of increased iron content, lead to magnetic field shifts that are generally inhomogeneous over the scale of an image voxel. This results in incoherent phase accumulation and therefore to T2 ∗ reduction and signal loss in gradient echo imaging (GRE). In addition, these inclusions may lead to a net frequency shift of the voxel-averaged signal, which manifests itself as a voxel phase shift. This effect has been recently exploited at 7 T to improve visualization of subtle anatomical variations in grey and white matter of human brain. (8)

3.1.1. Resolution

The choice of image resolution is important as it directly affects scan times and image quality. Too low a resolution may lead to partial volume effects and affect the conspicuity of small anatomical variations. Too high a resolution may increase the sensitivity to patient motion and lead to low signal-to-noise ratio (SNR). This SNR can only partially be recovered with spatial averaging during image reconstruction. Using 32 channel detectors at 7 T, image resolutions of 0.2 × 0.2 × 1 mm or 0.3 × 0.3 × 0.3 mm are feasible within scan times of about 10 min.

3.1.2. Echo Time and Bandwidth

In choosing optimal echo time (TE) and bandwidth, one needs to take into account the contrast-to-noise ratio (CNR). While increasing echo time (TE) leads to increased percentage signal loss and absolute phase shift, it also reduces image SNR. It turns out that one can optimize the CNR of both signal amplitude and phase by choosing TE equal to the average T2 ∗ values of the brain structures involved (8). At 7 T, this means TE needs to be chosen

High-Field MRI of Brain Iron

243

in the range of 10–30 ms, which covers much of the range of T2 ∗ values found in (normal) human brain. Furthermore, CNR optimization requires minimization of the acquisition bandwidth, i.e., capture as much of the signal decay curve by maximizing the duration of the data acquisition window (TACQ) (Fig. 11.3). One caveat with this is that increasing TACQ increases image blurring and off-resonance related pixel shifts, some of which can be corrected in post-processing. Typical TACQ values at 7 T range from 10 to 30 ms.

Fig. 11.3. Optimization of sensitivity (SNR). Maximum SNR is obtained when the acquisition duration (TACQ) is maximized. As TACQ is generally centered around the gradient echo time (TE), TACQ < 2∗ TE.

3.1.3. Multi-slice 2D Versus 3D Techniques

GRE MRI can be performed either in a multi-slice or true 3D fashion. The latter may have a significant SNR advantage if a large area of interest (or the entire brain) needs to be imaged. 3D techniques excite the entire slab of interest with each RF pulse rather than sub-sections in a sequential fashion. The SNR advantage comes about when the time to run through all the sub-sections in a multi-slice scan exceeds the longitudinal relaxation time of the tissue. Under this condition, multi-slice techniques become rather inefficient (in term of SNR per square root of total scan time) compared to true 3D techniques. A caveat with 3D techniques is that the generated MRI signal may have a larger dynamic range and therefore put increased demands on the MRI acquisition hardware. Further, 3D techniques may be less robust in the presence of head motion.

3.1.4. Image Acceleration with Parallel Imaging

Higher resolution MRI requires the acquisition of large data matrices, leading to long scan times. A typical 20-slice highresolution 2D acquisition with image matrix of 1,024 × 768 requires 15,360 repeated RF excitations. With T2 ∗ of up to 30 ms, most of the signal decay curve can be sampled in about 50 ms. Assuming each repeated excitation to last 50 ms, the

244

Duyn

scan time for the entire acquisition would be close to 13 min. Increasing volume coverage beyond the 20 slice example given above may lead to prohibitively long scan times, necessitating the use of image acceleration approaches such as parallel imaging. Methods, such as SENSE (10) and SMASH (17), allow some of the acquisition matrix element to be estimated from the spatial information contained in the sensitivity profiles of the detector elements. The saving in acquisition time resulting from this is ultimately limited by the number of detector elements. With 32channel detectors and acceleration in one dimension, acceleration rates of 3–4 (i.e., scan time reduction of 66–75%) are feasible without significant degradation of image quality. 3.1.5. Multi-echo Techniques

The image intensity in GRE MRI data is dependent not only on tissue T2 ∗ values but also on other MR parameters, such as spin density and T1 , and sequence parameters, such as TR and flip angle. This complicates extraction of quantitative and reproducible measures. To overcome this, one can acquire multiple sequential echoes, each of which can be reconstructed into a separate image. The varying T2 ∗ weighting of each image can be used to extract quantitative T2 ∗ values. The generation of multiple echoes can be effectuated by repeated reversal of the read gradient (Fig. 11.4).

Fig. 11.4. Multi-echo acquisition. Multiple reversals of the read gradient are used to generate a number of echo signals with increasing T2 ∗ -weighting.

The added benefit of the multi-echo approach is that each echo is acquired in a shorter time (higher bandwidth) and therefore is less affected by off-resonance artifacts and T2 ∗ blurring (18). Furthermore, there is no significant SNR penalty in doing this, as the multiple echo data can be recombined into a single image with SNR similar to that of the low bandwidth T2 ∗ weighted image of the conventional single echo approach.

High-Field MRI of Brain Iron

245

Drawbacks are the increased stress on the gradient system, the increased acoustic noise, and the increased sensitivity to motion. The latter originates from the fact that it becomes more difficult (and less efficient) to employ motion compensation strategies that rely on gradient moment nulling. (19) 3.2. Reconstruction Methods

A number of processing steps are required to convert the raw data into interpretable images that can provide a measure of iron content. These include combining of the coil signals with or without parallel imaging reconstruction and calculation of frequency (or phase) maps. Additional steps can include the calculation of magnetic susceptibility and T2 ∗ maps. We will discuss each of these briefly.

3.2.1. Coil Combining

Combination of coil signals can be performed with the standard SENSE reconstruction, as described previously (20). This is appropriate for both standard non-accelerated acquisitions (R=1) and accelerated acquisitions (R>1). It may be beneficial to use subject-specific coil sensitivity data, which can be used to generate the required coil sensitivity reference maps (20). For this purpose, a fast, low resolution scan can be performed using the same slice locations as the high-resolution data. Preferably, a scan with minimal T2 ∗ contrast is used. This can be achieved by using short TE.

3.2.2. Calculation of Frequency Maps

The complex image data generated with the SENSE reconstruction can be converted into both magnitude (i.e., signal amplitude) and phase maps, both of which are sensitive to tissue iron content. The phase maps are then further processed to remove unwanted spatial variations associated with large-scale bulk susceptibility effects at, e.g., air-tissue interfaces. This can be effectively achieved with spatial high-pass filtering through homodyne methods or polynomial fitting (8, 21, 22). These methods also allow convenient removal of any phase jumps (at boundaries of the [–π, π] phase range) that may be present in the raw data. The remaining signal phase can be attributed to off-resonance effects that reflect the underlying tissue properties, including the local iron content. The amplitude of this effect (in Hertz) can be calculated by dividing the local phase shift (in cycles) by the echo time (in seconds).

3.2.3. Susceptibility Maps

Although phase/frequency images have been used to directly estimate local iron content (7), one caveat is that the two are only indirectly related. One important confound is that local resonance frequency is dependent in a complicated manner on geometry and orientation of both local and surrounding distribution of iron inclusions (6, 23, 24). A number of research groups are

246

Duyn

addressing this problem and are attempting to reconstruct susceptibility maps from 3D phase distributions (25–28). The former would not be dependent on orientation and geometry and more directly represent the local tissue composition. Preliminary experience in brain suggests that susceptibility calculation is indeed possible (28); however, there are still unresolved issues that impact the quality of the susceptibility maps. These issues include the presence of streaking artifacts due to focal areas of ill-defined phase (e.g,. in vessels or near air-tissue interfaces) and noise amplification for structures that are at the magic angle relative to the main magnetic field (28). It is anticipated that these issues will be resolved, at least partly, in coming years. The data available with multi-echo techniques allow calculation of quantitative T2 ∗ values (or R2 ∗ values; R2 ∗ = 1/T2 ∗ ), which may supplement susceptibility information for the study of brain iron content. T2 ∗ values can be derived for multi-echo data by simple T2 ∗ fitting of the signal decay with increasing echo time. When both positive and negative echoes in a GRE echo train are used, correction of off-resonance related distortions may be required prior to fitting. This can be done based on Bo maps that can be derived from the phase data. Sample R2 ∗ and T2 ∗ maps are shown in Fig. 11.5.

Fig. 11.5. Example of various contrasts that can be derived from a multi-echo data acquisition. Shown are magnitude signal of the first echo (primarily proton-density weighted) (a), frequency (b), T2 ∗ (c), and R2 ∗ (d).

High-Field MRI of Brain Iron

3.3. Data Interpretation

247

Although it has been over two decades since the first MRI study of brain iron distribution, the development of a quantitative method to estimate brain iron content from MRI-derived measures is still a work in progress. The primary reason for this is that MRI contrast mechanisms are generally complex, and this is certainly the case for the phase shifts and T2 ∗ values derived from GRE data at high field. For example, the T2 ∗ relaxation caused by intra-voxel phase dispersion may originate from a number of sources in addition to iron, including exchange effects with amide protons (29), and inhomogeneous fields generated by susceptibility inclusions, such as proteins, myelin, and deoxyhemoglobin (8, 30, 31). These effects may have a geometry and an orientation dependence. This is also the case for MRI frequency maps, which are affected by many of the same contributors. Nevertheless, in regions where iron dominates the contrast, T2 ∗ and susceptibility values may have a relatively well-defined dependence on iron content. In these regions, these measures may provide reasonable estimates of tissue iron content (7, 22). However, it is expected that the relative contribution of the sources contributing to contrast in susceptibility weighted MRI will vary across brain regions. For example, in white matter, some of the paramagnetic susceptibility of iron may be cancelled out by diamagnetic contributions of myelin. This could explain the absence of a paramagnetic phase shift in WM (relative to cerebrospinal fluid) (8) and a remaining diamagnetic shift after iron extraction (32). Also, because of the generally highly ordered microscopic structure of WM, the orientation of this structure may affect the MRI susceptibility measures (30, 33). Accurate analysis of brain iron content with MRI will likely require a comprehensive understanding of the mechanisms and relative importance of contributing compounds to the various MRI contrast parameters. It is likely that a combined analysis of the various MRI contrasts will contribute to this understanding. For example, a combined analysis of T2 ∗ and phase data may be helpful in quantification of myelin and iron content, as these compounds differentially contribute to the two contrasts.

Acknowledgments My colleagues in the laboratory of Advanced MRI are acknowledged for their contributions to this work. This research was supported by the Intramural Research Program of NIH, NINDS.

248

Duyn

References 1. Morris, C. M. et al. Brain iron homeostasis (Translated from Eng). J Inorg Biochem 1992;47(3–4):257–265 (in Eng). 2. Burdo, J. R., Connor, J. R. Brain iron uptake and homeostatic mechanisms: An overview (Translated from Eng). Biometals 2003;16(1):63–75 (in Eng). 3. Zecca, L., Youdim, M. B., Riederer, P., Connor, J. R., Crichton, R. R. Iron, brain ageing and neurodegenerative disorders (Translated from Eng). Nat Rev Neurosci 2004;5(11):863–873 (in Eng). 4. Drayer, B. et al. MRI of brain iron. Am J Roentgenol 1986;147(1):103–110. 5. Schenck, J. F. Magnetic resonance imaging of brain iron. J Neurol Sci 2003;207(1–2): 99–102. 6. Haacke, E. M. et al. Imaging iron stores in the brain using magnetic resonance imaging. Magn Reson Imaging 2005;23(1):1–25. 7. Ogg, R. J., Langston, J. W., Haacke, E. M., Steen, R. G., Taylor, J. S. The correlation between phase shifts in gradient-echo MR images and regional brain iron concentration. Magn Reson Imaging 1999;17(8): 1141–1148. 8. Duyn, J. H. et al. High-field MRI of brain cortical substructure based on signal phase. Proc Natl Acad Sci USA 2007;104(28):11796–11801. 9. Sodickson, D. K., Manning, W. J. Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38(4):591–603. 10. Pruessmann, K. P., Weiger, M., Scheidegger, M. B., Boesiger, P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 1999;42(5):952–962. 11. van Gelderen, P., de Zwart, J. A., Starewicz, P., Hinks, R. S., Duyn, J. H. Real-time shimming to compensate for respirationinduced B0 fluctuations. Magn Reson Med 2007;57(2):362–368. 12. Derbyshire, J. A., Wright, G. A., Henkelman, R. M., Hinks, R. S. Dynamic scanplane tracking using MR position monitoring (Translated from Eng). J Magn Reson Imaging 1998;8(4):924–932 (in Eng). 13. Tremblay, M., Tam, F., Graham, S. J. Retrospective coregistration of functional magnetic resonance imaging data using external monitoring. Magn Reson Med 2005;53(1): 141–149. 14. Zaitsev, M., Dold, C., Sakas, G., Hennig, J., Speck, O. Magnetic resonance imaging of freely moving objects: Prospective

15.

16.

17. 18.

19.

20.

21.

22.

23.

24.

25.

real-time motion correction using an external optical motion tracking system (Translated from Eng). Neuroimage 2006;31(3): 1038–1050 (in Eng). Qin, L. et al. Prospective head-movement correction for high-resolution MRI using an in-bore optical tracking system (Translated from Eng). Magn Reson Med 2009;62: 924–934 (in Eng). Ooi, M. B., Krueger, S., Thomas, W. J., Swaminathan, S. V., Brown, T. R. Prospective real-time correction for arbitrary head motion using active markers (Translated from Eng). Magn Reson Med 2009;62(4): 943–954 (in Eng). Sodickson, D. K., Griswold, M. A., Jakob, P. M. SMASH imaging. Magn Reson Imaging Clin N Am 1999;7(2):237–254, vii–viii. Fischl, B. et al. Sequence-independent segmentation of magnetic resonance images (Translated from Eng). Neuroimage 2004;23(Suppl 1):S69–S84 (in Eng). Haacke, E. M., Lenz, G. W. Improving MR image quality in the presence of motion by using rephasing gradients (Translated from Eng). AJR Am J Roentgenol 1987;148(6):1251–1258 (in Eng). de Zwart, J. A., Ledden, P. J., Kellman, P., van Gelderen, P., Duyn, J. H. Design of a SENSE-optimized high-sensitivity MRI receive coil for brain imaging. Magn Reson Med 2002;47(6):1218–1227. Abduljalil, A. M., Schmalbrock, P., Novak, V., Chakeres, D. W. Enhanced gray and white matter contrast of phase susceptibilityweighted images in ultra-high-field magnetic resonance imaging. J Magn Reson Imaging 2003;18(3):284–290. Yao, B. et al. Susceptibility contrast in high field MRI of human brain as a function of tissue iron content (Translated from Eng). Neuroimage 2009;44(4):1259–1266 (in Eng). Salomir, R., de Senneville, B. D., Moonen, C. T. W. A fast calculation method for magnetic field inhomogeneity Due To an arbitrary distribution of bulk susceptibility. Concepts Magn Reson B 2003;19B(1):26–34. Marques, J. P., Bowtell, R. W. Using forward calculations of the magnetic field perturbation Due To a realistic vascular model to explore the BOLD effect. NMR Biomed 2008;21(6):553–565. de Rochefort, L., Brown, R., Prince, M. R., Wang, Y. Quantitative MR susceptibility mapping using piece-wise constant regularized inversion of the magnetic field

High-Field MRI of Brain Iron

26.

27.

28.

29.

(Translated from Eng). Magn Reson Med 2008;60(4):1003–1009 (in Eng). Kressler, B. et al. Nonlinear regularization for per voxel estimation of magnetic susceptibility distributions from MRI field maps (Translated from Eng). IEEE Trans Med Imaging 2010;29(2):273–281 (in Eng). Liu, T., Spincemaille, P., de Rochefort, L., Kressler, B., Wang, Y. Calculation of susceptibility through multiple orientation sampling (COSMOS): A method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI (Translated from Eng). Magn Reson Med 2009;61(1):196–204 (in Eng). Shmueli, K., Li, J., Duyn, J. H. Magnetic susceptibility mapping of brain tissue in-vivo using MRI phase data (Translated from Eng). Magn Reson Med 2009;62(6):1510–1522 (in Eng). Zhong, K., Leupold, J., von Elverfeldt, D., Speck, O. The molecular basis for gray and

30.

31.

32.

33.

249

white matter contrast in phase imaging. Neuroimage 2008;40(4):1561–1566. He, X., Yablonskiy, D. A. Biophysical mechanisms of phase contrast in gradient echo MRI (Translated from Eng). Proc Natl Acad Sci USA 2009;106(32):13558–13563 (in Eng). Lee, D., Hirano, Y., Fukunaga, M., Silva, A. C., Duyn, J. H. On the contribution of deoxy-hemoglobin to MRI gray-white matter contrast at high field. Neuroimage 2010;49(1):193–198. Fukunaga, M. et al. Layer-specific variation of iron content in cerebral cortex as a source of MRI contrast (Translated from Eng). Proc Natl Acad Sci USA 2009;107(8):3834–3839 (in Eng). Lee, J. et al. Sensitivity of MRI resonance frequency to the orientation of brain tissue microstructure (Translated from Eng). Proc Natl Acad Sci USA 2010;107(11):5130– 5135 (in Eng).

wwwwwww

Chapter 12 Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications Manisha Aggarwal, Jiangyang Zhang, and Susumu Mori Abstract In this chapter, we introduce modern magnetic resonance imaging (MRI)-based mouse brain atlases. Although unable to match the resolution and specificity of their histology-based counterparts, MRI-based mouse brain atlases feature higher anatomical fidelity and can facilitate high-throughput computerassisted analysis of certain brain phenotypes. This chapter discusses several technical aspects of MRI-based mouse brain atlases, which are important to understand the usefulness as well as limitations of existing atlases. We focus on a novel MRI technique, diffusion tensor imaging (DTI), which provides rich tissue contrasts and is uniquely suited for studying white matter structures and immature mouse brains. The chapter then demonstrates several applications of MRI-based mouse brain atlases in anatomical phenotyping and guiding stereotaxic operations. Key words: Mouse brain atlas, magnetic resonance imaging, brain morphology, diffusion tensor imaging.

1. Introduction The laboratory mouse is widely used in neuroscience research, because it shares many genes, physiological processes, and disease loci with humans and is relatively easy to handle. With modern gene technology and the availability of the mouse genome database, it is relatively easy to generate genetically modified mouse strains in order to investigate the mechanisms of genetic controls of the brain. Numerous mouse models of human diseases have been established in the last few decades, and they have played essential roles in advancing our knowledge on the mechanisms M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_12, © Springer Science+Business Media, LLC 2011

251

252

Aggarwal, Zhang, and Mori

of diseases in the brain and developing new therapeutic treatments. Atlases of the mouse brain are important tools in examining changes in the mouse brain due to genetic mutations or pathological conditions. A brain atlas presents neuroanatomical information in the forms of anatomical images and structural delineations, from which the morphological properties of brain structures and their spatial relationships can be appreciated. The information can be used to guide surgical operations and target-specific delivery of drugs or cells. A brain atlas can serve as the central platform for data analysis and reporting. For instance, data on cellular and molecular events in various parts of the brain can be conveniently incorporated in a brain atlas to study their distributions and structural differences using advanced bioinformatics tools (1, 2). There are several well-established mouse brain atlases based on histology (3–11), which provides rich cellular and molecular information at high resolution. The staining methods used in these atlases reveal the biochemical composition and microstructural information of brain structures. The contrasts presented in the stained histological sections are used to delineate detailed structures in the mouse brain, e.g., the thalamic nuclei, and have been widely regarded as the gold standard for distinguishing brain structures. One limitation of histology based brain atlases is that they have limited anatomical fidelity. Histology is performed on ex vivo brain specimens. The sectioning and embedding processes can cause significant tissue injury and deformation, which can be more severe for embryonic or neonatal brain samples, since the immature brain tissue is relatively soft and easily deformed compared to the adult brain. Furthermore, the sectioning process can disrupt the spatial relationships between structures, which are difficult to recover even with complex reconstruction algorithms. Lack of anatomical fidelity limits the application of histology based atlases. For example, the precision of stereotaxic operations strongly depends on whether the atlas in use to guide such operations is an accurate representation of the target brain anatomy. This point will be discussed in detail in a later section. One solution to ensure high anatomical fidelity is to construct mouse brain atlases based on images acquired using three dimensional (3D) non-destructive imaging techniques, which do not require sectioning to acquire a stack of cross-sectional images, and therefore preserve high anatomical fidelity. 3D imaging techniques also have several unique advantages. Images can often be acquired from live animals, and therefore the anatomical information presented in such images is the least perturbed by the imaging procedure. The results of 3D imaging techniques can be viewed in any required oblique orientation and can be processed and analyzed directly without complex 3D reconstruction procedures. This property potentially allows direct quantitative

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

253

analyses of anatomical variability in the brain and facilitates the generation of population-averaged atlases. There are several 3D imaging techniques currently available. Micro-computed tomography (CT) can generate moderate to high-resolution images (1–100 μm) depending on the instrument and available X-ray source. It can be used to study bone and skull features, and, when used together with certain contrast agents, can reveal vascular information in the mouse brain and other organs (12, 13). However, micro-CT often fails to generate satisfactory contrast in the brain for structural delineation. Optical projection tomography (OPT) (14) is a recently developed technique, which provides high spatial resolution (∼5 μm) and imaging speed. However, the image contrast in the brain provided by OPT is limited because it is difficult to stain intact specimens. Compared to micro-CT and OPT, magnetic resonance imaging (MRI) provides far richer soft tissue contrast in the brain with modest spatial resolution (up to 10–20 μm). There are several unique tissue MRI contrasts that can be used to create an MRI-based atlas, including T1 , T2 , T2 ∗ , diffusion, and magnetization transfer. These MRI contrasts reflect the physical and chemical microenvironment of tissue water molecules, for example, water and myelin contents. Figure 12.1 shows examples of in vivo and ex vivo mouse brain MR images acquired with different contrast techniques. These MRI contrasts have been widely used to study normal brain anatomy and physiology, as well as various pathological conditions in the brain, in diseases, such as multiple sclerosis (15–17) and stroke (18, 19). Although the sensitivity and specificity of MRI contrasts cannot compete with the contrasts provided by histology, the ability to monitor anatomical and physiological changes

Fig. 12.1. Appearances of live and postmortem mouse brains with different MR contrasts. The in vivo images were acquired with a resolution of 0.1 × 0.1 × 0.4 mm3 . The ex vivo images were acquired with a resolution of 0.125 × 0.125 × 0.125 mm3 . DW, diffusion weighted; FA, fractional anisotropy; MT, magnetization transfer. The FA images represent one of the different contrasts derived from diffusion tensor imaging. The in vivo MT image is not coregistered with other in vivo images.

254

Aggarwal, Zhang, and Mori

in vivo makes MRI the ideal imaging technique for many studies, and therefore it is imperative to have MRI-based atlases of mouse brains to guide and facilitate such studies. Given the advantages of MRI, it is not surprising that several mouse brain atlases already exist. Modern MRI-based mouse brain atlases contain high-resolution population-averaged brain images with detailed structural segmentations and annotations. An excellent overview of currently existing MRI-based mouse brain atlases can be found in Dorr et al. (20). The major applications of MRI-based mouse brain atlases include visualization of mouse brain anatomy, mapping and analyses of experimental data, and anatomical phenotyping in the mouse brain. In the following sections, we have presented the details involved in creating an MRI-based mouse brain atlas and have outlined the usefulness of such atlases through several applications.

2. Considerations in Creating an MRI-Based Mouse Brain Atlas

2.1. MR Images: The Foundations of MRI-Based Mouse Brain Atlases

An MRI-based mouse brain atlas may contain the following several components: MR images with one or multiple contrasts illustrating the mouse brain anatomy; detailed structural segmentations; and along with a user interface and image analysis tools. Two key features that are often used to compare existing MRIbased atlases are the spatial resolution and contrasts of the MR images in each atlas. These two important and interlocking factors ultimately determine the usefulness of an MRI-based atlas. Satisfactory image contrast and high spatial resolution are essential for resolving miniature structures in the brain. While it is possible to generate MRI-based atlases with multiple image contrasts, doing so will inevitably prolong the time needed to acquire a complete set of high-resolution data, which may become impractical due to degradation of the specimen and instrument instability. The choice of image contrast therefore often depends on the intended applications of the atlas. For example, if the intended applications of an atlas are analyses of T2 images of the mouse brain, then it is necessary to have T2 images in the atlas. Most current MRIbased mouse brain atlases are based on T2 or T2 ∗ MRI, because they are widely used to study mouse brain anatomy and pathology and provide satisfactory tissue contrast for structural delineation in the adult mouse brain. As for resolution is considered, high resolution requires longer imaging time and is not always practical. Clinically, anatomical images of the human brain can be routinely acquired at a resolution of 1 × 1 × 1 mm. Considering

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

255

that the human brain is approximately 3,000 times larger than the mouse brain in terms of total brain volume (21), a comparable resolution in the mouse brain would be 0.07 × 0.07 × 0.07 mm. Several existing adult mouse brain atlases provide resolution in the range of 0.03 × 0.03 × 0.03–0.05 × 0.05 × 0.05 mm (13, 22–24). Development of new imaging techniques that can improve image contrast and resolution is an important area of research. Recently, the use of “active staining” techniques (25) has also been introduced in the mouse brain that use contrast agents in conjunction with perfusion fixatives for increasing the tissue contrast and signal-to-noise ratio in MR images of the postmortem mouse brain (25, 26). Multi-spectral MR acquisition with enhanced T2 contrasts in the actively stained images has been shown to reveal more detailed morphological aspects of the mouse brain compared to conventional T2 imaging. With partial k-space acquisition and contrast agents that shorten tissue T1 (26), high-resolution mouse brain imaging with up to 21.5 μm isotropic resolution has been achieved (27). In the last decade, diffusion tensor imaging (DTI) has emerged as a novel MR technique that can reveal tissue microstructure with endogenous contrasts (28, 29). Tissue water diffusion can be characterized by diffusion constants along six different orientations. The distribution of diffusion constants along each direction is under the influence of local tissue microstructure. DTI measures these diffusion constants and fits them into a tensor model, from which the diffusion anisotropy and principal diffusion orientation are calculated. These parameters have been shown to have inherent sensitivity to tissue architecture and physiological conditions. The degree of anisotropy is sensitive to the existence of axonal projections and the degree of myelination (30). The principal diffusion orientation provides an additional white matter tissue contrast that is based on the trajectory of white matter tracts. The orientation and diffusion anisotropy data can be used to reconstruct the trajectories of white matter tracts in 3D, using the so-called fiber-tracking technique (31), as shown in Fig. 12.2b. Applications of DTI on human and animal brains have shown great potential for elucidating the complex white matter structures. Readers can find more information on this technique in detailed review articles (32–34). One important advantage of DTI over T1 and T2 MRIs is that it can provide superior contrasts to delineate anatomical structures in premature mouse brains (35, 36). Figure 12.3 shows ex vivo high-resolution T2 and diffusion tensor images of developing mouse brains from postnatal day 0 (p0, at birth) to postnatal day 80 (p80). Because the myelination process starts at approximately p7 in the mouse and is not completed until late postnatal stages and since tissue T2 is heavily influenced by myelin content, T2

256

Aggarwal, Zhang, and Mori

Fig. 12.2. Diffusion tensor imaging of an adult mouse brain. a Nissl-stained histology (left) and diffusion tensor images of a perfusion-fixed adult mouse brain. b Reconstructed white matter tracts from the DTI data. 2n, optic nerve; ac, anterior commissure; cc, corpus callosum; cp, cerebral peduncle; DG, dentate gyrus; ec, external capsule; f, fornix; fi, fimbria; H, hippocampus; ml, medial lemniscus; opt, optic tract; py, pyramidal tract; sm, stria medularis. The scale bar represents 1 mm. The color arrows indicate the color coding used for diffusion anisotropy orientation. Red, green, and blue represent rostral-caudal, medial-lateral, and dorsal-ventral orientations, respectively.

images of early postnatal mouse brains show limited contrast for white matter structures. In comparison, diffusion tensor images provide superior white matter contrast, and the contrast is consistent from p0 to p80 and later stages. White matter structures, for example, the cerebral peduncle (cp) and optic tract (opt), can be easily identified in diffusion tensor images, but not in T2 images. DTI can also reveal anatomical structures in embryonic mouse brains, for example, the cortical plate and subventricular zone in the embryonic mouse cortex. DTI is therefore an important MR contrast for MRI-based atlases of embryonic and neonatal mouse brains. 2.2. Structural Segmentation

As described in the introduction, an atlas needs both highresolution anatomical images with rich anatomical contrasts and detailed structural delineations to be useful. In several histology based mouse brain atlases, numerous structures are manually delineated on images of stained tissue sections based on

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

257

Fig. 12.3. T2 and diffusion tensor color map images of postnatal mouse brains. In the color map images, we have used R, G, B colors to visualize white matter orientation. Red indicates that local tissue orientation is perpendicular to the plane, green indicates horizontal orientation, and blue indicates vertical orientation. The orange arrows indicate the locations of forceps major of the corpus callosum during p0–p10 or the locations of splenium of the corpus callosum during p20–p80. The yellow arrows indicate the locations of medial lemniscus, and the pink arrows indicate the locations of fasciculus retroflexus.

expert knowledge of the unique cellular and molecular markers of each structure and their spatial relationships. In MRI-based brain atlases, because the images are stored in 3D format, structural segmentation in the 3D images, in which voxels that belong to a particular structure are selected and classified as a 3D entity, is the common form of structural delineation. Segmentations in mouse brain MR images have been done either manually (20, 37) or semi-automatically (21, 38). Figure 12.4 shows examples of structural segmentations in images of an adult mouse brain based on both manual segmentation and fiber tracking for white matter structures. The numbers of structural segmentations in existing MRI-based mouse brain atlases vary approximately from 20 to 70 and are limited compared to hundreds of structures defined in histology based mouse brain atlases. In addition to structural segmentations, a user-friendly interface also adds to the value of a brain atlas. A user interface should allow users to browse through the anatomical images and structural annotations and can be located in users’ computers or on the web. The lower panel of Fig. 12.4 demonstrates the interface of our current MRI/DTI-based mouse brain atlas. It allows users to navigate through multiple 3D MR images with different tissue contrasts, read coordinates, display segmented structures and view overlays of 2D and 3D structural definitions, and to visualize the reconstructed structures in 3D.

258

Aggarwal, Zhang, and Mori

Fig. 12.4. Structural segmentations and user interface of an MRI-based mouse brain atlas. Upper panels: coronal images of an adult mouse brain from our MRI/DTI-based mouse brain atlas. The top panel shows the T2 -weighted (T2 ), fractional anisotropy (FA), and direction-encoded color map (DEC) images, and the bottom panel shows the images overlaid with reconstructed white matter and gray matter structures currently included in the atlas. The color schemes for gray matter structures in T2 and FA images are as follows: white, neocortex; blue, hippocampus; purple, striatum. The color schemes for white matter structures are as follows: yellow: fimbria (fi); pink: corpus callosum (cc); magenta: fasciculus retroflexus (fr); red: stria terminalis (st); light green: fornix (fx); dark green: the trigeminal nerve (5n); light blue: cerebral peduncle (cp); dark blue: stria medularis (sm); light purple: mammillothalamic tract (mt); purple: optic tract (opt). Bottom panel: user interface of our data-viewing software, “AtlasView.” The software allows users to navigate through 3D MRI multicontrast data of a mouse brain. It overlays 2D and 3D structural definitions and visualizes reconstructed structures in 3D.

2.3. In Vivo and Ex Vivo Mouse Brain Atlases

Most current MRI-based atlases are based on images acquired from postmortem samples because it is relatively easy to acquire high-resolution images from postmortem specimens than from

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

259

live animals. Postmortem specimens, mostly formaldehyde fixed, can be trimmed to fit into small and sensitive MR coils, which, when combined with a high-performance gradient system, can generate high-resolution images with superior quality. Such combinations of optimized imaging hardware are often not available for in vivo MRI due to the added requirement of animal monitoring and support system. In addition, ex vivo MRI can last for 12 h or even longer, which can further improve the image quality by increasing signal averaging, while in vivo imaging may be susceptible to artifacts due to subject motion or the stability of animals, and is mostly limited to 2–3 h. All these factors have made ex vivo imaging the ideal choice for acquiring high-resolution and highquality images of the mouse brain. In comparison, in vivo MRI is limited by the sensitivity of the imaging instruments and the time that animals can stay relatively stable in the magnet under anesthesia. Even with various triggering techniques, e.g., ECG and respiratory triggers, it is still difficult to completely eliminate the effects of motion on the acquired images. The residual motion will cause a certain degree of degradation in the resolution and contrast of the images, as shown in Fig. 12.1. Even with the aforementioned disadvantages of in vivo MRI compared to ex vivo MRI, atlases based on in vivo MRI are still useful if the intended application of the atlas is to analyze in vivo MR images of the mouse brain. It has been reported that tissue contrasts in in vivo and ex vivo images are not entirely the same. The fixation process can alter the local physical and chemical environments experienced by water molecules due to the crosslinking of fixatives with macromolecules, which alters the appearance or contrast of the brain structures in ex vivo MRI compared to in vivo MRI. Another consideration is that there are noticeable morphological changes between the brains in live animals and in perfusion-fixed brain specimens. The lateral ventricles in postmortem specimens often have significantly reduced volumes or are completely collapsed due to removal of the cerebrospinal fluid (CSF) pressure. Also depending on the osmotic pressure of the fixation solution, the brain tissue may show slight enlargement and/or shrinkage. In Table 12.1, we have compared the structural volumes of the same mouse brains measured in vivo and ex vivo. The drastic changes in the volumes of the lateral ventricles can be appreciated. Although more detailed analyses are still needed, the results here suggest that ex vivo MRI-based atlases may not represent in vivo brain morphology accurately, necessitating the generation of in vivo MRI-based atlases (39). 2.4. Single-Subject and PopulationAveraged Brain Atlases

While histology based atlases are mostly based on one or a few specimens, the 3D images generated by MRI and the recent advances in computational techniques make it possible to perform spatial normalization and averaging of images from multiple

260

Aggarwal, Zhang, and Mori

Table 12.1 Comparisons of regional brain volumes (mm3 ) measured using in vivo MRI and ex vivo MRI from perfusion-fixed mouse brains. Data are expressed as mean ± SE (n=3). Brain region

In vivo MRI

Ex vivo MRI

Whole brain

486.95±9.32

465.65±2.12

4.4

Striatum

29.87±0.94

23.37±0.07

21.8

Hippocampus

28.45±2.27

26.34±0.84

7.4

119.02±2.09

110.74±1.92

6.9

64.05±1.98

59.30±1.81

7.4

Cortex Cerebellum Lateral ventricle

4.88±1.27

0.05±0.001

% shrinkage

99.0

specimens. It is also possible to use these techniques to generate so-called minimal deformation atlases that approximate the geometrical average of normal brains and reflect the average morphological characteristics of the sample population. Detailed procedures for generating unbiased population-averaged brain images can be found in (37, 40, 41). An additional benefit of group averaging is that the averaged images have higher signal-to-noise ratio than the individual images, and therefore, facilitate structural segmentation and visualization (20). While it is beneficial to have population-averaged brain atlases, the usefulness of these atlases can be constrained by the image normalization techniques adopted. High accuracy of the image normalization step is critical for atlases based on population averaging. Since the spatial normalization of images is based on similarity of their pixel intensity values, the subject and template images should have similar contrast patterns for the mapping to be accurate. Ex vivo images have high spatial resolution and strong structural contrasts, which makes it relatively easy to construct mappings between ex vivo images in order to create population-averaged atlases. In vivo images, on the other hand, have limited structural contrasts as explained before. It is therefore more challenging to achieve accurate mapping between in vivo images. Averaging of inaccurately normalized images can cause further degradation of the tissue contrast and result in poor quality in registration of subsequent in vivo subject images to the population-averaged brain image. Figure 12.5 A shows our current population-averaged ex vivo mouse brain atlas with averaged T2 and DTI data and structural segmentation. With the recent advances in fast in vivo brain imaging, we now demonstrate that it is possible to construct in vivo population-averaged brain atlases using DTI. Images of 2month-old female C57BL/6 mouse brains were acquired with a

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

261

Fig. 12.5. In vivo and ex vivo group-averaged mouse brain atlases based on T2 MRI and DTI. a Group-averaged ex vivo T2 and diffusion tensor images of adult mouse brains (C57BL/6, 2 month old, n = 10) and structural segmentations (Atlas). b Group-averaged in vivo mouse brain atlas with preliminary structural segmentations (C57BL/6, 2 month old, n = 9). c Mid-sagittal view of the population-averaged in vivo images and structural segmentations. In the direction-encoded colormap (DEC) images, red, green, and blue represent the medial-lateral, anterior-posterior, and superior-inferior axes, respectively.

spatial resolution of 0.1 × 0.1 × 0.4 mm3 and total imaging time of 2–3 h. After iterative spatial normalization and averaging, the population-averaged images from nine mouse brains were generated (Fig. 12.5b). The rich tissue contrasts provided by DTI and high spatial resolution provided by fast 3D imaging techniques enable accurate spatial normalization, as can be appreciated from the sharp structural boundaries in the averaged images. Because the images were acquired in 3D fashion, the data can be viewed and manipulated in 3D (Fig. 12.5c), which is important for examining 3D structural properties. We have performed initial segmentation of 58 structures in the averaged brain images (Fig. 12.5 Atlas) based on our ex vivo atlas. In general, ex vivo T2 images have richer tissue contrasts than in vivo T2 images, probably due to tissue fixation and higher spatial resolution. The volume of the lateral ventricles in the in vivo images is significantly larger than the ex vivo images due to the lack of CSF pressure in postmortem samples. In comparison, the DTI contrast patterns are relatively undisturbed from in vivo to ex vivo, a finding that has also been reported by other groups (42, 43).

3. Applications of MRI-Based Mouse Brain Atlas 3.1. Morphological Studies

Measuring changes in brain morphology (shapes and volumes of specific anatomical areas) is important for many studies, but not always straightforward even with 3D MRI data. Manual delineation of structures in serial MR images is time-consuming and the results from different operators and different laboratories may

262

Aggarwal, Zhang, and Mori

not be compatible due to differences in definition of anatomical boundaries and inter-rater variations. With the availability of brain atlases and image registration and mapping techniques, we now can characterize brain morphology efficiently and consistently. In this section, an application of atlas-based analysis in quantitative characterization of neurodegenerative atrophy in the R6/2 mouse brain is demonstrated. The R6/2 transgenic mouse is a widely used model of Huntington’s disease (HD) in pre-clinical therapeutic trials. It has progressive HD-like gross atrophy in the brain and especially in the striatum (44). We collected high resolution 3D T2 -weighted images of R6/2 mice (n = 7) and wild-type littermates (n = 8) longitudinally from 3 to 12 weeks after birth. To analyze the differences in structural volumes between R6/2 and wild-type mice, a single-subject mouse brain atlas was first created. The mouse brain images used in the atlas were selected from a set of in vivo T2 -weighted MR images of C57BL/6 mice (n = 10, female, 12 weeks old), one of the background strains of the R6/2 strain. The atlas image has whole brain and ventricular volumes (obtained via manual segmentation) close to the median values of the 10 mice. The image was then manually adjusted to the orientation defined in the Paxinos’ atlas (3) and resampled to an isotropic resolution of 0.1 mm × 0.1 mm × 0.1 mm per pixel. This atlas contains manual segmentation of 10 brain structures that follow closely the definition by Paxinos (3). The brain was segmented from the rostral ends of the olfactory bulbs to the caudal end of the cerebellum; the cortex was defined by the corpus callosum and external capsule, with the ventral boundary by rhinal fissure (we excluded the part ventral to the rhinal fissure in this study); the striatum was defined by the corpus callosum, external capsule, and anterior commissure; the hippocampus was defined by the external capsule, lateral and third ventricle, and thalamus; and the ventricle was defined by intense signal from CSF. The boundary of striatum and accumbens is often difficult to identify even with histology slides, but borders of the corpus callosum, external capsule, anterior commissure, and globus pallidus are clear in MRI. We used clearly identifiable anatomical landmarks and the Paxinos’ atlas to define the boundary as reproducibly as possible. Linear and nonlinear image transformations were used for spatial normalization of R6/2 and wild-type mouse brain images. Figure 12.6 shows the images from three R6/2 and three wildtype mouse brains after linear affine transformation and nonlinear registration based on the LDDMM technique (45). Affine transformation can adjust the overall size and orientation of each image to the atlas, but to normalize detailed morphology, e.g., the shape of the lateral ventricles, to the atlas requires accurate nonlinear transformation. LDDMM can accurately transform both wild-type and R6/2 mouse brain images to the atlas so that the

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

263

Fig. 12.6. A diagram that illustrates atlas-based analysis of brain atrophy in the R6/2 mouse, a model of Huntington’s disease. The atlas is a single-subject atlas with embedded structural delineations. For purpose of illustration, we have used the yellow dashed line along the brain surface, the blue curve on the corpus callosum and external capsule, and the solid green line outlining the ventricles to represent the structural delineations in the atlas. Images from R6/2 and wild-type mice (shown here with three mice in each group) were spatially normalized to the atlas image using affine transformation and LDDMM. After spatial normalization, the structure delineation defined in the atlas image can be transferred and overlaid on the transformed subject images.

structural delineations embedded in the atlas can be directly transferred to the normalized images. Using inverse transformations that deform the atlas image and structural delineations to the subject images, automated segmentation of major gray matter structures (neocortex, striatum, hippocampus, cerebellum, etc) can be achieved. The accuracy of this segmentation approach was examined using expert manual segmentation as the gold standard. Strong correlations between the manual and atlas-based segmentation results (Table 12.2) indicate the accuracy of the atlas-based approach. Figure 12.7 further demonstrates the quality of atlas-based automated segmentation of p21, p42, and p84 mouse brain images with LDDMM. The transformations generated by LDDMM carry structural segmentation in the atlas to each subject image. The segmented structures are outlined and overlaid on in vivo MR images, and the reconstructed surfaces were visualized in 3D. Volumes of major brain structures were obtained via automated segmentation followed by manual correction (Fig. 12.7 b–e). The R6/2 mice show significantly reduced whole brain volume and hippocampal volume at 4 weeks of age (p28). Differences in the striatal volume and lateral ventricles become significant at 5 weeks of

264

Aggarwal, Zhang, and Mori

Table 12.2 R2 values of correlation analysis of the manual segmentation and LDDMM-based automated segmentation results at 4, 6, and 12 weeks of age in both the wild-type and R6/2 (within the parentheses) mice. Structures

4 weeks

6 weeks

12 weeks

Striatum

0.841 (0.997)

0.959 (0.925)

0.956 (0.571)

Hippocampus

0.985 (0.985)

01.977 (0.653)

0.899 (0.704)

Lateral ventricles

0.968 (0.969)

0.942 (0.933)

0.952 (0.919)

Cortex

0.966 (0.942)

0.618 (0.985)

0.771 (0.871)

age (p35). The cerebellar volume showed no significant change. These results show that MRI can detect atrophy in R6/2 mice as early as p28, and the atrophy progresses afterwards. These results show the usefulness of the atlas for detecting atrophy in mouse brains. 3.2. Combined Micro-CT and DTI-Based Atlas for Guidance of Stereotaxic Operations

The precision of targeting structures during stereotaxic surgery in the mouse brain depends on the accuracy of the brain atlas used to guide such operations. 2D histology based atlases that provide stereotaxic coordinates of brain sections relative to skull landmarks are conventionally used for guidance of stereotaxis in the mouse brain. MRI-based mouse brain atlases can provide anatomical information of the brain, but lack bone tissue contrasts for identification of cranial landmarks, which are essential for stereotaxis. Recently, 3D stereotaxic mouse brain atlases based on combining MR images with micro-CT have been developed (37, 46). To construct single-subject stereotaxic atlases, C57BL/6 mouse head specimens at different stages of postnatal development were scanned with MRI and subsequent micro-CT. To incorporate the brain images in skull-based stereotaxic coordinates, the 3D MR images were coregistered to micro-CT images of the same subject, that provide fine bone tissue contrasts for delineation of skull surface landmarks (Fig. 12.8). The coregistered CT-MRI images were oriented with the lambda and bregma landmarks in the same horizontal plane, which is the standard orientation used in most histology based stereotaxic atlases of murine brains (3, 47). The atlases contain CT and multiple MR contrasts (T2 -weighted, diffusion-weighted, fractional anisotropy and diffusion orientation). As mentioned previously, DTI contrasts are particularly useful for delineation of anatomical structures during early postnatal stages, when incomplete myelination

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

265

Fig. 12.7. a Automated segmentation of R6/2 mouse brain images. Pre-defined segmentation of major gray matter structures (caudate putamen: yellow; hippocampus: green; cerebellum: purple; lateral ventricle: blue) in the atlas mouse brain images was transformed to individual images. b–e MR-based volume measurements of the brain, lateral ventricles, striatum, and hippocampus. ∗ Significant (p<0.05) difference between wild-type and R6/2 mice.

in the CNS limits the level of contrast revealed by T2 -weighted or diffusion-weighted MR imaging (Fig. 12.8). At the level of resolution currently achievable in MRI atlases, the level of anatomical detail and structural delineation in the

266

Aggarwal, Zhang, and Mori

Fig. 12.8. Coregistration of micro-CT skull and MR brain images of C57BL/6 mice at different developmental stages from postnatal day 7 (p7) to adult. CT images (displayed in metallic color) are overlaid on isotropically diffusion-weighted images (displayed in gray-scale). The right panel shows DTI-derived anisotropy contrasts with red, green, and blue representing the medial-lateral, rostral-caudal, and dorsal-ventral orientations respectively. Scale bar represents 2 mm.

mouse brain is not comparable to that of histology based atlases. However, the ability to visualize brain anatomy in 3D renders the CT-MRI atlases advantageous for several stereotaxic applications. For instance, the atlases can be important for surgical targeting of a specific brain region, where it is not desirable to damage important adjoining structures during the operation. An example of the application of the atlas for pre-surgical planning of stereotaxic operations is demonstrated in Fig. 12.9; with 3D visualization of a “hypothetical” needle path that can be virtually moved and rotated in the atlas space, it is possible to determine the target coordinates and precise angle of injection in order to avoid hitting or damaging adjoining structures, which may be difficult with serial 2D histological sections. The atlases can be rotated and virtually sliced in any desired orientation, and the stereotaxic coordinates can be redefined relative to any user-specified land-

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

267

Fig. 12.9. User interface of the CT-MRI stereotaxic mouse brain atlas for guidance of stereotaxic operations. The stereotaxic coordinates of any brain region can be directly read by moving the mouse cursor to its location. Pre-segmented anatomical structures can be visualized in 3D for surgical planning of the stereotaxic injection. In the example illustrated, the angle of injection for targeting the thalamus (yellow) without damaging adjoining structures such as the caudateputamen (green) and hippocampus (brown) can be determined by visualization of a hypothetical needle path (shown in red) that can be virtually moved and rotated.

mark in addition to the conventional bregma–lambda coordinate system, thereby allowing greater flexibility in surgical placement of the experimental animal. As mentioned previously, since stereotaxic surgery is typically performed in the mouse brain in vivo, atlases based on ex vivo specimens may not offer an accurate representation of the target brain anatomy. To account for postmortem tissue shrinkage, as well as anatomical variability in the mouse brain, we developed a population-averaged in vivo MRI-based brain template for the adult mouse brain (C57BL/6, n = 9, 5 months old). High resolution ex vivo MR images were corrected for postmortem tissue shrinkage by LDDMM-based nonlinear mapping to the in vivo reference brain template and then incorporated in population-averaged skull-based stereotaxic coordinates. The shrinkage-corrected stereotaxic atlas offers high anatomical fidelity to in vivo adult mouse brains. A comparison of the shrinkage-corrected CT-DTI atlas with existing histology based stereotaxic atlases of the adult C57BL/6 mouse brain (3, 4) is

268

Aggarwal, Zhang, and Mori

Fig. 12.10. Comparison of postmortem shrinkage-corrected CT-DTI atlas of the adult C57BL/6 mouse brain with existing histology based stereotaxic atlases by Franklin and Paxinos and the Allen Institute for Brain Sciences. a Mid-sagittal section (x = 0 mm): top and bottom panels show color-coded orientation maps derived from DTI overlaid on the Paxinos and the Allen Institute atlas sections, respectively. b and c Coronal sections at y = –1.06 mm (top) and –2.94 mm (bottom). In each coronal section, semi-sections from the Franklin and Paxinos atlas and the Allen Institute atlas are overlaid on the left and right halves of the DTI section, respectively. White arrows indicate differences in stereotaxic locations of brain structures between the atlases, and white circles indicate the regions with perfect alignment between the atlases. Crosshairs indicate the location of the bregma landmark, which is chosen as the origin of the stereotaxic coordinate system. All scales are in millileter.

illustrated in Fig. 12.10, and reveals significant differences in stereotaxic coordinate locations of several internal brain structures, as well as in the overall medial-lateral width and horizontal orientation of the brain. These results highlight the usefulness of 3D CT- and MRIbased stereotaxic mouse brain atlases. They also demonstrate the potential applications of in vivo MRI and advanced brain mapping techniques for further improving the accuracy of surgical targeting in the mouse brain.

4. Summary In this chapter, we have presented several technical considerations involved in creating MRI-based mouse brain atlases and important applications of such atlases in neuroscience. The ultimate goal of MRI-based mouse brain atlases is to help scientists detect and monitor anatomical changes in mouse brains due to either genetic mutations, disease processes, or other causes. Technical advances in the field of MRI, which will enable high-resolution imaging of the mouse brain with novel image contrasts, and computational

Magnetic Resonance Imaging-Based Mouse Brain Atlas and Its Applications

269

techniques, which will enhance our ability to use brain atlases to analyze images, will propel the further development and broaden the application of MRI-based mouse brain atlases. References 1. Lein, E. S., Hawrylycz, M. J., Ao, N. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 2007;445: 168–176. 2. MacKenzie-Graham, A., Jones, E. S., Shattuck, D. W., Dinov, I. D., Bota, M., Toga, A. W. The informatics of a C57BL/6j mouse brain atlas. Neuroinformatics 2003;1: 397–410. 3. Paxinos, G., Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates, 2nd ed. San Diego, CA: Academic Press; 2003. 4. Dong, H., Science TAIfB. The Allen Reference Atlas: A Digital Color Brain Atlas of the C57BL/6j Male Mouse (DVD Edition). Hoboken, NJ: Wiley; 2008. 5. Jacobowitz, D., Abbott, L. Chemoarchitectonic Atlas of Developing Mouse Brain. Boca Raton, FL: CRC; 1997. 6. Kaufman, M. Atlas of Mouse Development. Maryland Heights, MO: Academic Press; 1992. 7. Paxinos, G., Halliday, G. H., Watson, C., Koutcherov, Y., Wang, H. Atlas of the Developing Mouse Brain at E17.5, P0 and P6. Maryland Heights, MO: Academic Press; 2006. 8. Schambra, U. Prenatal Mouse Brain Atlas. New York, NY: Springer; 2008. 9. Sidman, R. L., Angevine, J., Pierce, E. Atlas of the Mouse Brain and Spinal Cord (Commonwealth Fund Publications). Cambridge, MA: Harvard University Press; 1971. 10. Valverde, F. Golgi Atlas of the Postnatal Mouse Brain. New York, NY: Springer; 2004. 11. Baldock, R., Bard, J., Brune, R. et al. The Edinburgh mouse atlas: Using the CD. Brief Bioinform 2001;2:159–169. 12. McDonald, D. M., Choyke, P. L. Imaging of angiogenesis: From microscope to clinic. Nat Med 2003;9:713–725. 13. Dorr, A., Sled, J. G., Kabani, N. Threedimensional cerebral vasculature of the CBA mouse brain: A magnetic resonance imaging and micro computed tomography study. Neuroimage 2007;35: 1409–1423. 14. Sharpe, J., Ahlgren, U., Perry, P. et al. Optical projection tomography as a tool for 3D microscopy and gene expression studies. Science 2002;296:541–545.

15. Catalaa, I., Grossman, R. I., Kolson, D. L. et al. Multiple sclerosis: Magnetization transfer histogram analysis of segmented normal-appearing white matter. Radiology 2000;216:351–355. 16. Turner, B., Lin, X., Calmon, G., Roberts, N., Blumhardt, L. D. Cerebral atrophy and disability in relapsing-remitting and secondary progressive multiple sclerosis over four years. Mult Scler 2003;9:21–27. 17. Agosta, F., Absinta, M., Sormani, M. P. et al. In vivo assessment of cervical cord damage in MS patients: A longitudinal diffusion tensor MRI study. Brain 2007;130: 2211–2219. 18. Werring, D. J., Toosy, A. T., Clark, C. A. et al. Diffusion tensor imaging can detect and quantify corticospinal tract degeneration after stroke. J Neurol Neurosurg Psychiatry 2000;69:269–272. 19. Thomalla, G., Glauche, V., Koch, M. A., Beaulieu, C., Weiller, C., Rother, J. Diffusion tensor imaging detects early wallerian degeneration of the pyramidal tract after ischemic stroke. Neuroimage 2004;22: 1767–1774. 20. Dorr, A. E., Lerch, J. P., Spring, S., Kabani, N., Henkelman, R. M. High resolution three-dimensional brain atlas using an average magnetic resonance image of 40 adult C57Bl/6j mice. Neuroimage 2008;42: 60–69. 21. Ali, A. A., Dale, A. M., Badea, A., Johnson, G. A. Automated segmentation of neuroanatomical structures in multispectral MR microscopy of the mouse brain. Neuroimage 2005;27:425–435. 22. MacKenzie-Graham, A., Lee, E. F., Dinov, I. D. et al. A multimodal, multidimensional atlas of the C57BL/6j mouse brain. J Anat 2004;204:93–102. 23. Ma, Y., Hof, P. R., Grant, S. C. et al. A three-dimensional digital atlas database of the adult C57BL/6j mouse brain by magnetic resonance microscopy. Neuroscience 2005;135:1203–1215. 24. Badea, A., Ali-Sharief, A. A., Johnson, G. A. Morphometric analysis of the C57BL/6j mouse brain. Neuroimage 2007;37: 683–693.

270

Aggarwal, Zhang, and Mori

25. Sharief, A. A., Johnson, G. A. Enhanced T2 contrast for MR histology of the mouse brain. Magn Reson Med 2006;56:717–725. 26. Sharief, A. A., Badea, A., Dale, A. M., Johnson, G. A. Automated segmentation of the actively stained mouse brain using multi-spectral MR microscopy. Neuroimage 2008;39:136–145. 27. Johnson, G. A., Ali-Sharief, A., Badea, A. et al. High-throughput morphologic phenotyping of the mouse brain with magnetic resonance histology. Neuroimage 2007;37: 82–89. 28. Basser, P. J., Pierpaoli, C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209–219. 29. Le Bihan, D., Breton, E., Lallemand, D., Grenier, P., Cabanis, E., Laval-Jeantet, M. MR imaging of intravoxel incoherent motions: Application to diffusion and perfusion in neurologic disorders. Radiology 1986;161:401–407. 30. Beaulieu, C. The basis of anisotropic water diffusion in the nervous system – a technical review. NMR Biomed 2002;15:435–455. 31. Mori, S., Zijl, P. C. M. V. Fiber tracking: Principles and strategies. NMR Biomed 2002:1–14. 32. Basser, P. J., Jones, D. K. Diffusion-tensor, MRI: Theory, experimental design and data analysis – a technical review. NMR Biomed 2002;15:456–467. 33. Le Bihan, D. Looking into the functional architecture of the brain with diffusion MRI. Nat Rev Neurosci 2003;4:469–480. 34. Mori, S., Zhang, J. Principles of diffusion tensor imaging and its applications to basic neuroscience research. Neuron 2006;51:527–539. 35. Mori, S., Itoh, R., Zhang, J. et al. Diffusion tensor imaging of the developing mouse brain. Magn Reson Med 2001;46:18–23. 36. Zhang, J., Richards, L. J., Yarowsky, P., Huang, H., van Zijl, P. C., Mori, S. Three-dimensional anatomical characterization of the developing mouse brain by diffusion tensor microimaging. Neuroimage 2003;20:1639–1648. 37. Aggarwal, M., Zhang, J., Miller, M. I., Sidman, R. L., Mori, S. Magnetic resonance imaging and micro-computed tomography combined atlas of developing and adult

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

mouse brains for stereotaxic surgery. Neuroscience 2009;162:1339–1350. Chen, X. J., Kovacevic, N., Lobaugh, N. J., Sled, J. G., Henkelman, R. M., Henderson, J. T. Neuroanatomical differences between mouse strains as shown by highresolution 3D MRI. Neuroimage 2006;29: 99–105. Ma, Y., Smith, D., Hof, P. R. et al. In vivo 3D digital atlas database of the adult C57BL/6j mouse brain by magnetic resonance microscopy. Front Neuroanat 2008;2:1. Kovacevic, N., Henderson, J. T., Chan, E. et al. A three-dimensional MRI atlas of the mouse brain with estimates of the average and variability. Cereb Cortex 2005;15: 639–645. Lee, E. F., Jacobs, R. E., Dinov, I., Leow, A., Toga, A. W. Standard atlas space for C57BL/6j neonatal mouse brain. Anat Embryol (Berl) 2005;210:245–263. Sun, S. W., Liang, H. F., Le, T. Q., Armstrong, R. C., Cross, A. H., Song, S. K. Differential sensitivity of in vivo and ex vivo diffusion tensor imaging to evolving optic nerve injury in mice with retinal ischemia. Neuroimage 2006;32:1195–1204. Sun, S. W., Neil, J. J., Liang, H. F. et al. Formalin fixation alters water diffusion coefficient magnitude but not anisotropy in infarcted brain. Magn Reson Med 2005;53:1447–1451. Stack, E. C., Kubilus, J. K., Smith, K. et al. Chronology of behavioral symptoms and neuropathological sequela in R6/2 huntington’s disease transgenic mice. J Comp Neurol 2005;490:354–370. Miller, M. I., Trouve, A., Younes, L. On the metrics and Euler-Lagrange equations of computational anatomy. Annu Rev Biomed Eng 2002;4:375–405. Chan, E., Kovacevic, N., Ho, S. K., Henkelman, R. M., Henderson, J. T. Development of a high resolution three-dimensional surgical atlas of the murine head for strains 129S1/svimj and C57Bl/6j using magnetic resonance imaging and micro-computed 2007;144: tomography. Neuroscience 604–615. Paxinos, G., Watson, C. The Rat Brain in Stereotaxic Coordinates, 4th ed. San Diego, CA: Academic Press; 1998.

Chapter 13 CEST MRI Reporter Genes Guanshu Liu, Jeff W.M. Bulte, and Assaf A. Gilad Abstract In recent years, several reporter genes have been developed that can serve as a beacon for non-invasive magnetic resonance imaging (MRI). Here, we provide a brief summary of recent advances in MRI reporter gene technology, as well as detailed “hands-on” protocols for cloning, expression, and imaging of reporter genes based on chemical exchange saturation transfer (CEST). Key words: Chemical exchange saturation transfer, reporter gene, molecular imaging, cell tracking, cloning, expression vector.

1. Introduction One of the major challenges of the post-genome era is to study the expression of genes in a physiologically viable context. This will involve not just the histological analysis and extraction of DNA, RNA, and protein from tissue specimens, but rather, imaging gene expression patterns in the whole organism in vivo. The development of recombinant DNA cloning techniques in the 1970s enabled the transfer of genetic material from one organism to another. This subsequently led to the developing of reporter genes, with cells expressing xenogeneic proteins. The encoding gene is either fused to the gene of interest or replaces it. The main applications for these reporters are (1) monitoring gene expression levels, (2) investigating dynamic molecular interactions between proteins, (3) studying cellular interactions, (4) tracking cells in normal/abnormal development or in cell transplantation therapy, and (5) monitoring gene M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_13, © Springer Science+Business Media, LLC 2011

271

272

Liu, Bulte, and Gilad

replacement therapy. The majority of reporter genes has been developed for optical imaging. The genes can encode proteins that emit light upon absorption of photons at a specific frequency (fluorescence) or at the cleavage of a substrate (bioluminescence). In both cases, the released photon is detected in a frequencydependent manner using a charge-coupled device (CCD) to generate an image. Throughout the years, reporter genes have been developed for other imaging modalities, including nuclear imaging (1). The quest to develop reporter genes for magnetic resonance (MR) began with overexpressing creatine kinase (2) and arginine kinase (3). The transgene expression was detected via metabolic changes with MR spectroscopy. Later, the overexpression of proteins that are naturally involved in iron metabolism (4, 5) and storage (6, 7) was used for MR imaging. The rationale was that iron is a (super) paramagnetic metal, which can provide MRI contrast. A different strategy was to express a transgenic enzyme, βgalactosidase, which converts certain compounds to (super) paramagnetic contrast agents (8–10). With the exception of ferritin (6, 7) and green fluorescent protein (GFP), that contains specific protons visible with magnetization transfer (MT) MRI (11), all other MRI reporter genes rely on the administration of a substrate. An important consideration is that the pharmacokinetics of these substrates may not allow access to all tissues, especially to the central nervous system (CNS). A recently developed alternative reporter gene is based on previous observations that poly-L-lysine has a uniquely high chemical exchange rate, suitable for CEST MRI (12, 13). An artificial gene called LRP (lysine-rich protein), which has a high percentage of lysine residues, was designed de novo (Fig. 13.1). This gene provided contrast based on CEST MRI that was sufficient to distinguish 9 L rat glioma cells overexpressing the transgene from control 9 L cells (14) (Fig. 13.2). A broader overview on MR reporter genes can be found elsewhere (15, 16). Chemical exchange saturation transfer (CEST) is a new type of MRI contrast that has been recently developed to highlight the exchangeable protons by reducing the bulk water signal through chemical exchange (17). CEST is implemented by applying a saturation radiofrequency (RF) pulse, or pulses, at the exchangeable proton resonance frequency for a long duration to saturate the proton’s magnetization namely chemical exchange, since these protons constantly exchange with bulk water protons, they can be detected as a reduction in the water proton MR signal (18). In addition to a vast number of applications (a comprehensive summary can be found in a recent paper (18, 19)), CEST imaging shows great promise for the imaging of reporter genes (14). There are three major advantages for using a CEST

CEST MRI Reporter Genes

273

Fig. 13.1. CEST imaging and cloning of the LRP reporter gene. a Frequency selective radiofrequency pulses label amide protons (green). These exchange with water protons, reducing MRI signal intensity. b Outline of cloning of LRP. c, d Immunofluorescent polyclonal-antibody staining against LRP (red) and nuclear staining (blue) in rat 9 L-glioma cells. Only cells expressing LRP (c) show intense cytoplasmic antibody binding; control cells (d) show mainly non-specific peri-nuclear staining. e MTS assay for mitochondrial metabolic rates (mean±SD) shows no inhibitory effect of LRP on cell metabolism/proliferation. Reproduced, with permission, from Ref. 14.

reporter gene. First, biocompatible peptides produced through the genetic manipulation can be used, rather than metal-based exogenous contrast agents, as MRI-visible reporters, which can reduce the risk of disturbing the microenvironment of cells. Second, CEST contrast is switchable, which does not interfere with other MRI contrasts, such as T1 and T2 , when the saturation pulse is switched off. The most desirable advantage, however, is the potential to create different “colors,” by applying a saturation pulse at different frequencies, for simultaneously imaging more than one target, as optical imaging often does (20). With the aid of this multicolor approach, it is possible to highlight multiple cells or the expression of multiple genes in a CEST MR image.

274

Liu, Bulte, and Gilad

Fig. 13.2. In vivo imaging of LRP-transfected rat glioma cells. a, e anatomical images; b CEST SI difference map overlaid on (a) distinguishes LRP-expressing and control xenografts; c SI differences (mean±SD; six mice, each containing two xenografts; ∗ p=0.03, two-tailed, unpaired t-test); d RT-PCR of xenografts from rat brains. f Eosin–hematoxilyn stain of a frozen section corresponding to LRP tumor in (e); g magnification of (f) shows a uniform tumor mass. Reproduced, with permission, from Ref. (14).

2. Materials 2.1. General Requirements

1. A high-field, small animal narrow bore or clinical MR scanner with a relatively homogeneous main magnetic field, fast and reliable gradient coils, and a high signal-to-noise RF coil. 2. Devices and equipment for animal anesthesia, motion restraint, physiological monitoring, and body temperature and respiration maintenance. 3. Imaging processing software, e.g., Matlab (Mathworks, Natick, MA) or IDL.

2.2. Gene Cloning, Vector Construction, Cell Culture Conditions, and Transfection 2.2.1. Genes

One main advantage of using artificial CEST reporter genes is the availability of a large variability in amino acid sequence since

CEST MRI Reporter Genes

275

different sequences can provide similar CEST contrast. LRP (14) has been developed as a prototype reporter gene, and a higher level of expression might be achieved due to codon optimization (minimizing repeats) and amino acid sequence optimization (20). The gene can be generated by cloning, in tandem, of DNA oligomers encoding to lysine (AAA or AAG), or, alternatively, genes can be synthesized and commercially purchased (e.g., Blue Heron Biotechnology, BOTHELL WA). 2.2.2. Expression Vectors

The expression vectors, either plasmids or viral vectors, should be selected for the specific experiment. The vector should have a promoter that allows the expression of the protein tailored to the cell type being used.

2.3. MRI

The following pulse sequences should be available. 1. A fast 3D localizer sequence and sequences for anatomical imaging, such as multi-slice spin echo (SE) and multi-slice gradient echo (GRE). 2. A localized spectroscopy sequence, such as PRESS (pointresolved spectroscopy). 3. A fast spin-echo sequence, such as RARE (rapid acquisition with relaxation enhancement), or a fast Gradient Echo sequence, such as EPI (Echo Planar Imaging), including a magnetization transfer (MT) module, with which a saturation pulse can be manipulated for the desired pulse shape, power, duration, and offset.

3. Methods 3.1. Cloning and Transfection 3.1.1. Cloning of CEST Reporter Genes

Two complimentary synthetic oligonucleotides (84 base pairs long), encoding the artificial desired sequence, should be designed, so that, after annealing, they will retain corresponding endonuclease restriction site overhangs (e.g., to Bgl II at the 5 end and BamH I at the 3 end). This double-strand DNA is referred to as the monomer. The monomer should be cloned into the expression vector in the context of the desired promoter that will allow expression in the target cells. In addition, a start (ATG) codon is required to initiate translation, preferably in the context of a Kozak sequence. Cloning the gene in the frame with an antigen tag (such as HA, V5, myc, etc.) is useful for subsequent detection with immunohistochemistry. After cloning, the new vector should be digested with Bgl II and BamH I, and the released insert should be ligated into the Bgl II sites of the

276

Liu, Bulte, and Gilad

parental vector (see Note 1). The ligation will result in a dimer (i.e., a sequence encoding to polypeptide with double length). Next, the new vector should be digested with Bgl II and BamH I, and the released insert (dimer) should be ligated into the Bgl II sites of the parental vector to form a tetramer. This process should be repeated until the DNA sequence is sufficiently long (see Note 2). 3.1.2. Cell Culture and Transfection

The cell line of choice is transfected with the new construct. In general, any transfection or viral infection protocol is adequate; here, we have briefly described a general protocol for transfection using lipofectamine 2000 from Invitrogen (see Note 3). A day before transfection, cells are plated in 10-cm tissue culture dish in a dilution such that the cells will be at 80% confluence on the day of transfection. 1. Lipofectamine 2000 (60 μl) is diluted in 1 ml optiMEM, mixed gently, and allowed to stand for 5 min. Then, 24 μg DNA is added to 1 ml optiMEM. The diluted DNA and lipofectamine solution are combined and incubated for 20 min at room temperature. Cells are washed with optiMEM or PBS, and the medium is replaced with 8 ml optiMEM. 2. The DNA transfection mixture (2 ml) is added gently to each well and incubated at 37◦ C for 5–6 h. After this, regular culture medium (with serum) is used, supplemented with the appropriate antibiotics for which the clones to be selected are engineered to be resistant.

3.1.3. Validation by Immunofluorescence

In order to validate protein expression, cells are stained using a specific antibody that recognizes the fused epitope (e.g., anti-V5 anti-HA, etc.). The protocol below can be modified and optimized depending on the antibodies and cells (see Note 4). Cells are grown on eight-well glass chamber slides (Lab-Tek II, Nalgc Nunc, USA) overnight. The volumes may need to be adjusted for different plates. 1. The cell culture medium is removed; cells are washed with PBS and fixed with cold acetone for 10 min at –20◦ C. 2. After air-drying for 15 min, cells are washed twice for 5 min with TBST (0.2 ml/wash). 3. Next, 0.1 ml of blocking solution (PBS containing serum) is added and the mixture is incubated for 60 min at room temperature to reduce non-specific binding of antibody. 4. The blocking solution is removed and 0.1 ml of PBS is added (containing serum), with the appropriate first antibody (1:100 to 1:400 dilution of antibody), followed by incubation for 60 min at room temperature or overnight at 4◦ C.

CEST MRI Reporter Genes

277

5. Cells are washed 3 × 5 min with PBS and then incubated in 0.1 ml of PBS containing the appropriate secondary antibody conjugated to a fluorescent dye or enzyme for 60 min at room temperature in the dark. 6. Cells are washed three times for 5 min with PBS and counter-stained. 7. Cells are embedded with mounting media, cover-slipped, and screened with a microscope. 3.1.4. Protein Extraction

As CEST reporter genes are extremely sensitive, the changes in pH and the exchange rate are dramatically reduced when the cells are fixed. The best way to measure the CEST contrast, in vitro, is to extract the proteins. To this end, the cells are washed twice and collected in ice-cold 10-mM PBS (pH=7.1, without Mg2+ /Ca2+ ). After centrifugation at 2,500g at 4◦ C for 10 min, R (Mammalian Prothe pellet is suspended in 1 ml of M-PER tein Extraction Reagent PIERCE) and shaken gently at 4◦ C for 10 min. Cell debris is removed by centrifugation at 14,000g at 4◦ C for 15 min. The supernatant is transferred to a dialysis tube (cutoff=3.5 kDa) and dialyzed twice against PBS. A protease inhibitor cocktail is added (e.g., PIERCE; HaltTM ) and the protein extraction is stored at –80◦ C. Protein concentrations are then determined using the Bradford assay (PIERCE).

3.1.5. Transplantation

Animal procedures should be conducted in accordance with the guidelines for the care and use of research animals. The cells are collected and suspended in a low volume of PBS or saline. The cells are inoculated into the target tissue while the animal is kept anesthetized by isoflurane inhalation (1–2%).

3.2. General MRI Protocol 3.2.1. Localizer

1. In vivo MR imaging should be performed several days after cell transplantation. The animals should be restrained using a holder, centering at both the center of the RF coil and the center of the magnet, and kept anesthetized using 1–2% isoflurane gas inhalation throughout the imaging procedure. 2. After the animal is positioned appropriately so that the region of interest is set as close to the magnetic field and transmitter RF coil isocenter as possible, providing optimal image quality and minimal B1 and B0 inhomogeneities, tuning and matching settings must be manually adjusted in order for the scanner to be in a resonance capable mode. 3. A localizer sequence (tripilot RARE, MSME, or FLASH on Bruker small animal scanners) is obtained, with large FOVs to acquire three images along the XY, YZ, and XZ planes.

278

Liu, Bulte, and Gilad

4. The optimal MR acquisition parameters are determined, including B0 homogeneity, resonance frequency offset, transmit gain for 90 and 180◦ flip angles, and receiving gain. This can be achieved by using automatic global parameter toolboxes provided by the manufacturer. 5. T2-w anatomical images are then acquired, using a multipleslice fast SE sequence (i.e., RARE), and the subsequent region of interest for single-slice CEST imaging should be defined. The following typical parameters can be used: acquisition bandwidth = 50 kHz, 15 slices; 0.7-mm slice thickness, TE as short as possible (∼4–5 ms), TR = 1,000 ms, RARE factor (echo number) = 16, FOV=20 mm × 20 mm, and 128 × 64 matrix size. 6. The desired slice for the CEST experiment should be chosen, based on the anatomic indicators, such as needle tract, distance to the center, or T1 /T2 contrast. Due to the fact that CEST imaging employs a long saturation pulse, a singleslice approach is typically preferred in a CEST acquisition to save time. The acquisition is then repeated using the defined single-slice geometry to ensure that the correct region of interest (ROI) will be sufficiently covered in the following CEST imaging section. 3.3. CEST Imaging and Z-Spectra Acquisition 3.3.1. Cest Imaging

Typical RARE imaging parameters used are as follows: acquisition bandwidth = 50 kHz, slice thickness = 0.7 mm, TE = 5 ms, TR = 5,000 ms, RARE factor = 8, FOV= 2 cm × 2 cm, matrix size =128 × 64, continuous wave (CW) pulse = 2,000 ms, B1 strength=3.6 microT (153 Hz), and number of averages (NA) = 2. The overall acquisition time is approximately 80 s.

4. Notes 1. The pair of restriction enzymes, BamH I and bgl II, has compatible ends. While BamH I cuts after the first G of the sequence (GGATCC), Bgl II cuts after the first A in the sequence (AGATCA). Therefore, after ligation, the newly formed sequence is GGATCA if BamH I is located in the 3 end or AGATCA if Bgl II is in the 3 end. In both cases, the new sequence cannot be cleaved any further, which is required for elongation of the gene. It is critical to make sure

CEST MRI Reporter Genes

279

that at least one of these enzymes does not cut anywhere in the expression vector. If both enzymes cut, there are other pairs of restriction enzymes that can be used. 2. A reporter encodes to a longer protein will give a higher CEST signal since the signal is proportional to the number of the exchangeable NH protons. However, since the gene is constructed from repetitive elements, cloning and amplifying the plasmid in bacteria become more and more complicated with each cloning cycle. One way to address this problem is to grow plasmid Escherichia coli strands, which are designed for cloning direct repeats, such as Stbl3 (Invitrogen). An alternative way is to use a commercially available full-length synthetic gene (e.g., Blue Heron Biotechnology, Bothell, WA). 3. Varying the concentrations of DNA and transfection reagent can optimize the transfection efficiency. It is important not to disturb the transfection complexes (DNA and reagent) by mixing or pipetting. 4. If the protein is rich with lysine residues, then the isoelectric point (pI) will be too high to perform immunoblotting. In this case, immunofluorescence is the method of choice, which also provides information about the spatial cellular localization of the proteins.

Acknowledgments The protocols described in this chapter were developed through funding with NIH Roadmap R21 EB005252 to J.W.M.B. References 1. Serganova, I., Ponomarev, V., Blasberg, R. Human reporter genes: Potential use in clinical studies. Nucl Med Biol 2007;34:791–807. 2. Koretsky, A. P., Brosnan, M. J., Chen, L. H., Chen, J. D., Van Dyke, T. NMR detection of creatine kinase expressed in liver of transgenic mice: Determination of free ADP levels. Proc Natl Acad Sci USA 1990;87:3112–3116. 3. Walter, G., Barton, E. R., Sweeney, H. L. Noninvasive measurement of gene expression in skeletal muscle. Proc Natl Acad Sci USA 2000;97:5151–5155. 4. Alfke, H., Stoppler, H., Nocken, F. et al. In vitro MR imaging of regulated gene expression. Radiology 2003;228:488–492.

5. Weissleder, R., Moore, A., Mahmood, U. et al. In vivo magnetic resonance imaging of transgene expression. Nat Med 2000;6: 351–355. 6. Cohen, B., Ziv, K., Plaks, V. et al. MRI detection of transcriptional regulation of gene expression in transgenic mice. Nat Med 2007;13:498–503. 7. Genove, G., Demarco, U., Xu, H., Goins, W. F., Ahrens, E. T. A new transgene reporter for in vivo magnetic resonance imaging. Nat Med 2005;11:450–454. 8. Kodibagkar, V. D., Yu, J., Liu, L., Hetherington, H. P., Mason, R. P. Imaging beta-galactosidase activity using (19)f chem-

280

9.

10.

11.

12.

13.

Liu, Bulte, and Gilad ical shift imaging of lacz gene-reporter molecule 2-fluoro-4-nitrophenol-beta-Dgalactopyranoside. Magn Reson Imaging 2006;24:959–962. Louie, A. Y., Huber, M. M., Ahrens, E. T. et al. In vivo visualization of gene expression using magnetic resonance imaging. Nat Biotechnol 2000;18: 321–325. Bengtsson, N. E., Brown, G., Scott, E. W., Walter, G. A. Lacz as a genetic reporter for real-time MRI. Magn Reson Med 2010;63:745–753. Perez-Torres, C. J., Massaad, C. A., Hilsenbeck, S. G., Serrano, F., Pautler, R. G. In vitro and in vivo magnetic resonance imaging (MRI) detection of GFP through magnetization transfer contrast (MTC). Neuroimage 2010;50:375–382. Goffeney, N., Bulte, J. W., Duyn, J., Bryant, L. H., Jr., van Zijl, P. C. Sensitive NMR detection of cationic-polymer-based gene delivery systems using saturation transfer via proton exchange. J Am Chem Soc 2001;123: 8628–8629. McMahon, M., Gilad, A., Zhou, J., Sun, P., Bulte, J., van Zijl, P. Quantifying exchange rates in chemical exchange saturation transfer agents using the saturation time and saturation power dependencies of the magnetization transfer effect on the magnetic resonance imaging signal (QUEST and QUESP): ph calibration for poly-L-lysine and a star-

14.

15.

16.

17.

18.

19.

20.

burst dendrimer. Magn Reson Med 2006;55: 836–847. Gilad, A., McMahon, M., Walczak, P. et al. Artificial reporter gene providing MRI contrast based on proton exchange. Nat Biotechnol 2007;25:217–219. Gilad, A. A., Winnard, P. T., Jr., van Zijl, P. C., Bulte, J. W. Developing MR reporter genes: Promises and pitfalls. Nmr Biomed 2007;20:275–290. Gilad, A. A., Ziv, K., McMahon, M. T., van Zijl, P. C., Neeman, M., Bulte, J. W. MRI reporter genes. J Nucl Med 2008;49:1905– 1908. Ward, K. M., Aletras, A. H., Balaban, R. S. A new class of contrast agents for MRI based on proton chemical exchange dependent saturation transfer (CEST). J Magn Reson 2000;143:79–87. Zhou, J., van Zijl, P. C. Chemical exchange saturation transfer imaging and spectroscopy. Prog Nucl Magn Reson Spectrosc 2006;48:109–136. Sherry, A. D., Woods, M. Chemical exchange saturation transfer contrast agents for magnetic resonance imaging. Annu Rev Biomed Eng 2008;10:391–411. McMahon, M. T., Gilad, A. A., DeLiso, M. A., Berman, S. M., Bulte, J. W., van Zijl, P. C. New “multicolor” polypeptide diamagnetic chemical exchange saturation transfer (DIACEST) contrast agents for MRI. Magn Reson Med 2008;60:803–812.

Chapter 14 Longitudinal Functional Magnetic Resonance Imaging in Animal Models Afonso C. Silva, Junjie V. Liu, Yoshiyuki Hirano, Renata F. Leoni, Hellmut Merkle, Julie B. Mackel, Xian Feng Zhang, George C. Nascimento, and Bojana Stefanovic Abstract Functional magnetic resonance imaging (fMRI) has had an essential role in furthering our understanding of brain physiology and function. fMRI techniques are nowadays widely applied in neuroscience research, as well as in translational and clinical studies. The use of animal models in fMRI studies has been fundamental in helping elucidate the mechanisms of cerebral blood-flow regulation, and in the exploration of basic neuroscience questions, such as the mechanisms of perception, behavior, and cognition. Because animals are inherently non-compliant, most fMRI performed to date have required the use of anesthesia, which interferes with brain function and compromises interpretability and applicability of results to our understanding of human brain function. An alternative approach that eliminates the need for anesthesia involves training the animal to tolerate physical restraint during the data acquisition. In the present chapter, we review these two different approaches to obtaining fMRI data from animal models, with a specific focus on the acquisition of longitudinal data from the same subjects. Key words: Anesthesia, awake, brain, BOLD, cerebral blood flow, cerebral blood volume, neurovascular coupling, non-human primates, rodents, songbirds.

1. Introduction Functional magnetic resonance imaging (fMRI) has made a remarkable impact on brain research, establishing itself as the most prominent research tool in cognitive neuroscience (1, 2) and showing great promise in translational and clinical studies (3). fMRI relies on the neurovascular coupling, a tight relationship between changes in neural activity and local regulation of M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_14, © Springer Science+Business Media, LLC 2011

281

282

Silva et al.

cerebral blood flow (CBF), cerebral blood volume (CBV), and oxygen consumption (CMRO2 ) (4). Functional MRI provides excellent contrast-to-noise ratio, sub-millimeter spatial resolution, coverage of the whole brain, and relative ease of implementation. On the other hand, the temporal resolution of fMRI is relatively low (particularly with respect to the time scale of neuronal events) and the underlying fMRI signal mechanism (5) and its functional specificity (6) are still a subject of research. The use of animal models has been essential to the development of fMRI techniques. Rodent models were the subjects in the first studies employing the blood oxygenation-level-dependent (BOLD) contrast (7, 8), as well as the first CBF measurements using exogenous MRI tracers, including deuterium (9, 10), fluorine (11, 12), and gadolinium chelates (13, 14), which also allowed the estimation of CBV in addition to CBF. The advantages of using an endogenous tracer led to the development of arterial spin labeling (ASL) techniques to quantify CBF noninvasively, with the first demonstrations once again occurring in rodents (15, 16). Animal models of functional brain activation have been widely employed to address issues related to spatial localization of the functional signals, the magnitude of signal changes as a function of stimulation parameters, as well as temporal aspects of the hemodynamic response, giving critical insight into the brain’s physiology and function (17). Furthermore, the use of animal models in fMRI has been particularly advantageous in preclinical and translational studies of various models of brain disease. A major practical issue related to performing fMRI in animals relates to compliance. The MRI environment poses stringent restrictions on subject positioning and movement, and typical fMRI studies last from a few to several hours. Notwithstanding the duration of the experiments, because animals are inherently non-compliant, most fMRI performed to date have required the use of anesthesia, which offers the important advantages of ensuring compliance, minimizing movement, and alleviating stress, at the expense of requiring the investigator to monitor and control the systemic physiological status of the animal. Another major disadvantage of the use of anesthesia is that it interferes with both neural activity and neurovascular coupling, thus compromising interpretability and applicability of results to the understanding of human brain function. An alternative choice to the use of anesthesia is to acclimate, condition, and train the animal to tolerate physical restraint during the data acquisition. This approach offers the advantage of minimizing the need for physiological monitoring and maintenance. On the other hand, it is difficult to dynamically monitor the level of stress in awake animals and virtually impossible to establish its absence, even with extensive training.

Longitudinal fMRI in Animal Models

283

Stress can be a significant confound in the study of brain function, particularly when hemodynamic variables are used as surrogate markers of neural activity. Indeed, continued research on training protocols and evaluation of stress indicators is a subject of intense research (18). In the present chapter, we review these two different approaches to obtaining fMRI data from animal models, with a specific focus on the acquisition of longitudinal data from the same subjects.

2. fMRI of Anesthetized Animal Models: From Terminal to Longitudinal Preparations 2.1. Anesthetics for Terminal Experiments

Urethane (19–23) and α-chloralose (24–36) are the most widely used anesthetics in fMRI studies of rodents. A major advantage of both anesthetics is that they preserve the neurovascular coupling (37). However, both substances are toxic and thus not adequate for longitudinal use, being used only in terminal preparations (38, 39). The use of anesthesia makes the animal lose the ability to regulate its own physiology and body temperature, forcing the investigator to monitor and maintain the vital signs of the animal to ensure stable physiology throughout the entire experiment. Therefore, surgical preparation is required. This consists of (a) induction of anesthesia, (b) oral intubation or tracheostomy, and (c) placement of intravenous, intra-arterial, and intraperitoneal catheters. Figure 14.1 shows a typical surgical preparation station and all associated equipment. Fifteen to thirty minutes prior to induction of anesthesia, it is useful to treat the animal with atropine sulfate (Vedco Inc, Saint Joseph, MO), given subcutaneously or intramuscularly at a single dose of 0.5 mg/kg, to decrease bronchial secretions and salivation during anesthesia, and as an anesthesia adjuvant. Anesthesia induction in rodents usually uses halogenated anesthetics (halothane, sevoflurane, and isoflurane). In contrast, intramuscular bolus of ketamine (either alone, or in combination with its potentiators xylazine, acepromazine or medetomidine hydrochloride) is frequently used for induction in both cats and non-human primates. Following the anesthesia induction, the animal is typically orally intubated or tracheostomized and mechanically ventilated to allow dynamic adjustments of respiratory parameters

284

Silva et al.

Fig. 14.1. Typical surgical station for preparation of anesthetized animals. a Medical gases and vacuum scavenger; b gas flowmeters and mixer; c isoflurane vaporizer and compliance balloon; d small animal ventilator; e pulse oximeter/capnograph; f temperature regulator, heated water bath, and water mat; g surgical microscope.

as to ensure stable physiology – under the effect of the anesthetic – throughout the experiment. It is extremely important to monitor and control the animal’s vital signs. The body temperature must be monitored via a rectal temperature probe, and a heated water pad or another suitable source of heat must be used to keep the body temperature at normal values (e.g., 37.5 ± 0.5 ◦ C for rats; 38.5 ± 0.5 ◦ C for cats and marmosets). Pulse oximetry and capnography, when available, can be started at this point. These techniques provide a good assessment of the quality of the animal’s respiration, giving the investigator the necessary information to adjust the breathing gas mixture and the ventilator parameters to ensure good oxygenation and proper ventilation rate and volume to the animal. Following the start of mechanical ventilation, catheters are placed into the femoral artery, the femoral vein (or the lateral tail vein), and the intraperitoneal cavity. These catheters allow the administration of anesthesia and other drugs that control the physiological status of the animal, the measurement of arterial blood pressure via a pressure sensor, and the sampling of arterial blood gases. Once all surgical procedures are completed and all open wounds are sutured, the inhalational anesthetic can be

Longitudinal fMRI in Animal Models

285

discontinued and the anesthesia switched to either α-chloralose (α-chloralose, product #C8091, Sigma-Aldrich, St. Louis, MO) or urethane (urethane, product #U2500, Sigma-Aldrich). Typical dosage for α-chloralose is 80 mg/kg initial IV bolus (37) followed by 27 mg/kg·h constant IV infusion (33). For urethane, the typical dosage is 1.25 g/kg single dose IP (23), as urethane is known to be a long-acting (8–10 h) anesthetic. Both α-chloralose and urethane can be dissolved in slightly warmed phosphate-buffered saline (PBS) at a concentration of 10 mg/ml or 100 mg/ml, respectively. However, urethane is a known carcinogen and should only be manipulated in the hood using gloves. A muscular relaxant such as pancuronium bromide (Pavulon, 2 mg/ml, Teva Pharmaceuticals USA, North Wales, PA) may be periodically administered at a dose of 2 mg/kg/hr IV or IP to aid with immobilization. 2.2. Anesthetics for Longitudinal Experiments

Due to the extensive surgical preparation described above and also to the toxicity and adverse side effects of both α-chloralose and urethane, different anesthetics need to be used for longitudinal studies. These studies require minimal interventions on the animal so that its overall physiological state is stable throughout the duration of the study and especially each time the animal is tested. Because of this, inhalational compounds such as halothane (30) and isoflurane (40–48) are attractive in that they can be administered via a face mask to the animal, thus obviating catheterizations for vascular access. Furthermore, these agents provide stable physiology and allow for easy control of the plane of anesthesia and quick and smooth recovery of the animal upon withdrawal of the anesthetic. However, while both anesthetics are safe to use in repetitive studies in the same animal, they suppress neuronal activity (49) and cerebral metabolism (50) and affect both cerebrovascular tone as well as cerebrovascular reactivity (51), thus greatly influencing the cerebrovascular coupling (52) and requiring optimization of the stimulus parameters to produce robust activation (46). In fact, the influence of halogenated inhalational anesthetics on neurovascular coupling and cerebrovascular regulation can be long lasting. A recent study on the long-term effects of a short hypoxic episode on CBF regulation in isoflurane-anesthetized rats showed a dramatic reduction in the CBF response to hypoxia in animals that were exposed to isoflurane 5 days earlier (53). As an alternative to inhalational anesthetics, injectable agents such as propofol and medetomidine are attractive in providing satisfactory depth of anesthesia, quick onset of action, and smooth recovery of the animals. Propofol is an injectable anesthetic with rapid mechanism of action that is increasingly used in fMRI experiments in animal models (40, 54–57). The depth of anesthesia under propofol can be readily adjusted by varying the rate

286

Silva et al.

of infusion, and the animals quickly recover at the end of the experiment, thus facilitating longitudinal studies. Another agent that has been recently proposed as suitable for repetitive studies is the α2-adrenoreceptor agonist medetomidine hydrochloride (22, 58–61), which has been shown to allow robust fMRI responses (58, 59) and the measurement of resting state signal fluctuations in the brain (59–61). The sedative and analgesic effects of medetomidine can be quickly reversed with the application of atipamezole hydrochloride. However, some limitations of prolonged exposures to medetomidine include a gradual rise in blood pressure and heart rate, lower pO2 values, and a more difficult control of the plane of anesthesia (61). When using injectable, intravenous anesthetics, special attention must be given to the total infusion volume in relation to the total blood volume, especially in small animals such as rodents and small non-human primates, so that the infusion does not change blood chemistry. For example, the dose of propofol (10 mg/ml) required to maintain immobilization in rats and marmosets is 0.5–1.0 mg/kg/min (3–6 ml/kg/hr), which is about five times higher than that required in humans (0.05–0.2 mg/kg/min). Yet the total blood volume of a 400-g rat is ∼25 ml (62), circa 1/200 of the total human blood volume. Thus, intravenous anesthetics have a much greater potential to disturb blood chemistry in small animals than in humans, and drug preparations should be chosen carefully. Propofol, as made by certain manufacturers, has a pH value as low as 4, posing a significant risk of inducing acidosis in small animals, especially in prolonged and/or longitudinal experiments. Like in clinical medicine, concurrent intravenous infusion of lactated Ringer solution with the anesthetics may partially alleviate the detrimental effects of blood chemical disturbance. 2.3. Placement of the Animal in an MRI-Compatible Bed

Magnetic resonance imaging poses stringent requirements on immobilization. After surgery, the animals are placed in a stereotaxic-like head holder and strapped to an MRI-compatible bed. A number of beds are available from the major vendors of small animal MRI scanners, such as Bruker (BrukerBiospin, Corp., Ettlingen, Germany) or Varian (Varian, Inc., Palo Alto, CA), or from third-party vendors (e.g., Rapid Biomedical, GmbH, Rimpar, Germany; Ekam Imaging, Inc., Shrewsbury, MA). Figure 14.2 shows a picture of the bed that we have used for fMRI of small animals. The bed is an essential part of the experiment, as it integrates, in a single platform, a resting place for the animal, the stereotaxic head holder, the physiological maintenance and monitoring devices, a stage for receive RF coils and preamplifiers, and another stage for the functional stimulation devices. Therefore, significant effort needs to be put on the design, fabrication, and adaptation of an animal bed to include

Longitudinal fMRI in Animal Models

287

Fig. 14.2. a MRI-compatible animal bed, containing a stereotaxic head holder with ear pieces and a bite bar (b), to which the head of the animal is secured. Mechanical ventilation is provided by the gas lines. c Pulse oximetry sensor and rectal temperature probe. d Heated water mat (shown unfolded), which can be wrapped around the animal’s body to maintain temperature. e Multi-channel receive RF coils. f Multi-channel RF preamplifiers.

and integrate all the features necessary to the successful execution of the experiment. 2.4. Physiological Monitoring

As mentioned above, the use of anesthesia makes the animal lose the ability to regulate its own physiology and body temperature, and it is the responsibility of the investigator to monitor and maintain the vital signs of the animal throughout the remainder of the experiment. As soon as the animal has been moved to the MRI-compatible bed, physiological monitoring must start – or resume if started during animal preparation. All essential physiological parameters, including, but not limited to the list below, should be monitored and maintained at normal values. 1. Rectal temperature: via temperature probe inserted in rectum. This is one of the most critical parameters to be monitored and maintained, as anesthesia impedes the animal’s own temperature regulation. Normal rectal temperature values for rats are 37.5 ± 0.5◦ C, and 38.5 ± 0.5◦ C for cats and marmosets. 2. Mean arterial blood pressure (MABP): via pressure transducer hooked up to femoral arterial line, if available, or via pressure cuff wrapped around the tail or thigh. Typical MABP values in rats are 110–120 mmHg (63, 64). Typical MABP values in conscious marmosets are 100–110 mmHg (65). 3. Pulse oximetry (SPO2 ): via transducer placed on forelimb or hindlimb. Usually SPO2 values remain above 90%.

288

Silva et al.

4. Heart rate (HR): either derived from MABP trace or reported by pulse oximeter. For rats under α-chloralose, HR typically rises above 300 BPM but stays below that value under other anesthetics. For marmosets under propofol, HR typically decreases from above 350 BPM in the conscious, awake condition to 150–250 BPM in anesthetized animals. 5. End-tidal CO2 (ETCO2 ): via micro-capnometer hooked up to face mask or to the respiratory line. The distance between the mechanical ventilator and the MRI magnet forces the use of long respiration lines, resulting in substantial mixing between expired and recirculated gas. Therefore, the microcapnometer tends to display significantly attenuated values of ETCO2 , especially in small animals. Yet, these values are valid relatives and a significant correlation exists between ETCO2 and PaCO2 , as shown in Fig. 14.3. 6. Respiratory pressure: monitored via pressure transducer in mechanical ventilator. It is important to check the respiratory compliance of the intubated or tracheostomized animal throughout the experiment. The end-inspiratory pressure can be set by adjusting the flow of the air mixture in the ventilator. Typically, end-inspiratory pressures are in the range of 8–12 cm H2 O. 7. Arterial blood gases, including pH, PaCO2 , and PaO2 : sampled from the femoral artery, when available. Even though

Fig. 14.3. Pooled data plot of ETCO2 as measured with a capnograph versus PaCO2 sampled from arterial blood (n= 34 rats, average of six points per animal) in normocapnia and hypercapnia. The dashed line is the line of identity, and the correlation coefficient between ETCO2 and PaCO2 is 0.77 (r 2 = 0.59). Even though the absolute value of ETCO2 is influenced by the size of the animal relative to the total flow and volume of air in the ventilator, the length of the gas lines, and the flow of expired air into the capnograph, the significant correlation between ETCO2 and PaCO2 allows ETCO2 values to be used as a relative index of changes in PaCO2 .

Longitudinal fMRI in Animal Models

289

arterial blood sampling in mice is restricted to much fewer samples than in rats, latest-generation blood gas analyzers are able to work with samples as small as 30 μl of blood (typically 60 μl) (e.g., ABL80 FLEX, Radiometer America, Westlake, OH). Periodic sampling of arterial blood is also performed to yield information on arterial blood gases, hematocrit, and electrolytes. Data collected during unstable physiological states are best discarded. Furthermore, large deviations in the arterial blood gases often call for administration of corrective pharmacological agents. All the physiological parameters listed above can be recorded during the experiment using a multi-channel data acquisition system, such as the system MP150 (Biopac Systems, Inc., Goleta, CA). Figure 14.4 shows the physiological monitoring graph of a typical fMRI experiment to measure the BOLD and CBF response to electrical stimulation of the forepaws in a rat anesthetized with α-chloralose. Traces of the arterial blood pressure, respiratory pressure, and heart rate were recorded along with the forepaw stimulation epochs, the EPI acquisition times, and the gradient temperature. Monitoring of the animal’s physiology during the experiment ensures that the data are acquired under stable conditions. For example, no changes in arterial blood pressure are noticed during the forepaw stimulation epochs. Another advantage of acquiring physiological parameters is the ability to do retrospective correction of the influence of either the respiratory or cardiac cycles on the fMRI signal (66).

Fig. 14.4. Physiological and fMRI data acquisition traces recorded during a typical fMRI experiment of electrical stimulation of both forepaws of a rat anesthetized with α-chloralose. The experiment lasted 4 min. From top to bottom, the traces show the arterial blood pressure (red), respiratory pressure (green), EPI acquisition tics (blue), stimulation of the left (green) and the right (blue) forepaws, the MRI gradient temperature (magenta), and the heart rate derived from the ABP trace (blue).

290

Silva et al.

2.5. Recovery from Anesthesia

3. fMRI of Conscious, Awake Animals

In longitudinal experiments in which artificial ventilation is used, the animal’s autonomous control of the respiration may be suppressed even after anesthetics are withdrawn. Thus, recovery from anesthesia needs to be closely monitored, and emergency procedures should be planned in advance. Throughout the recovery, artificial ventilation and rectal temperature monitoring should continue, and the intravenous infusion site should remain patent, until the animal is clearly able to breathe on its own after being disconnected from the breathing circuit. Antidotes for some anesthetics are available, such as naloxone for opioids and atipamezole for medetomidine chloride, and should be administered. In the event of respiratory and/or cardiovascular suspension after removal of intubation tube, a respiratory stimulant (e.g., doxapram) and sympathetic stimulant (e.g., epinephrine) can be administered through the intravenous infusion tube, which should not be removed from the animal until it is fully alert.

The use of anesthesia for MRI and fMRI studies in animal models has the advantages of effectively ensuring compliance and of minimizing stress via sedation. However, any anesthetic interferes, in different ways, with neural activity and cerebrovascular reactivity, representing a complex confound to the interpretability and applicability of the obtained data to the understanding of human brain function. The alternative to the use of anesthesia as a means of ensuring compliance is to condition the conscious, awake animal, to tolerate the rigid head restraint required to allow the acquisition of good quality data with acceptable levels of motion artifacts. The use of conscious, awake animals in the MRI setting is becoming increasingly popular, as exemplified by studies performed in rodents (18, 29, 42, 55, 67–71) and monkeys (71–84). To provide effective, yet relaxed restraint to the animals, the animal bed needs to be designed taking into account comfortable support for their head and body (77, 85). Equally important is the development of acclimatization and training procedures to condition the animal to tolerate long periods of restraint with minimal stress (18), as a stressed-out animal is as useful as an unresponsive one! In this section, we describe our own experience with obtaining longitudinal fMRI data from conscious, awake marmosets in a 7T horizontal MRI scanner. Traditionally, the use of awake, conscious animals in the MRI requires the surgical implantation of head posts that can be rigidly secured by clamps to a specially

Longitudinal fMRI in Animal Models

291

designed frame (85). This approach allows maximum restraint of the animals but has many disadvantages. The head implants typically generate susceptibility artifacts that degrade image quality by introducing geometric distortions and/or signal dropouts in the MR images. In addition, they require constant aseptic cleaning to prevent infections. Moreover, if an infection appears, the animal needs to be treated with antibiotics and anti-inflammatory drugs that may interfere with neurovascular coupling and confound interpretation of the data. Further, the use of a head implant significantly detracts from a major advantage of MRI as a noninvasive technique. An alternative to the use of head implants is to secure the animal to an MRI-compatible stereotaxic head holder by means of ear pieces and a bite bar. To do this, the animals need to be sedated with a short-acting anesthetic, such as medetomidine, so that they can be attached to the stereotaxic head holder, after which the head posts are allowed to wake up by reversal of the anesthesia (e.g., with atipamezole) (18, 42, 70, 80). While it has been shown that animals can be successfully conditioned not to fight the head restraint upon regaining consciousness (18), this approach has the disadvantages of requiring the use of anesthesia and of utilizing a head holder that may potentially hurt the animal due to the presence of the ear bars. Our approach to restrain the head of the animals is different than either of the two above. We opted to eliminate the need for implanting head posts or to administer any anesthetics or sedatives altogether by acclimatizing the animals to being restrained by a custom-fit helmet that was specifically designed to match the contour of each individual head exactly, providing a comfortable, yet effective restraint. 3.1. Design of the Animal Bed and Restraint Device

As shown in Fig. 14.5, the bed for fMRI of awake marmosets consists of a cylindrical tube of inner diameter 111 mm, made out of fiberglass impregnated with an epoxy resin (NEMA grade G10, FPI Industries, Arnold, PA), cut in half length-wise, in which the marmoset lies in the prone, sphinx position. Two lateral support bars made out of Delrin are attached to either inner side of the bed to support the helmet and cover pieces. The bed attaches to the sliding mechanism on one end via the hanger (Fig. 14.5), so as to float cantilevered inside the magnet without touching the gradients or the transmit RF coil. Prior to positioning the marmoset in the cradle, a sleeveless jacket (Lomir Biomedical, Inc. Malone, NY) is placed on the marmoset. Next, a plastic semicylindrical cover made of Lexan is attached to the back of the marmoset’s jacket using plastic cable ties. The marmoset is then gently placed into the cradle and the cover is secured by screwing attached nylon thumb screws into the bars on the cradle. The animal is now loosely but effectively restrained from sliding out of the cradle anteriorly or posteriorly. The arms, legs, and tail of the animal are free to move unimpededly. Additionally, the

292

Silva et al.

Fig. 14.5. Illustration of a restrained marmoset in the MRI-compatible bed. The body of the animal is loosely attached to a back cover via zip ties secured to a sleeveless jacket worn by the animal. The back cover is screwed to the side bars on the cradle, while the arms, legs, and tail of the animal are free to move. The head of the marmoset is secured to a two-piece, custom-built helmet made specifically for that individual. The chin piece on the bottom supports the chin of the animal, and the head piece on the top prevents head motion. Note that the helmet pieces are lined up with foam on the inside to provide a comfortable support to the entire head. The animal sits in the sphinx position looking out towards the back of the magnet. The bed is secured to the bed sliding mechanism on one end via the hanger.

semi-cylindrical cover is a few centimeters from the marmoset’s back, allowing the animal to stretch and to adjust its body position as needed for comfort, as shown in Fig. 14.5. Because the shape and size of marmosets’ skulls vary, individual custom-built helmets are designed for each animal. For this, a 3D spin-echo MRI of the entire head and neck is acquired from each animal. Next, a 3D surface-rendering algorithm is applied to obtain the contour of the head, which is then fed into Rhinoceros 3d (McNeel North America, Seattle, WA), a 3D modeling program, to design the top (head) and bottom (chin) helmet pieces. After the head and chin helmet pieces are designed (see Fig. 14.6), they are sent to a 3D printer (ProJet HD3000, 3D Systems Corp., Rock Hill, SC), which builds the helmets, layer by

Longitudinal fMRI in Animal Models

293

Fig. 14.6. Detail of the construction of the custom-fit helmet. Based on a 3D MRI of the entire head of the marmoset, a 3D model of the helmet is created consisting of two pieces: the chin piece (blue) to support the chin of the animal and the head piece (red) that supports the head and prevents motion. Once the model is created, it is sent to a 3D plastic printer that creates the helmet.

layer, from liquid ABS plastic that hardens during the manufacturing process. After the helmet printing is complete, 3-mm thick foam is glued onto the inside surface of both top and bottom pieces to provide greater comfort for the animal. Proper and perfect fitting of the pieces to each animal is guaranteed by design. Since the shape of each animal is used to produce a helmet manufactured specifically for its head, we can ensure that the animal is comfortable yet immobilized. After the animal is lowered into the cradle and secured by the cover piece, the helmet top and bottom pieces are carefully placed around the animal’s head and screwed into the bars. 3.2. Acclimatization of the Animal to Body and Head Restraint

Once the individualized helmet pieces are built, the animal needs to be acclimatized to the bed and to physical restraint during the MRI sessions. This acclimatization procedure consists of three phases, as illustrated in Fig. 14.7: 1. Phase 1: Acclimatization of Awake Marmosets to Being Contained in the Bed. In this phase, the animals are dressed with the jackets, attached to the back cover by zip ties, and lowered into the bed. The cover is secured to the bed by nylon thumb screws. The bed is then inserted into a mock MRI tube and the animal is observed from a distance via a webcam. For phase 1, training begins with 15 min on day 1, and progresses to an hour by day 4 (see Table 14.1). As a reward, 3–6 cc of infant milk formula (Pediasure) and 3–5 mini-marshmallows are given at the beginning and end of each training session. Prior to, during, and after each of the acclimatization phases, a behavioral assessment takes place to provide a measure of the tolerance of the animals to the acclimatization procedures. The scoring of each animal is performed using

294

Silva et al.

Fig. 14.7. Illustration of the three phases of training. In phase 1 (top), the body of the animal is loosely attached to the MRI bed and the animal is conditioned to staying in the bed for increasing periods of time. In phase 2 (middle), reinforcement of the training in phase continues while the animal gets used to MRI sounds. In phase 3 (bottom), the individualized helmet is introduced to restrain the head of the animal.

Table 14.1 Three-week acclimatization schedule Phase

Monday

Wednesday

Friday

Sunday

Procedures

1

0:15

0:30

0:45

1:00

Jacket

2

1:00

1:20

1:40

2:00

Jacket + MRI sounds

3

1:00

1:20

1:40

2:00

Jacket + MRI sounds + helmet

the Behavioral Assessment Scale shown in Table 14.2 (86). It is our experience that all animals successfully complete the training in phase 1, starting with an average score of 4 on the Behavioral Assessment Scale and acclimatizing to an average score of 2 by the end of phase 1. 2. Phase 2: Acclimatization of Awake Marmosets to Being Contained in the Bed in the Presence of MRI Sounds. In spite of being entirely non-invasive, MRI is, unfortunately, loud. Therefore, it is necessary that the animals get properly acclimatized to the sounds generated by the MRI scanner during imaging. In phase 2, the animals are restrained as in phase 1 for increasing periods (Table 14.1). While in the mock MRI tube, they were allowed to hear the sounds produced by the MRI scanner, played out at a softer level than in a real MRI session. This schedule reinforces the adaptation to the body restraint initiated in phase 1 and conditions the animals to ignore the MRI sounds produced by

Longitudinal fMRI in Animal Models

295

Table 14.2 Behavioral assessment scale Score

Behavior

1

Quiet: marmoset calm and relaxed

2

Mostly quiet, agitated only initially

3

Mostly quiet, with brief, intermittent mild agitation

4

Quiet after initial struggle, increasingly agitated over time

5

Mild agitation for about half of the restraint period

6

Moderate agitation during half of the restraint period

7

Restless and agitated during most of the restraint period

8

Extremely agitated during most of the restraint period

the scanner. As in phase 1, Pediasure and mini-marshmallows are used as rewards at the beginning and at the end of each training session. The same Assessment Scale for Behavioral Responses (Table 14.2) is used to assess the response of the marmosets to the restraint device in the presence of MRI sounds. It has been our experience that all animals successfully complete this phase of the training, starting again with an average score between 3 or 4 on the first day and moving down to an average score of 2 or better by the end of phase 2. At this point, if an animal proves difficult to train by systematically scoring above the mean score, the investigator may choose to adapt the exact timeline of the acclimatization procedures to the idiosyncrasies of that individual, or may choose to drop the animal out of the study. 3. Phase 3: Acclimatization of Awake Marmosets to Head Restraint in the Presence of MRI Sounds. In phase 3, marmosets are restrained to the bed by the body cover, as in the previous phases, and then fitted with their custom-built helmets, which are attached to the MRI bed as shown in Figs. 14.5, 14.6, and 14.7. The use of the helmet effectively – yet comfortably – restrains head motion. The ears are fitted with earplugs made of silicon jelly (Insta-Putty, Insta-Mold Products, Inc., Oaks, PA), which are pressed gently into the ear canals by foam pads from both sides of the head. This further restrains head motion and protects the animals from MRI scanner’s noise. In phase 3, the animals are conditioned for increasing periods, as in previous phases (Table 14.1). While in the mock MRI tube, they were allowed to hear the sounds produced by the MRI scanner, played out at the same level of loudness as in a real MRI session (Fig. 14.7). This schedule reinforces the adaptation

296

Silva et al.

to the body restraint of the previous phases and further conditions the animals to ignore the MRI sounds produced by the scanner, while enforcing full head fixation. As in previous phases, the animals’ response to training will be evaluated by the Behavioral Response Scale shown in Table 14.2. It has been our experience that all animals are able to successfully complete this last phase of the training, starting again with an average score between 3 or 4 on the first day and moving down to an average score of 2 or better by the end of phase 3. Out of 14 marmosets trained to date, we only dropped one animal out of the study due to increased agitation during phase 2 of the acclimatization. After the animals have successfully completed the 3-week training program, they proceed to undergoing the actual MRI studies. Continued exposure to the MRI scanner in actual studies fully consolidates the acclimatization, and all animals are tolerant of all procedures after a couple of MRI sessions. Overall, this new method of restraint is completely non-invasive, comfortable for the animals, and of great scientific payoff in eliminating the need to either perform surgery for installation of head-post implants or use sedatives to restrain the animals, allowing any marmoset to be acclimatized to restraint and engaged in longitudinal experiments. The advantages of using awake subjects can be fully appreciated by the high amplitudes and reproducibility of fMRI responses to somatosensory stimuli, as shown in Fig. 14.8. In awake marmosets (Fig. 14.8a), electrical stimulation of one arm by repeated pulses (1.5 mA, 0.4 ms duration, 1–125 Hz) evoked fMRI responses in a series of brain regions including thalamus, caudate, putamen, primary (SI), and secondary (SII) somatosensory cortices. Responses were significant in contralateral thalamus, SI and SII, as well as in ipsilateral thalamus and SII. While the fMRI responses were much stronger on the contralateral side,

Fig. 14.8. Map of fMRI response evoked by electrical stimulation of the left arm of the marmoset. Eight contiguous coronal slices immediately posterior to the anterior commissure (AC) are shown in a representative session in awake state (a) and under propofol anesthesia (b). Robust BOLD responses can be detected in the thalamus (Tha.) and in the primary (SI) and secondary (SII) somatosensory cortex, although the responses are much stronger in the awake than in anesthetized animals.

Longitudinal fMRI in Animal Models

297

occasionally robust ipsilateral responses were also detected in SI. In marmosets anesthetized with propofol (Fig. 14.8b), however, both the spatial extent as well as the fMRI response amplitudes were significantly smaller compared to awake subjects. In particular, responses in ipsilateral thalamus and ipsilateral SI were insignificant in many anesthetized sessions, consistent with previous studies using anesthetized rats. Responses in caudate and putamen were also much weaker, although detectable. Furthermore, the response amplitudes in thalamus and cortical areas in anesthetized subjects were less than half of those in awake subjects. Figure 14.9 shows the fMRI response obtained from the same animal in different MRI sessions in a time span of 10 months. Excellent reproducibility of the amplitude and spatial and

Fig. 14.9. fMRI response obtained from the same animal in different MRI sessions in a time span of 10 months. a T-map demonstrating the main active areas of the brain, including SI, SII, and caudate putamen. b BOLD time-course in response to a 2-s electrical stimulus of both hands (2 mA, 0.3 ms, 64 Hz), obtained at five different times (0 weeks, 3 weeks, 6 weeks, 2 months, and 10 months) post-acclimatization. Excellent reproducibility of the amplitude and temporal characteristics of the BOLD response are achieved.

298

Silva et al.

Table 14.3 The amplitude percent signal changes, standard deviation (SD) of baseline, timeto-peak (TTP), and full width at half maximum (FWHM) of the BOLD response from a representative marmoset undergoing longitudinal fMRI (2 s stimulus duration, 2 mA, 0.3 ms pulse width, 64 Hz) of the somatosensory cortex Signal increase (%)

First scan

3 weeks

6 weeks

2 months

10 months

Mean ± SD

1.95

2.50

2.69

2.57

2.47

2.44 ± 0.28

SD of baseline (%)

0.08

0.14

0.17

0.17

0.21

0.15 ± 0.08

TTP (s)

3.84

4.41

3.41

3.80

3.78

3.85 ± 0.36

FWHM (s)

3.92

4.70

4.34

4.14

4.55

4.33 ± 0.31

temporal characteristics of the BOLD response were obtained (see Table 14.3), demonstrating that longitudinal fMRI studies can be successfully carried out in awake, behaving animals.

4. Conclusions In the present chapter, we have presented and reviewed different methodological approaches to obtaining fMRI data from animal models in longitudinal studies. When opting for which approach to follow, the reader is confronted with a balance of advantages and disadvantages that may not always point to the same choice, depending on the purpose of the study. The use of anesthesia has the advantages of ensuring compliance, minimizing movement, and alleviating stress in an effective way. However, anesthesia suppresses many homeostatic pathways and requires the investigator to monitor and control the physiology of the animal. Another major disadvantage of the use of anesthesia is that it interferes with neural activity and neurovascular coupling, severely restricting the choice of the functional paradigm and compromising the interpretability and applicability of the data to the understanding of human brain function. An alternative choice to the use of anesthesia is to acclimate, condition, and train the animal to tolerate physical restraint during the data acquisition. This approach offers the advantage of minimizing the need for physiological monitoring and maintenance but must be done in a well-devised and gradual manner so as to eliminate, as much as possible, stress of the animal. Stress can introduce as much of a confound to the data as anesthesia, and thus the task of acclimatizing animals to awake studies cannot be taken lightly. Nevertheless, when performed consistently and correctly, acclimatization can be very effective in condition-

Longitudinal fMRI in Animal Models

299

ing the animal to participate cooperatively in the study, producing results that, in principle, are experimentally as close as possible to mimicking the setup in human studies.

Acknowledgments This work was supported by the Intramural Research Program of the NIH, NINDS (Alan P. Koretsky, Scientific Director). References 1. Poldrack, R. A. The role of fMRI in cognitive neuroscience: Where do we stand? Curr Opin Neurobiol 2008;18:223–227. 2. Dolan, R. J. Neuroimaging of cognition: Past, present, and future. Neuron 2008; 60:496–502. 3. Matthews, P. M., Honey, G. D., Bullmore, E. T. Applications of fMRI in translational medicine and clinical practice. Nat Rev Neurosci 2006;7:732–744. 4. Attwell, D., Iadecola, C. The neural basis of functional brain imaging signals. Trends Neurosci 2002;25:621–625. 5. Logothetis, N. K. The underpinnings of the BOLD functional magnetic resonance imaging signal. J Neurosci 2003;23:3963–3971. 6. Logothetis, N. K. What we can do and what we cannot do with fMRI. Nature 2008;453:869–878. 7. Ogawa, S., Lee, T. M., Kay, A. R., Tank, D. W. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990;87: 9868–9872. 8. Ogawa, S., Lee, T. M. Magnetic resonance imaging of blood vessels at high fields: In vivo and in vitro measurements and image simulation. Magn Reson Med 1990;16:9–18. 9. Kim, S. G., Ackerman, J. J. Quantification of regional blood flow by monitoring of exogenous tracer via nuclear magnetic resonance spectroscopy. Magn Reson Med 1990;14:266–282. 10. Detre, J. A., Eskey, C. J., Koretsky, A. P. Measurement of cerebral blood flow in rat brain by 19F-NMR detection of trifluoromethane washout. Magn Reson Med 1990;15:45–57. 11. Detre, J. A., Williams, D. S., Koretsky, A. P. Nuclear magnetic resonance determination of flow, lactate, and phosphate metabolites

12.

13.

14.

15.

16. 17.

18.

during amphetamine stimulation of the rat brain. NMR Biomed 1990;3:272–278. Barranco, D., Sutton, L. N., Florin, S., Greenberg, J., Sinnwell, T., Ligeti, L., McLaughlin, A. C. Use of 19F NMR spectroscopy for measurement of cerebral blood flow: A comparative study using microspheres. J Cereb Blood Flow Metab 1989;9:886–891. Villringer, A., Rosen, B. R., Belliveau, J. W., Ackerman, J. L., Lauffer, R. B., Buxton, R. B., Chao, Y. S., Wedeen, V. J., Brady, T. J. Dynamic imaging with lanthanide chelates in normal brain: Contrast due to magnetic susceptibility effects. Magn Reson Med 1988;6:164–174. Rosen, B. R., Belliveau, J. W., Buchbinder, B. R., McKinstry, R. C., Porkka, L. M., Kennedy, D. N., Neuder, M. S., Fisel, C. R., Aronen, H. J., Kwong, K. K. Contrast agents and cerebral hemodynamics. Magn Reson Med 1991;19:285–292. Williams, D. S., Detre, J. A., Leigh, J. S., Koretsky, A. P. Magnetic resonance imaging of perfusion using spin inversion of arterial water. Proc Natl Acad Sci USA 1992;89:212–216. Detre, J. A., Leigh, J. S., Williams, D. S., Koretsky, A. P. Perfusion imaging. Magn Reson Med 1992;23:37–45. Van der Linden, A., Van, C. N., RamosCabrer, P., Hoehn, M. Current status of functional MRI on small animals: Application to physiology, pathophysiology, and cognition. NMR Biomed 2007;20:522–545. King, J. A., Garelick, T. S., Brevard, M. E., Chen, W., Messenger, T. L., Duong, T. Q., Ferris, C. F. Procedure for minimizing stress for fMRI studies in conscious rats. J Neurosci Methods 2005;148:154–160.

300

Silva et al.

19. Lowe, A. S., Williams, S. C., Symms, M. R., Stolerman, I. P., Shoaib, M. Functional magnetic resonance neuroimaging of drug dependence: Naloxone-precipitated morphine withdrawal. Neuroimage 2002;17:902–910. 20. Wu, G., Luo, F., Li, Z., Zhao, X., Li, S. J. Transient relationships among BOLD, CBV, and CBF changes in rat brain as detected by functional MRI. Magn Reson Med 2002;48:987–993. 21. Kannurpatti, S. S., Biswal, B. B. Effect of anesthesia on CBF, MAP and fMRIBOLD signal in response to apnea. Brain Res 2004;1011:141–147. 22. Boumans, T., Theunissen, F. E., Poirier, C., Van der Linden, A. Neural representation of spectral and temporal features of song in the auditory forebrain of zebra finches as revealed by functional MRI. Eur J Neurosci 2007;26:2613–2626. 23. Huttunen, J. K., Grohn, O., Penttonen, M. Coupling between simultaneously recorded BOLD response and neuronal activity in the rat somatosensory cortex. Neuroimage 2008;39:775–785. 24. Hyder, F., Behar, K. L., Martin, M. A., Blamire, A. M., Shulman, R. G. Dynamic magnetic resonance imaging of the rat brain during forepaw stimulation. J Cereb Blood Flow Metab 1994;14:649–655. 25. Kerskens, C. M., Hoehn-Berlage, M., Schmitz, B., Busch, E., Bock, C., Gyngell, M. L., Hossmann, K. A. Ultrafast perfusionweighted MRI of functional brain activation in rats during forepaw stimulation: Comparison with T2 -weighted MRI. NMR Biomed 1996 Feb;9(1) 1996;9:20–23. 26. Gyngell, M. L., Bock, C., Schmitz, B., Hoehn-Berlage, M., Hossmann, K. A. Variation of functional MRI signal in response to frequency of somatosensory stimulation in alpha-chloralose anesthetized rats. Magn Reson Med 1996;36:13–15. 27. Bock, C., Krep, H., Brinker, G., HoehnBerlage, M. Brainmapping of alphachloralose anesthetized rats with T2∗ weighted imaging: Distinction between the representation of the forepaw and hindpaw in the somatosensory cortex. NMR Biomed 1998;11:115–119. 28. Silva, A. C., Lee, S. P., Yang, G., Iadecola, C., Kim, S. G. Simultaneous blood oxygenation level-dependent and cerebral blood flow functional magnetic resonance imaging during forepaw stimulation in the rat. J Cereb Blood Flow Metab 1999;19:871–879. 29. Peeters, R. R., Tindemans, I., De Schutter, E., Van der, L. A. Comparing BOLD fMRI

30.

31.

32.

33.

34.

35.

36.

37.

38. 39.

40.

41.

signal changes in the awake and anesthetized rat during electrical forepaw stimulation. Magn Reson Imaging 2001;19:821–826. Austin, V. C., Blamire, A. M., Allers, K. A., Sharp, T., Styles, P., Matthews, P. M., Sibson, N. R. Confounding effects of anesthesia on functional activation in rodent brain: A study of halothane and alpha-chloralose anesthesia. Neuroimage 2005;24:92–100. Keilholz, S. D., Silva, A. C., Raman, M., Merkle, H., Koretsky, A. P. BOLD and CBV-weighted functional magnetic resonance imaging of the rat somatosensory system. Magn Reson Med 2006;55: 316–324. Yang, J., Shen, J. Increased oxygen consumption in the somatosensory cortex of alpha-chloralose anesthetized rats during forepaw stimulation determined using MRS at 11.7 Tesla Neuroimage 2006;32: 1317–1325. Stefanovic, B., Bosetti, F., Silva, A. C. Modulatory role of cyclooxygenase-2 in cerebrovascular coupling. Neuroimage 2006;32:23–32. Stefanovic, B., Schwindt, W., Hoehn, M., Silva, A. C. Functional uncoupling of hemodynamic from neuronal response by inhibition of neuronal nitric oxide synthase. J Cereb Blood Flow Metab 2007;27:741–754. Sanganahalli, B. G., Herman, P., Hyder, F. Frequency-dependent tactile responses in rat brain measured by functional MRI. NMR Biomed 2008;21:410–416. Herman, P., Sanganahalli, B. G., Hyder, F. Multimodal measurements of blood plasma and red blood cell volumes during functional brain activation. J Cereb Blood Flow Metab 2009;29:19–24. Ueki, M., Mies, G., Hossmann, K. A. Effect of alpha-chloralose, halothane, pentobarbital and nitrous oxide anesthesia on metabolic coupling in somatosensory cortex of rat. Acta Anaesthesiol Scand 1992;36:318–322. Soma, L. R. Anesthetic and analgesic considerations in the experimental animal. Ann N Y Acad Sci 1983;406:32–47. Silverman, J., Muir, W. W., III A review of laboratory animal anesthesia with chloral hydrate and chloralose. Lab Anim Sci 1993;43:210–216. Willis, C. K., Quinn, R. P., McDonell, W. M., Gati, J., Parent, J., Nicolle, D. Functional MRI as a tool to assess vision in dogs: The optimal anesthetic. Vet Ophthalmol 2001;4:243–253. Heinke, W., Schwarzbauer, C. Subanesthetic isoflurane affects task-induced brain activation in a highly specific manner: A functional

Longitudinal fMRI in Animal Models

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

magnetic resonance imaging study. Anesthesiology 2001;94:973–981. Sicard, K., Shen, Q., Brevard, M. E., Sullivan, R., Ferris, C. F., King, J. A., Duong, T. Q. Regional cerebral blood flow and BOLD responses in conscious and anesthetized rats under basal and hypercapnic conditions: Implications for functional MRI studies. J Cereb Blood Flow Metab 2003;23:472–481. Liu, Z. M., Schmidt, K. F., Sicard, K. M., Duong, T. Q. Imaging oxygen consumption in forepaw somatosensory stimulation in rats under isoflurane anesthesia. Magn Reson Med 2004;52:277–285. Abo, M., Suzuki, M., Senoo, A., Miyano, S., Yamauchi, H., Yonemoto, K., Watanabe, S., Edstrom, L. Influence of isoflurane concentration and hypoxia on functional magnetic resonance imaging for the detection of bicuculline-induced neuronal activation. Neurosignals 2004;13:144–149. Dashti, M., Geso, M., Williams, J. The effects of anaesthesia on cortical stimulation in rats: A functional MRI study. Australas Phys Eng Sci Med 2005;28:21–25. Masamoto, K., Kim, T., Fukuda, M., Wang, P., Kim, S. G. Relationship between neural, vascular, and BOLD signals in isofluraneanesthetized rat somatosensory cortex. Cereb Cortex 2007;17:942–950. Duong, T. Q. Cerebral blood flow and BOLD fMRI responses to hypoxia in awake and anesthetized rats. Brain Res 2007;1135:186–194. Sommers, M. G., van, E. J., Booij, L. H., Heerschap, A. Isoflurane anesthesia is a valuable alternative for alpha-chloralose anesthesia in the forepaw stimulation model in rats. NMR Biomed 2009;22:414–418. Hentschke, H., Schwarz, C., Antkowiak, B. Neocortex is the major target of sedative concentrations of volatile anaesthetics: Strong depression of firing rates and increase of GABAA receptor-mediated inhibition. Eur J Neurosci 2005;21:93–102. Todd, M. M., Drummond, J. C. A comparison of the cerebrovascular and metabolic effects of halothane and isoflurane in the cat. Anesthesiology 1984;60:276–282. Drummond, J. C., Todd, M. M., Scheller, M. S., Shapiro, H. M. A comparison of the direct cerebral vasodilating potencies of halothane and isoflurane in the New Zealand white rabbit. Anesthesiology 1986;65:462–467. Masamoto, K., Fukuda, M., Vazquez, A., Kim, S. G. Dose-dependent effect of isoflurane on neurovascular coupling in rat cerebral cortex. Eur J Neurosci 2009;30: 242–250.

301

53. Wegener, S., Wong, E. C. Longitudinal MRI studies in the isoflurane-anesthetized rat: Long-term effects of a short hypoxic episode on regulation of cerebral blood flow as assessed by pulsed arterial spin labelling. NMR Biomed 2008;21:696–703. 54. Scanley, B. E., Kennan, R. P., Cannan, S., Skudlarski, P., Innis, R. B., Gore, J. C. Functional magnetic resonance imaging of median nerve stimulation in rats at 2.0 T Magn Reson Med 1997;37:969–972. 55. Lahti, K. M., Ferris, C. F., Li, F., Sotak, C. H., King, J. A. Comparison of evoked cortical activity in conscious and propofolanesthetized rats using functional MRI. Magn Reson Med 1999;41:412–416. 56. Kalisch, R., Elbel, G. K., Gossl, C., Czisch, M., Auer, D. P. Blood pressure changes induced by arterial blood withdrawal influence bold signal in anesthesized rats at 7 Tesla: Implications for pharmacologic MRI. Neuroimage 2001;14:891–898. 57. Makiranta, M. J., Lehtinen, S., Jauhiainen, J. P., Oikarinen, J. T., Pyhtinen, J., Tervonen, O. MR perfusion, diffusion and BOLD imaging of methotrexate-exposed swine brain. J Magn Reson Imaging 2002;15:511–519. 58. Weber, R., Ramos-Cabrer, P., Wiedermann, D., Van Camp, N., Hoehn, M. A fully noninvasive and robust experimental protocol for longitudinal fMRI studies in the rat. Neuroimage 2006;29:1303–1310. 59. Zhao, F., Zhao, T., Zhou, L., Wu, Q., Hu, X. BOLD study of stimulation-induced neural activity and resting-state connectivity in medetomidine-sedated rat. Neuroimage 2008;39:248–260. 60. Pawela, C. P., Biswal, B. B., Cho, Y. R., Kao, D. S., Li, R., Jones, S. R., Schulte, M. L., Matloub, H. S., Hudetz, A. G., Hyde, J. S. Resting-state functional connectivity of the rat brain. Magn Reson Med 2008;59: 1021–1029. 61. Pawela, C. P., Biswal, B. B., Hudetz, A. G., Schulte, M. L., Li, R., Jones, S. R., Cho, Y. R., Matloub, H. S., Hyde, J. S. A protocol for use of medetomidine anesthesia in rats for extended studies using task-induced BOLD contrast and resting-state functional connectivity. Neuroimage 2009;46:1137–1147. 62. Lee, H. B., Blaufox, M. D. Blood volume in the rat. J Nucl Med 1985;26:72–76. 63. Hinojosa-Laborde, C., Greene, A. S., Cowley, A. W., Jr. Autoregulation of the systemic circulation in conscious rats. Hypertension 1988;11:685–691. 64. Skarlatos, S., Brand, P. H., Metting, P. J., Britton, S. L. Spontaneous changes in arterial blood pressure and renal interstitial

302

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

Silva et al. hydrostatic pressure in conscious rats. J Physiol 1994;481(Pt 3):743–752. Schnell, C. R., Wood, J. M. Measurement of blood pressure and heart rate by telemetry in conscious, unrestrained marmosets. Am J Physiol 1993;264:H1509–H1516. Hu, X., Le, T. H., Parrish, T., Erhard, P. Retrospective estimation and correction of physiological fluctuation in functional MRI. Magn Reson Med 1995;34:201–212. Lahti, K. M., Ferris, C. F., Li, F., Sotak, C. H., King, J. A. Imaging brain activity in conscious animals using functional MRI. J Neurosci Methods 1998;82:75–83. Khubchandani, M., Mallick, H. N., Jagannathan, N. R., Mohan, K. V. Stereotaxic assembly and procedures for simultaneous electrophysiological and MRI study of conscious rat. Magn Reson Med 2003;49:962–967. Tenney, J. R., Duong, T. Q., King, J. A., Ferris, C. F. FMRI of brain activation in a genetic rat model of absence seizures. Epilepsia 2004;45:576–582. Ferris, C. F., Febo, M., Luo, F., Schmidt, K., Brevard, M., Harder, J. A., Kulkarni, P., Messenger, T., King, J. A. Functional magnetic resonance imaging in conscious animals: A new tool in behavioural neuroscience research. J Neuroendocrinol 2006;18:307–318. Febo, M., Shields, J., Ferris, C. F., King, J. A. Oxytocin modulates unconditioned fear response in lactating dams: An fMRI study. Brain Res 2009;1302:183–193. Dubowitz, D. J., Chen, D. Y., Atkinson, D. J., Grieve, K. L., Gillikin, B., Bradley, W. G., Jr., Andersen, R. A. Functional magnetic resonance imaging in macaque cortex. Neuroreport 1998;9:2213–2218. Zhang, Z., Andersen, A. H., Avison, M. J., Gerhardt, G. A., Gash, D. M. Functional MRI of apomorphine activation of the basal ganglia in awake rhesus monkeys. Brain Res 2000;852:290–296. Dubowitz, D. J., Bernheim, K. A., Chen, D. Y., Bradley, W. G., Jr., Andersen, R. A. Enhancing fMRI contrast in awakebehaving primates using intravascular magnetite dextran nanopartieles. Neuroreport 2001;12:2335–2340. Ferris, C. F., Snowdon, C. T., King, J. A., Duong, T. Q., Ziegler, T. E., Ugurbil, K., Ludwig, R., Schultz-Darken, N. J., Wu, Z., Olson, D. P., Sullivan, J. M., Jr, Tannenbaum, P. L., Vaughan, J. T. Functional imaging of brain activity in conscious monkeys responding to sexually arousing cues. Neuroreport 2001;12:2231–2236. Vanduffel, W., Fize, D., Mandeville, J. B., Nelissen, K., Van Hecke, P., Rosen, B. R., Tootell, R. B., Orban, G. A. Visual motion

77.

78. 79.

80.

81.

82.

83.

84.

85.

86.

processing investigated using contrast agentenhanced fMRI in awake behaving monkeys. Neuron 2001;32:565–577. Andersen, A. H., Zhang, Z., Barber, T., Rayens, W. S., Zhang, J., Grondin, R., Hardy, P., Gerhardt, G. A., Gash, D. M. Functional MRI studies in awake rhesus monkeys: Methodological and analytical strategies. J Neurosci Methods 2002;118:141–152. Orban, G. A. Functional MRI in the awake monkey: The missing link. J. Cogn Neurosci 2002;14:965–969. Leite, F. P., Tsao, D., Vanduffel, W., Fize, D., Sasaki, Y., Wald, L. L., Dale, A. M., Kwong, K. K., Orban, G. A., Rosen, B. R., Tootell, R. B., Mandeville, J. B. Repeated fMRI using iron oxide contrast agent in awake, behaving macaques at 3 Tesla. Neuroimage 2002;16:283–294. Ferris, C. F., Snowdon, C. T., King, J. A., Sullivan, J. M., Jr., Ziegler, T. E., Olson, D. P., Schultz-Darken, N. J., Tannenbaum, P. L., Ludwig, R., Wu, Z., Einspanier, A., Vaughan, J. T., Duong, T. Q. Activation of neural pathways associated with sexual arousal in non-human primates. J Magn Reson Imaging 2004;19:168–175. Pfeuffer, J., Shmuel, A., Keliris, G. A., Steudel, T., Merkle, H., Logothetis, N. K. Functional MR imaging in the awake monkey: Effects of motion on dynamic off-resonance and processing strategies. Magn Reson Imaging 2007;25: 869–882. Goense, J. B., Ku, S. P., Merkle, H., Tolias, A. S., Logothetis, N. K. FMRI of the temporal lobe of the awake monkey at 7 T. Neuroimage 2008;39:1081–1093. Maier, A., Wilke, M., Aura, C., Zhu, C., Ye, F. Q., Leopold, D. A. Divergence of fMRI and neural signals in V1 during perceptual suppression in the awake monkey. Nat Neurosci 2008;11:1193–1200. Wang, Z., Guo, Y., Bradesi, S., Labus, J. S., Maarek, J. M., Lee, K., Winchester, W. J., Mayer, E. A., Holschneider, D. P. Sex differences in functional brain activation during noxious visceral stimulation in rats. Pain 2009;145:120–128. Stefanacci, L., Reber, P., Costanza, J., Wong, E., Buxton, R., Zola, S., Squire, L., Albright, T. FMRI of monkey visual cortex. Neuron 1998 Jun 1998;20: 1051–1057. Schultz-Darken, N. J., Pape, R. M., Tannenbaum, P. L., Saltzman, W., Abbott, D. H. Novel restraint system for neuroendocrine studies of socially living common marmoset monkeys. Lab Anim 2004;38:393–405.

Chapter 15 Combining EEG and fMRI Karen Mullinger and Richard Bowtell Abstract The combination of electroencephalography (EEG) with functional magnetic resonance imaging (fMRI) forms a powerful tool for the investigation of brain function, but concurrent implementation of EEG and fMRI poses many technical challenges. Here, the motivation for combining EEG and fMRI is explored and methods underlying the combination are described. After a brief introduction to the two different techniques, the advantages and disadvantages of different methods of data recording are detailed, followed by a description of the artefacts encountered when performing EEG and fMRI measurements simultaneously, and the methods which have been developed to eliminate these artefacts. Important safety considerations and potential pitfalls associated with simultaneous recording are also described. The ways in which EEG and fMRI data analysis can be integrated are then described along with examples of key work which illustrate the power of combined EEG/fMRI measurements. The chapter concludes with a brief discussion of future directions for combined EEG/fMRI research. Key words: Simultaneous EEG–fMRI, gradient artefact correction, pulse artefact correction, B0 inhomogeneity, B1 inhomogeneity, EEG–fMRI safety, EEG–fMRI data fusion.

1. Introduction Functional brain imaging is a growing field of research with many different, but complementary, imaging methods available to investigators. The combination of functional imaging techniques to produce more powerful tools for investigating brain activity is becoming an increasing feature of this research. In particular, the combination of electroencephalography (EEG) with functional magnetic resonance imaging (fMRI) has been shown to be a valuable method, which is increasingly being applied to

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_15, © Springer Science+Business Media, LLC 2011

303

304

Mullinger and Bowtell

the investigation of the function of the brain and its dysfunction in the presence of disease. In this chapter, we explore the motivation for combining EEG and fMRI and describe the methods underlying the combination. After a brief introduction to the two different techniques, we detail the advantages and disadvantages of different methods of data recording before exploring the artefacts encountered when performing EEG and fMRI measurements simultaneously, and the methods which have been developed to eliminate these artefacts. We then discuss the ways in which EEG and fMRI data can be integrated and given examples of key work that illustrates the power of combined EEG/fMRI measurements. The chapter concludes with a brief discussion of future directions for combined EEG/fMRI research.

2. Basics of EEG and fMRI 2.1. Electroencephalography (EEG)

Electroencephalography (EEG) measures the voltages generated at the surface of the scalp by the synchronous firing of networks of neurons in the brain. This technique was first developed by Hans Berger in 1929 (1, 2) and is now used in a wide range of clinical investigations and research studies. EEG can be used to measure both on-going oscillatory activity, which occurs across different frequency bands (theta: 4–8 Hz, alpha: 8–13 Hz, beta: 13–30 Hz and gamma: 30–200 Hz), as well as evoked potentials, which are immediate neuronal responses to stimuli. Since EEG directly measures the signature of electrical activity in the brain, its temporal resolution is excellent, of the order of milliseconds. However, EEG does suffer from a number of limitations related to sensitivity and spatial resolution. The voltages generated at the scalp due to neuronal currents in the brain result from electrical current flowing through the brain and skull to reach the scalp, and are consequently spatially disperse and weak in magnitude. For example, a single postsynaptic potential from an individual neuron firing at a depth of 5 cm produces a voltage of the order of a few pico-volts at the surface of the scalp. Consequently the synchronous firing of large numbers of neurons is needed to produce signals in the microvolt range which can easily be recorded by EEG equipment. Unfortunately, identification of the location of the neuronal source from measurements of voltages at the scalp surface is not straightforward. The inverse problem (3) which must be solved for this purpose does not have a unique solution and so generally requires additional constraints to be imposed (4) (e.g. through minimising the energy of the

Combining EEG and fMRI

305

calculated currents). Solution of the inverse problem requires an ability to calculate the voltages produced at the scalp by a known distribution of neuronal currents (the forward problem), and a model characterising the electrical properties of the head is needed when solving this forward problem. Such models range in complexity from a simple homogeneous sphere to complex boundary element models (3) formed from magnetic resonance image data which capture the proper shape of the brain and skull. The difficulties in accurately modelling the head and the ill-posed nature of the inverse problem mean that the accuracy of source localisation in EEG is limited. 2.2. Functional Magnetic Resonance Imaging (fMRI)

Functional magnetic resonance imaging (fMRI) can be performed using a number of different approaches, including those based on blood oxygen-level-dependent (BOLD) contrast (5–7) arterial spin labelling (ASL) (8, 9) and the use of injected contrast agents (10, 11). All of these techniques rely upon measuring the effects of the changes in blood flow and blood volume which accompany neuronal activation and thus only provide an indirect measure of neuronal activity. To date, combined EEG–fMRI has mainly been carried out using BOLD-based fMRI and thus the remainder of this chapter will focus on the combination of EEG with fMRI based on BOLD contrast. BOLD contrast relies on using haemoglobin as an endogenous contrast agent, which is possible because of the different magnetic properties of oxy- and deoxyhaemoglobin. Deoxygenated haemoglobin is paramagnetic, whilst oxygenated haemoglobin and brain tissues are diamagnetic. This means that inside an MR scanner, the magnetic field is distorted around blood vessels containing deoxygenated blood, and the resulting dephasing of the MR signal increases the rate of MR signal decay. This reduces the T2 ∗ relaxation time which is the time constant characterising the signal decay. Blood flow increases in active brain tissue, without a concomitant increase in oxygen consumption, and as a result, the concentration of deoxyhaemoglobin in the capillaries and venules is decreased. The increased blood oxygenation therefore causes a decrease in the susceptibility difference between the vasculature and surrounding tissue, and consequently reduces the rate of MR signal decay. The consequence of activation is therefore a local increase in the T2 ∗ relaxation time. By considering the difference of the signal intensity in images collected during a rest period and whilst a task is performed, areas of the brain involved in the task can be identified. The BOLD effect was initially identified in the early 1990s (5–7) and is now widely used in fMRI studies by neuroscientists. fMRI provides much better spatial localisation of brain activity than is possible with EEG. In fMRI, the accuracy of spatial localisation is limited by the effect of draining veins, since blood oxygenation changes occurring in the venous vasculature that

306

Mullinger and Bowtell

drains the active tissue can cause signal changes a few millimetres away from the site of activation. In addition, the minimum size of the voxels used in fMRI is limited by a number of factors, including the signal-to-noise ratio (SNR) and demands on scanner hardware, that are also linked to the extent of brain coverage that is required. Standard fMRI experiments typically employ voxels that have dimensions of a few millimetres. The temporal resolution of fMRI is far poorer than that of EEG, since it is based on detecting the effects of blood oxygenation changes which are delayed with respect to the increased electrical activity. The peak of the BOLD response typically occurs 5–8 s after the stimulus onset (12) and this time to peak varies across brain areas and subjects, making it difficult to infer information about the timing of activity occurring in the brain on the millisecond timescale that is accessible to EEG.

3. Approaches to Combining EEG and fMRI

The poor spatial resolution, but good temporal resolution of EEG, and conversely the poor temporal resolution, but good spatial resolution of fMRI, make the two techniques highly complementary, so that significant advantages can potentially be realised by combining EEG and fMRI in measuring brain function. EEG and fMRI can be combined at three different levels where data are collected by applying the techniques (i) separately, (ii) in an interleaved fashion, or (iii) simultaneously. Simultaneous acquisition of fMRI and EEG data obviously represents the best integration of the two techniques, as it ensures that haemodynamic and electrical responses corresponding to exactly the same series of brain states are recorded. This, for example, guarantees that the responses seen by the two modalities reflect exactly the same degree of habituation or changes in attentional state. Simultaneous acquisition is essential when unpredictable neuronal activity is the focus of a study. This includes studies of the haemodynamic correlates of natural fluctuations in oscillatory activity (e.g. alpha or beta activity) or of changes in the timing or strength of event-related potentials (ERPs) in individual trials, as well as monitoring of BOLD effects linked to epileptiform activity. There are, however, also a number of disadvantages of simultaneously acquiring EEG and fMRI data. The biggest of these is the degradation of the quality of data that are recorded simultaneously rather than separately. As discussed below, this is primarily a problem for EEG data recorded inside the hostile environment of an MR scanner, with the gradient artefact produced by the time-varying magnetic field gradients

Combining EEG and fMRI

307

needed for imaging and the pulse artefact produced by cardiacpulse-driven motion in the strong magnetic field of the MR scanner forming the main confounds. The quality of fMRI data can also be degraded by the presence of EEG recording equipment. In addition, the paradigm that is employed in simultaneous recordings must be suitable for both modalities, which may often mean that it is sub-optimal for one or the other, or both. For example, the many repeated stimulus presentations with very short inter-stimulus intervals that are often used in EEG-only experiments are not compatible with the measurement of well-defined haemodynamic responses using fMRI. It is important therefore to ensure that simultaneous recording of EEG and fMRI signals is strictly necessary before embarking on such experiments. For example, measurements aimed at exploring the average relationship between haemodynamic and electrical responses to wellcontrolled stimuli can readily be made from EEG and fMRI data recorded separately. Interleaved recording offers a halfway house between simultaneous and separate recordings and involves recording EEG data inside the MR scanner, but only during ‘quiet’ periods when no time-varying magnetic field gradients are being applied. Interleaved recordings generally rely upon the fact that the BOLD response is delayed relative to the neuronal activity. This means, for example, that by monitoring the EEG signal one can identify an evoked response to a stimulus or a spontaneous brain oscillation of interest and use this to trigger the recording of fMRI data. This approach allows EEG data to be collected without contamination from the gradient artefacts and also has the advantage that the EEG response can be monitored easily in real time. Information from the EEG data, such as the strength of an evoked response or spiking activity, in the case of epilepsy, can then be used in the analysis of fMRI data, an approach which has been successfully employed by a number of groups (13, 14). However, this method has some important limitations. The first, and perhaps the most significant, is that neither modality is continually monitoring brain activity, so that valuable information about the activation may not be available from the recorded data. Interleaving the acquisitions may also result in longer experiment times which can result in more subject discomfort. Finally, the nature of this type of acquisition prevents a steady state being reached prior to MR image acquisition and therefore spin history (15) effects may compromise the fMRI data. The remainder of this chapter focuses mainly on the simultaneous acquisition of EEG and fMRI data, since this approach enables questions to be addressed which cannot be answered using separate or interleaved recording. The power of this approach underlies the rapid expansion of the field of simultaneous EEG–fMRI over the last decade (16–21).

308

Mullinger and Bowtell

4. Sources of Artefact in Simultaneous EEG–fMRI Recordings

4.1. EEG Data Quality

The simultaneous acquisition of EEG and fMRI data poses a number of technical challenges. These challenges initially limited progress in exploiting the advantages of simultaneous recording and despite the rapid advances which have been made in recent years, further work is still needed to meet these challenges fully. Two physical effects are responsible for the degradation of EEG signals recorded in the magnet during MR scanning. The first effect is characterised by Faraday’s law of induction: ‘An electromotive force (emf) is induced in a loop of wire when the magnetic flux through a surface bounded by the loop changes in time’ (22). Such changes in flux can be produced by the application of timevarying magnetic fields or by movement of the wire loop in a static magnetic field. In EEG recordings, artefacts result from the emfs generated in the ‘loops’ formed by the EEG leads and the conducting tissues of the head. The second effect results from the flow of blood (an electrically conducting fluid) in a magnetic field. When the direction of flow is perpendicular to the magnetic field, the ions in the blood experience a force which acts across the vessel and as a result, charge accumulates at the vessel surface. This charge gives rise to electric fields around the vessel that produce voltages at the surface of the body (23). These effects, described in more detail by Allen et al. (24) generate artefacts in the EEG signal for a number of reasons: (i) The largest artefact in EEG data recorded concurrently with fMRI is produced by the time-varying magnetic field gradients that are used for spatial encoding of the MR signal. The temporal variation of the field gradients which can be as large as 200 Tm–1 s–1 produces large changes in magnetic flux over short periods of time (of order 1 mT in 100 μs) in the head and at the EEG leads and therefore, via Faraday’s law, produces voltages at the amplifier inputs which appear as artefacts in the signal. The resulting gradient artefact is large in magnitude (10–1,000 times larger than signals from the brain (25)) and it shows a topographical variation which depends on the position of the head and EEG leads with respect to the applied magnetic field gradients and has a predictable temporal form that depends on the derivative with respect to time of the gradient waveforms. (ii) The second most significant effect is the ballistocardiogram (BCG) or pulse artefact, which results from periodic motion linked to the cardiac cycle. The exact aetiology

Combining EEG and fMRI

309

of the pulse artefact is not well understood, but potential causes include the Hall voltage generated by blood flow in arteries in the brain and scalp, small head movements linked to the transfer of momentum to the head from inrushing arterial blood (26) and expansion of the scalp due to pulsatile blood flow. The magnitude of the pulse artefact scales with field strength (27) and is typically of order 100 μV in magnitude at 3 T. The pulse artefact waveform also shows a complicated pattern of spatial variation over the surface of the scalp, but its temporal and spatial characteristics are similar across repeated cardiac cycles. Although considerable progress has been made in understanding this artefact, research directed at elucidation of its spatiotemporal characteristics is ongoing (28). (iii) Away from the centre of the magnet, the static magnetic field becomes non-uniform. Movement of a lead in this inhomogeneous field causes a change in the flux linked by the loop formed by the leads and body, and consequently will induce an unwanted emf at the input to the amplifier. This source of artefact can be reduced significantly, however, by simply twisting together electrode leads running from the subject’s head to the amplifier inputs. This has the effect of reducing the area of the wire loops, thus reducing the induced emf. Securely fixing the leads to a stationary surface further reduces the induced artefact voltages by limiting movement (24). (iv) Head rotation (caused by talking, swallowing, coughing or turning) can produce large flux changes at the surface of the head and thus inducing a large emf on the scalp. This is not a significant additional problem for EEG/fMRI as steps are usually taken to avoid head motion in fMRI studies and any large movements will corrupt the fMRI data, which must then be discarded. The factors which most significantly limit the successful execution of simultaneous EEG and fMRI are (i) and (ii), as the other artefact sources can usually be minimised by use of an appropriate experimental setup. A variety of techniques for correcting the gradient and pulse artefacts have been developed, and the most widely used methods are described below. Artefact correction is still an active area of research and consequently there is not yet a single, accepted ‘best’ method for correcting artefacts in EEG data acquired during concurrent fMRI. 4.1.1. Gradient Artefact Elimination via the Stepping Stone Technique

The stepping stone technique allows the MR and EEG data acquisitions to be interleaved on a millisecond time scale (29). It uses a modified MR imaging sequence that includes short periods during which no temporal variation of the gradient occurs which

310

Mullinger and Bowtell

are regularly spaced throughout the acquisition. EEG data that are unaffected by gradient artefacts can therefore be acquired by only recording in the periodically occurring quiet periods of the sequence. This is a clever approach which allows gradient artefactfree data to be produced with minimal post-processing, although pulse artefacts are still present in the data. A disadvantage of the stepping stone approach is that it requires the MRI sequence to be significantly modified, potentially compromising image quality and certainly requiring reprogramming of the scanner to produce the non-standard gradient waveforms. Due to these limitations, the stepping-stone method has not been taken up by many investigators. Recently, however, this technique has been used in conjunction with post-processing artefact-reduction methods to enable the recording of ultra high-frequency (600 Hz) EEG signals (30) during concurrent fMRI, and this approach may become more widely used in the future. 4.1.2. Average Artefact Subtraction (AAS)

The average artefact subtraction (AAS) technique for removal of gradient and pulse artefacts via post-processing was developed by Allen et al. (24, 25) It was the first method which allowed combined EEG–fMRI to be implemented in a truly simultaneous manner and is currently the most commonly used correction technique. It involves calculation of an average artefact template, produced by averaging many repeats of the artefact waveforms, followed by subtraction of this template from the data at each occurrence of the artefact. There are a number of requirements that must be satisfied for successful implementation of this method. Firstly, the artefact waveform must be the same each time it appears; secondly the amplifiers must have a large enough dynamic range to be able to sample the artefact waveform without saturation of the signal on any channel. Thirdly, the artefact waveforms must be precisely sampled and their times of onset must be accurately known or measured.

4.1.2.1. Gradient Artefact Correction Using AAS

Correction of the gradient artefact using AAS (25) involves formation of an average artefact template, for each EEG lead, which spans one period of the repeated artefact waveform. The vast majority of fMRI experiments use multi-slice echo planar imaging (EPI) to repeatedly scan an image volume made up of multiple slices. The time taken to acquire each volume is usually written as TR. The gradient waveforms are identical across repeated acquisitions of image volumes and generally also exactly the same for acquisition of each of the N slices that make up the image volume, but it is not always the case that the individual slice acquisitions are uniformly distributed across the TR period. It is often therefore easier to define the period over which the gradient artefact template is formed as the time corresponding to

Combining EEG and fMRI

311

the acquisition of one volume (i.e. = TR) rather than that corresponding to the acquisition of one slice (although when the slice acquisitions are ‘equidistantly’ distributed across the TR period, the latter approach is readily possible and offers some advantages (31)). It is necessary to know the time in the EEG recording at which each MRI volume acquisition occurs in order to average across volume acquisitions when producing the artefact template. This is generally accomplished by recording a marker produced by the MR scanner at the beginning of each volume acquisition along with the EEG data. As a rule of thumb, it is necessary to average over at least 20 volumes in order to achieve a sufficient attenuation of true brain signals in the average artefact template (32). If this condition is not satisfied, then the artefact subtraction will corrupt the signals from the brain that are of interest. Once the average artefact has been formed, an adaptive noise cancellation (ANC) filter can then be applied to remove residual artefacts from the EEG trace. This filter is ideal for gradient artefact correction, as it reduces signal components that are correlated to a reference signal. Even though the AAS approach does an excellent job of reducing the amplitude of the gradient artefact, it is generally also necessary to apply some low-pass filtering to the corrected data to attenuate further the high-frequency components of the gradient artefact which contain a significant amount of power. Nowadays, it is typical to use a 70-Hz cut-off frequency for the low-pass filtering, although a value of 50 Hz was originally recommended by Allen et al. (25). The use of low-pass filtering has the significant disadvantage of placing a limit on the frequency range over which neuronal activity can be investigated using combined EEG/fMRI. In particular, such filtering causes problems for the study of the gamma-band activity (30–200 Hz) using combined EEG–fMRI. Although the gradient waveforms that are applied across repeated volume acquisitions are identical, the gradient artefact induced on each lead also depends on the position of the head relative to the scanner. Small changes in the position of the head during an EEG/fMRI study, thus, cause the artefact waveform to vary across volume acquisitions. This is problematic for the AAS approach, resulting in the presence of residual gradient artefacts in EEG data even after AAS has been applied. Moosmann et al. (33) have recently attempted to address this issue by using motion parameters calculated from the MR data to divide the data up into segments for which different gradient artefact templates are constructed. A new template is formed whenever a significant movement has occurred, thus allowing for changes in the artefact morphology across leads. This approach, which was shown to improve the performance of AAS when head movements are greater than 1 mm in magnitude (33) offers some promise for the future.

312

Mullinger and Bowtell

4.1.2.2. Pulse Artefact Correction Using AAS

Correction of the pulse artefact using AAS requires the formation of an artefact template by averaging the artefact waveform at each lead across multiple cardiac cycles. This is usually accomplished by identifying the onset of each cardiac cycle from the positions of the R-peaks in a simultaneously recorded electrocardiogram (ECG) (24). The peak in the pulse artefact occurs at a time of about 200 ms after the R-peak as a result of the delay between the contraction of the heart and the arrival of the arterial pulse in the head (24). The average artefact is generally formed by averaging artefact waveforms over cardiac cycles occurring in a sliding window of 10–20-s duration. A relatively short duration is necessary, because the pulse artefact has been found to vary across successive cardiac cycles. Averaging over long times, thus, yields an average artefact that is not representative of the individual occurrences of the pulse artefact, whilst the neuronal activity of interest will be attenuated by AAS when a sliding window of too short duration is employed. Correction of the pulse artefact using AAS is in general more problematic than correction of the gradient artefact, because of the variability of the timing and spatial form of the artefact produced by cardiac pulsation.

4.1.3. Optimal Basis Sets (OBS)

The OBS method, which was introduced by Niazy et al. (34), is designed to complement the AAS technique in correcting the gradient artefact and can also be used to remove the pulse artefact.

4.1.3.1. Gradient Artefact Correction Using the OBS Method

This method can be used to remove residual gradient artefacts that remain after the application of AAS. It involves applying principal component analysis (PCA) to the dataset after it has been segmented into a series of volume or slice acquisitions. This yields a basis set, which characterises the temporal form of the residual artefacts. A unique artefact template is then formed for each volume or slice acquisition (on each EEG channel) using a linear combination of the basis functions. Subtraction of these templates from the data has been shown to attenuate the residual gradient artefact (34).

4.1.3.2. Pulse Artefact Correction Using the OBS Method

Application of the OBS method to correction of the pulse artefact involves a similar process (34) except that before application of PCA, the data are divided up into a series of individual pulse artefacts (each corresponding to a single cardiac cycle), on the basis of R-peak occurrences in the ECG trace. Application of PCA to these data allows the identification of the principal components of the pulse artefact and the first three components are used to form the OBS. A fit to each occurrence of the pulse artefact is then formed using a linear combination of the OBS and subtracted from the data. The OBS approach can, thus, cope with some variation of the pulse artefact across the dataset and has been shown to perform better in removing the pulse artefact than

Combining EEG and fMRI

313

standard AAS correction (34) and also than independent component analysis (ICA) (discussed in the next section) (35). Application of OBS followed by ICA has been shown to offer further improvement in the efficacy of pulse artefact correction (35). 4.1.4. Independent Component Analysis (ICA)

ICA is another post-processing technique which may be employed for the removal of artefacts. ICA has been used in conventional EEG data analysis for more than 10 years (36–38) and more recently has been applied to the correction of the pulse artefact in EEG data recorded concurrently with fMRI. This is based on the sensible assumption that after gradient artefact correction, the recorded EEG data form a linear mixture of the pulse artefact and other components representing the neuronal signals of interest (39). After application of ICA, components which correspond to the pulse artefact are identified via visual inspection or by computing correlations with simultaneously acquired data from the ECG channel (39) and are then removed from the dataset. The remaining components can then be back-projected to form a pulse-artefact-free EEG dataset for further analysis. The use of ICA for pulse artefact correction has produced varying degrees of success for different groups, with many reporting an excellent ability to remove the pulse artefact (see for example (39, 40)). However, other groups reported that the use of ICA alone to remove the pulse artefact attenuates the neuronal activity of interest (35). The ICA technique is certainly viable in some situations: the best results have been achieved at lower magnetic field strengths where the pulse artefact is smaller in amplitude.

4.1.5. Other Correction Methods

Although the most widely used artefact correction techniques are described above, this is an active area of research and new approaches are still emerging, many of which exploit the use of blind source separation techniques (41, 42). Another method which has been shown to reduce the effect of residual artefacts in EEG data is the EEG beamformer (43–45). This is primarily a source localisation technique, but it also acts to attenuate artefactual signals. This is a consequence of the fact that the spatial filter used by the beamformer rejects sources of signal variance that do not appear dipolar in nature. Both the gradient and pulse artefacts are therefore attenuated since their spatial topographies differ from that of a dipole sited in the brain. It has been shown that the degree of artefact correction improves with the number of electrodes used in the EEG recording (45). The advantage of the beamformer over blind source separation techniques is that it is data-driven and the user does not have to identify noise and signal components. Beamforming is, however, best suited to the study of oscillatory activity rather than evoked potentials, thus limiting the range of EEG/fMRI studies to which it can be applied. It

314

Mullinger and Bowtell

also requires an accurate head model and knowledge of the precise location of the electrodes on the head, which may not always be available. A new hardware-based method for reducing the magnitude of the gradient and pulse artefacts at source has recently been developed (46). This approach involves using two layers of EEG electrodes and leads; the first layer of electrodes is attached to the scalp as in a standard EEG system, while the second layer is embedded in a conducting gel layer which is electrically isolated from the scalp. The electrodes and leads in this second, reference layer are arranged so that each directly overlays a similar electrode and lead that is attached to the scalp. With this arrangement, the magnetic flux changes produced by time varying gradients at the scalp are very similar to those produced in the reference layer. Consequently, the gradient artefacts produced in the two layers are very similar in form, and subtraction of the signals recorded from a reference layer channel from those recorded from the corresponding scalp layer channel greatly attenuates the gradient artefact. For similar reasons, pulse artefact contributions resulting from scalp expansion or head rotation are also cancelled by subtraction of the reference layer voltage. A nice feature of this approach is that the cancellation should occur independent of head position and also that it eliminates artefacts that are produced by other sources of head movement. In real measurements, the cancellation is not perfect and some residual artefacts remain after subtraction. Possible reasons for the imperfect cancellation include small differences in the paths of the two sets of wires that may cause minor discrepancies in the gradient and pulse artefact in the two layers, as well as differences between the current paths in the head and the gel layer. In addition, flow-induced potentials that contribute to the pulse artefact (28) will not be seen by the gel layer of the cap and so will not be cancelled by subtraction. Any residual artefacts remaining after the subtraction of the signals from the individual layers can be attenuated using adaptive noise cancellation. Detailed studies using this system are yet to be performed, but the approach provides a new way of reducing the artefacts at source, which importantly can be used in conjunction with all of the post-processing methods described above. 4.2. MR Data Quality

The issue of EEG data quality in combined EEG–fMRI experiments has received a great deal of attention because the gradient and pulse artefacts are so large compared with the signals of interest, meaning that artefact correction is essential. The effects of concurrent EEG recording on the quality of MR data are generally more subtle, but must also be considered to ensure that the information provided by the functional MR images is not compromised in combined EEG–fMRI experiments. The dominant sources of MR image degradation that must be considered when

Combining EEG and fMRI

315

collecting EEG–fMRI data are the perturbations of the static magnetic field and of the radio-frequency (RF) field involved in excitation and detection of the MR signal. These result from the introduction of the EEG cap and recording apparatus. The EEG cap produces a perturbation of the static magnetic field if the wires, electrodes or conductive gel have a significantly different magnetic susceptibility, χ, from the tissues of the head. Although the differences in magnetic susceptibility are likely to be small, the resulting field deviation can cause a significant variation in the NMR resonance frequency, ω. This can cause both image distortion and signal loss, and these effects increase with the strength of the static magnetic field, B0 . To identify if this is a problem for a specific EEG cap, MR-based B0 mapping can be carried out as described in Ref. (47). Perturbation of the RF field is caused by an interaction of the conducting elements of the EEG apparatus with the RF magnetic field that is used to excite and detect the MR signal. The signal is excited using an RF pulse which rotates the magnetisation away from its equilibrium state of alignment with the static field. The strength of the detected signal depends on the angle through which the magnetisation is rotated, and this flip-angle is directly proportional to the strength of the magnetic component of the RF field (written as B1 ). RF inhomogeneity due to the EEG apparatus will, thus, produce a variation of the flip-angle over the volume of interest and consequently an unwanted spatial variation of the signal intensity in an MR image. The strength of the MR signal received from a given region is also dependent on the value of B1 in that region and this can give rise to further image intensity variation in the presence of B1 inhomogeneity. The EEG cap and B1 field can interact in a variety of ways causing both spatial variation of the signal and a global decrease in the SNR of the image. When materials of high electrical conductivity are exposed to RF, large surface current densities which act to screen the RF field from the interior of the material are generated. These currents also disturb the B1 field in external regions that are in close proximity to the conductor. Consequently, it can be expected that electrically conducting material in EEG caps will cause some perturbation of the B1 field, leading to artefactual intensity variation in MR images. In addition, the interaction between the RF field and conducting material increases the effective resistance of the RF coil. Since a resistance acts as a source of noise, this effect can lead to a reduction in the SNR of images obtained in combined EEG/fMRI studies (48). A particularly strong perturbation of the RF field occurs when the length of an EEG lead is equal to an integer number of half wavelengths of the RF. In this situation, a standing wave can be set-up in the lead, leading to a strong distortion of the B1 field around the lead. Since the RF wavelength changes with B0 , the

316

Mullinger and Bowtell

same EEG cap and lead arrangement can produce very different effects in scanners operating at different fields. B1 mapping can be used to measure any perturbation of the RF field (49). As mentioned above, the degradation of MR image quality in combined EEG–fMRI is relatively minor compared to the effect on the EEG data, and as a consequence much less work has been carried out to ascertain and explain the effects of the EEG apparatus on MR images. However, the work that has been carried out has shown that the introduction of the EEG apparatus into the MR scanner can compromise MR data quality (50–53) and that the effects depend on the operating field and on the number of electrodes used in the EEG system (47, 48). It is therefore important to investigate the effect of a new EEG cap on MR data at all field strengths at which it will be used. Distortions and signal loss can be identified in EPI data by visual inspection. Effects may be relatively localised, so all slices acquired within a volume should be inspected. If image degradation is observed, then implementation of B0 and B1 mapping, as described earlier, will help identify the cause of the degradation and therefore guide the investigator to the relevant methods, discussed below, which might help overcome the problems. Even if no visible change in image quality is apparent, it is advisable to measure the signal-to-noise ratio (SNR) of images acquired with and without the cap in place (47) in order to evaluate whether the EEG cap causes a reduction in SNR that might compromise the detection of activation. The majority of the effects that are observed can be overcome by correctly selecting materials for the EEG cap (50). As previously discussed, the magnetic susceptibility of the materials must be considered in order to limit B0 perturbations. Interactions of the RF field with the EEG cap are more complex and can change with field strength. The use of higher resistance material (e.g. carbon or ink rather than copper) for fabrication of leads reduces the interaction with the RF field and this approach has recently been explored by Vasios et al. (53). It is worth noting that B1 inhomogeneity produced by the human head on its own increases with field strength, and consequently many groups (54, 55) are working on techniques for the amelioration of the effects of B1 variation. These techniques may also be employed to overcome the effects of EEG caps on the B1 field. 4.3. Safety

Data quality is of course an important consideration in combined EEG–fMRI experiments, but, as with any equipment that is taken into the environment of an MR scanner, patient safety is paramount and must be carefully evaluated. First and foremost, it must be ensured that any EEG system that it is to be used in combined EEG-fMRI experiments does not contain ferromagnetic material which would experience strong forces in the scanner’s fringe field (if a commercial system is used, then testing

Combining EEG and fMRI

317

would have been carried out by the company, but if the system is built in-house, then care must be taken in choosing components). Another important consideration is the isolation of the subject from external sources of electricity. Fibre optic cables are generally used to transmit the EEG signals out of the scanner room in MR-compatible EEG systems, thus providing optical isolation of the subject from the EEG recording apparatus. This also has the benefit of avoiding the conduction of externally generated RF interference into the scanner, which can occur when electrically conducting leads are run into the RF-screened room in which the scanner is sited. The potential for localised heating as a result of interaction of the applied RF with the EEG leads and electrodes must also be considered, as this could potentially harm the subject (56). Standard modifications of EEG systems to make them MR compatible include the introduction of a resistor (∼5 k) between each lead and electrode so as to reduce the current induced in the leads by the applied RF (56). In general, induced currents in leads and electrodes increase with the rate of change of the B1 field with time, and consequently scale with the RF frequency. Since the frequency of the applied RF must be increased with field strength to match the Larmor frequency, RF heating would rapidly increase with B0 if the magnitude of the RF field were kept constant. In most commercial systems, the maximum magnitude of the B1 field is decreased at higher field, so that the potential for RF heating does not scale as rapidly as might be expected with field strength. The standing wave effects which occur when a lead length matches a multiple of half the RF wavelength in fact mean that greater heating can occur at lower field strength. This shows the importance of considering the safety of experiments at individual field strengths and ensuring that safety guidelines are not exceeded. A number of investigators have made experimental measurements of RF heating in combined EEG/fMRI experiments (51, 53, 57–59) showing that temperature rises can be limited to acceptable levels when MR-compatible systems are used and regulatory limits on the specific absorption rate are observed. It is important that such heating tests are carried out before new combinations of EEG and RF hardware are used in experiments on human subjects. These should involve the measurement of temperatures around a number of electrodes and any observed temperature increases should be checked against safety guidelines (60–62). The EEG cap should be placed on a phantom which mimics the electrical properties of the human head (e.g. salineloaded agar) and connected to the EEG amplifier so that all relevant current paths are formed. Temperature recordings should be made for a period of time corresponding to the duration of a typical study with a simultaneous recording made at a site where no heating is expected. The propensity to cause heating (which

318

Mullinger and Bowtell

is characterised by the specific absorption rate) can be very different for different MRI sequences and it is therefore important that temperature recordings should be made for all relevant MR sequences which may be applied to a subject who is wearing the EEG cap.

5. Carrying Out Simultaneous EEG–fMRI Experiments 5.1. Experimental Setup

It is important that any EEG equipment that is used for simultaneous or interleaved recordings of EEG–fMRI is MR compatible. This means that the system is non-magnetic and that an evaluation of RF heating, as described in the previous section, has been carried out. The general setup of equipment for simultaneous recording is similar to that needed for separate recordings. Depending on the MR-compatible EEG system used, it may not be possible to reduce the electrode impedances to the low values that are accessible with conventional EEG systems due to the additional resistance of the built-in resistors. Care should be taken to ensure that MR-compatible gels and pastes are used to lower the impedances between the electrodes and scalp and that the minimum amount of gel/paste is used, as some of these products have been shown to cause MR image degradation (50). To ensure that the gradient artefact which is sampled by the EEG system is exactly the same for each volume acquisition, which is necessary for successful gradient artefact correction using AAS, the MR sequence must have a TR which is a multiple of the EEG sampling period. It is also worth noting that the TR which is actually applied by the scanner can be slightly different from the value which is set on the console. The difference is generally unimportant for MR data acquisition, being much less than the T1 or T2 relaxation times, but may be similar or greater in magnitude than the EEG sampling period and can therefore result in variation of the sampling of the gradient waveform across repeated volume acquisitions. External monitoring (e.g. by using an EEG system) allows the exact TR to be measured. For the purposes of gradient artefact correction, it is also helpful to record a marker in the EEG file at the beginning of each volume or slice acquisition. This is usually accomplished by recording external triggers produced by the scanner on the EEG system’s auxiliary input. If AAS is to be employed for gradient artefact correction, it is highly beneficial to synchronise the EEG sampling with the MR scanner’s clock. This can be accomplished

Combining EEG and fMRI

319

by driving the EEG system clock using the MR scanner’s 10-MHz reference signal, from which all scanner timings are derived. Apparatus for appropriate down-conversion of the 10-MHz signal to the generally lower EEG clock frequency, and subsequent use of this signal for synchronisation is now commercially available (e.g. Brain Products GmbH, SyncBox see www.brainproducts.com). Synchronisation ensures that the gradient artefacts are consistently sampled across repeated image acquisitions. This considerably improves the performance of AAS as a gradient artefact correction technique (63, 64). It is also necessary to record an ECG trace which is required by most techniques for pulse artefact correction. This can usually be recorded using an additional ECG electrode provided with the EEG cap or by using the MR scanner’s physiological monitoring system (64). The long length of the lead that is needed to record the ECG in an EEG system means that the gradient artefacts tend be larger on the ECG trace than on EEG channels. This, along with large amplitude of ECG signals, means that it is often the ECG channel that is most prone to amplifier saturation. In positioning the EEG system inside the scanner, it is important to avoid any looping of leads, which might lead to strong interactions with the applied RF and consequent local heating. Minimising the area of the circuits formed by following the reference lead from the amplifier to the head and then from an individual EEG electrode back to the amplifier helps to limit the size of the gradient artefacts and also the artefactual voltages produced by movement of the leads in the scanner’s strong static field. This can be achieved by bringing all of the leads together into a tight, twisted bundle as they leave the head. Steps should also be taken to limit movements of the leads and cable bundle in the field during experiments. When placing the EEG system within the MR scanner, it is also sensible to site the EEG amplifiers so that they are not resting on the MR scanner if possible (59). This prevents gradientinduced vibrations from inducing further artefacts in the EEG data. The cables running to the amplifiers should also be kept as straight as possible with no looping of wires which could result in significant RF heating. Although outside the MR scanner, EEG is often conducted with the subject seated; this is not possible in simultaneous EEG–fMRI recordings made using conventional MR scanners. Simultaneous recordings are usually made with the subject in a supine position, which means that the head rests on the electrodes that are sited over the posterior part of the head. This can result in excessive pressure being applied to small areas of the scalp leading to discomfort and consequent subject movement which degrade data quality. Such problems can be avoided by appropriately positioned padding, which can also be used to restrict head movement.

320

Mullinger and Bowtell

5.2. Data Analysis

The methods that can be employed for the analysis of simultaneously recorded EEG and fMRI data are extensive and will not be discussed in detail here. However, we do briefly discuss three different approaches to the integrated analysis of EEG and fMRI data. Following Kilner et al. (65), these approaches are categorised as (i) integration through prediction, (ii) integration through constraints and (iii) integration through fusion. Integration through prediction involves using the EEG data to form a model of the expected BOLD signal changes. This means convolving the time-course of a particular component of the EEG signal with the haemodynamic response function, so as to form a regressor. This forms a prediction of the time-course of BOLD signal changes linked to the particular EEG component allowing for the haemodynamic lag. Statistical analysis can then be used to identify voxels in the image whose signal variation is significantly correlated with the regressor, thus yielding a map depicting areas where the BOLD signals are consistent with the variation of the chosen EEG component. The choice of component will of course vary depending on the purpose of the study. The first studies in which this approach was employed, used the power of the EEG signal in the alpha-frequency band and tested for correlation with BOLD signal changes in the resting state, when alpha activity is most prevalent. Alpha activity at rest has a large amplitude with significant spontaneous modulation making it ideal for initial study with this multi-modal approach. Goldman et al. (66) were the first to perform this type of study, with complementary work subsequently carried out by a number of other groups (18, 67). Since then similar studies have evaluated correlations between BOLD signals and power in other frequency bands, both at rest and during task performance (21, 68, 69). Another way in which this approach can be applied is to use information from evoked responses, such as amplitude and latency, rather than rhythmic activity in forming the regressor. This approach, which relies upon the accurate recording of single trial events, has been successfully implemented by a number of investigators (16, 70). To use EEG data to predict the BOLD response in this way, one must be confident that both fMRI and EEG datasets have an SNR that is sufficient to allow single responses to be meaningfully analysed. One must also assume that variations in the measured power or evoked response are due to changes in brain activity rather than being physiological artefacts due to the effect of changes in blood flow or head movement which may also produce correlations between EEG and MR data. In the second approach (integration by constraints), fMRI data are used to provide spatial constraints when localising the sources of EEG signals. The premise of this approach is that

Combining EEG and fMRI

321

an improvement in EEG source localisation may be achieved by using the spatial information from the BOLD data as a prior in the solution of the inverse problem. Justification for this premise can be found in a number of combined EEG/fMRI studies (e.g. (48, 71)), which showed through separate analysis of data from the two modalities that haemodynamic activity and electrical responses were localised in the same brain areas. Phillips et al. (72) also developed a technique for using fMRI data to constrain the analysis of EEG data with a weighted minimum norm-source localisation method. However, in a review by Ritter et al. (19) it was suggested that this type of fusion should be performed with caution for a number of following reasons: deep sources which appear in the fMRI data may not have a correlate in the EEG signals, due to the strong depth dependence of the EEG signal; there may not be a corresponding BOLD response if low metabolic demand is generated by the neuronal activity responsible for a particular EEG signal; an EEG response may correlate to either a positive or negative BOLD signal making it difficult to know how to impose source constraints on the EEG. Nevertheless, provided the issues raised by Ritter et al. are taken into account, fMRI spatial information can still be employed as a weighted constraint (so that EEG source localisation is only guided by the fMRI data) and this can greatly assist source localisation as demonstrated by Wan et al. (73). Neither of the two approaches described above forms a true integration, since in each case a separate analysis of the EEG or fMRI data is required as a first step. A proper fusion of the data analysis, which would draw maximum benefit from simultaneous EEG/fMRI, requires a common temporal forward model that links the underlying neuronal dynamics of interest to the measured hemodynamic and electrical responses (65). Initial attempts at devising such a model have brought dividends in analysing data from particular paradigms (65, 74, 75), but further work is needed before this approach can be more widely exploited. Only a small portion of the seminal work demonstrating the integration of EEG–fMRI data has been discussed here. A broader overview of the full range of combined EEG/fMRI studies that have been carried out to date can be found in recent review papers (17, 19, 76). The clinical applications of combined EEG/fMRI are also highlighted in these reviews. Currently, the most promising area for clinical application is epilepsy. The advantage of simultaneous recordings in this instance is clear, as spiking activity occurring in epilepsy is unpredictable, but generally clearly depicted in EEG traces, while identification of the source of activity using EEG alone is difficult. The combination of EEG and fMRI could therefore potentially become a vital tool in the diagnosis and treatment of epilepsy, and the use of combined

322

Mullinger and Bowtell

EEG/fMRI in this area is currently being explored by a number of research groups (19, 76). Studies are also now emerging in which simultaneous EEG/fMRI is being used to study sleeping disorders and schizophrenia (77, 78) with early work indicating that EEG/fMRI could become a useful tool in helping to understand the differences in the brains of patients suffering from schizophrenia (77, 79).

6. The Future It is clear that combined EEG–fMRI acquisition will form a useful tool for studying brain function for decades to come. The rapid pace of developments in simultaneous EEG–fMRI, however, means that it is not yet possible to lay down a definitive method for conducting experiments or analysing data. Researchers in this field must continue to assess new techniques for their possible merits when applied to their own work. There are many different avenues still to be explored in using combined EEG/fMRI, ranging from technical development work targeted at improving data quality to applications aimed at improving understanding of the healthy and diseased brain. Particular areas of technical development that need further exploration include building on the work of Yan et al. (28) to produce a better understanding of the cause of the pulse artefact. This should enable the implementation of improved methods for pulse artefact removal which would allow low-frequency oscillatory and evoked responses to be recorded with greater fidelity. New methods for reducing the effect of gradient artefacts, like those proposed by Freyer et al. (30), also need to be developed further to allow the accurate recording of high-frequency neuronal activity during simultaneous EEG–fMRI acquisition. Neuronal activity in the gamma band is of great interest for combined EEG/fMRI as recent work (80, 81) suggests that activity in the high-frequency range may be most closely linked to the BOLD response. Unfortunately, when measured at the scalp, gamma-frequency signals tend to be significantly smaller in amplitude than lower frequency oscillatory rhythms. Gammaband activity is therefore harder to record even in ideal circumstances, and problems are exacerbated in EEG data recorded concurrently with fMRI as a result of the high-frequency residual gradient artefacts. Consequently, further work on reducing highfrequency gradient artefacts, including the development of EEG amplifiers with larger dynamic range, is needed to take forward studies of the correlation of BOLD and EEG measures of activation, which will underpin full data fusion.

Combining EEG and fMRI

323

References 1. Berger, H. Uber das elektrenkephalogramm des menschen. Arch Psychiatr Nervenkr 1929;87:527–570. 2. Berger, H. On the electroencephalogram of man. Electroencephalogr Clin Neurophysiol 1969;Supplement 28:37–73. 3. Phillips, C. Source Estimation in EEG. Combining Anatomical and Functional Constraints [PhD]. Belgium: University de Liege; 2000. 4. Michel, C. M., Thut, G., Morand, S. et al. Review: Electric source imaging of human brain functions. Brain Res Rev 2001;36:108–118. 5. Kwong, K., Belliveau, J., Chesler, D. A. et al. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci USA 1992;23:3963–3971. 6. Ogawa, S., Menon, R. S., Tank, D. W., Kim, S. -G., Merkle, H., Ellermann, J. M., Ugurbil, K. Functional brain mapping by blood oxygenation level dependent contrast magnetic resonance imaging. Biophys J 1993;64:803–812. 7. Ogawa, S., Lee, T. M., Kay, A. R., Tank, D. W. Brain magnetic-resonance-imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci U S A 1990;87:9868–9872. 8. Detre, J. A., Leigh, J. S., Williams, D. S., Koretsky, A. P. Perfusion imaging. Magn Reson Med Sci 1992;23:37–45. 9. Kim, S. G. Quantification of relative cerebral blood-flow changed by flow-sensitive alternating inversion -recovery (FAIR) techniqueapplication to functional mapping. Magn Reson Med Sci 1995;34:293–301. 10. Blockley, N. P., Francis, S. T., Gowland, P. A. Perturbation of the BOLD response by contrast agent and interpretation through a modified balloon model. Neuroimage 2009;48:84–93. 11. Frahm, J., Baudewig, J., Kallenberg, K., Kastrup, A., Merboldt, K. D., Dechent, P. The post-stimulation understoot in BOLD fMRI of human brain is not caused by elevated cerebral blood volume. Neuroimage 2008;40:473–481. 12. Buxton, R. B. Introduction to Functional Magnetic Resonance Imaging: Principles and Techniques. New York, NY: Cambridge University Press; 2002. 13. Bonmassar, G., Schwartz, D. P., Liu, A. K., Kwong, K., Dale, A. M., Belliveau, J. Spatiotemporal brain imaging of visual-evoked activity using interleaved

14.

15.

16.

17.

18.

19. 20.

21.

22. 23.

24.

25.

26.

EEG and fMRI recordings. Neuroimage 2001;13:1035–1043. Krakow, K., Woermann, F. G., Symms, M. R. et al. EEG-triggered functional MRI of interictal epileptiform activity in patients with partial seizures. Brain 1999;122: 1679–1688. Haacke, E. M., Brown, R. W., Thompson, M. R., Venkatesan, R. Magnetic Resonance Imaging: Physical Principles and Sequence Design. New York, NY: Wiley; 1999. Debener, S., Ullsperger, M., Siegel, M., Engel, A. K. Single-trial EEG-fMRI reveals the dynamics of cognitive function. Trends Cogn Sci 2006;10:558–563. Laufs, H. Endogenous brain oscialltions and related networks detected by surface EEG-combined fMRI. Hum Brain Mapp 2008;29:762–769. Laufs, H., Krakow, K., Sterzer, P. et al. Electroencephalographic signatures of attentional and cognitive default modes in spontaneous brain activity fluctuations at rest. Proc Natl Acad Sci USA 2003;100:11053–11058. Ritter, P., Villringer, A. Review: Simultaneous EEG-fMRI. Neurosci Behav Rev 2006;30:823–838. Vulliemoz, S., Thornton, R., Rodionov, R. et al. The spatio-temporal mapping of epileptic networks: Combination of EEGfMRI and EEG source imaging. Neuroimage 2009;46:834–843. Parkes, L. M., Bastiaansen, M. C. M., Norris, D. G. Combining EEG adn fMRI to investigate the post-movement beta rebound. Neuroimage 2006;29:685–696. Keller, F. J., Gettys, W. E., Skove, M. J. Physics: Classical and Modern, 2nd ed. New York, NY: McGraw-Hill; 1993. Tenforde, T. S., Gaffey, C. T., Moyer, B. R., Budinger, T. F. Cardiovascular alterations in macaca monkeys exposed to stationary magnetic fields: Experimental observations and theoretical analysis. Bioelectromagnetics 1983;4:1–9. Allen, P. J., Poizzi, G., Krakow, K., Fish, D. R., Lemieux, L. Identification of EEG events in the MR scanner: The problem of pulse artifact and a method for its subtraction. Neuroimage 1998;8:229–239. Allen, P. J., Josephs, O., Turner, R. A. Method for removing imaging artifact from continuous EEG recorded during functional MRI. Neuroimage 2000;12:230–239. Ives, J. R., Warach, S., Schmitt, F., Edelman, R. R., Schomer, D. L. Monitoring a patient’s EEG during echo planar

324

27.

28.

29.

30.

31.

32. 33.

34.

35.

36.

37.

38.

Mullinger and Bowtell MRI. Electroencephalogr Clin Neurophysiol 1993;87:417–420. Debener, S., Mullinger, K. J., Niazy, R. K., Bowtell, R. W. Properties of the ballistocardiogram artefact as revealed by EEG recordings at 1.5, 3 and 7 tesla static magnetic field strength. Int J Psychophysiol 2008;67:189–199. Yan, W. X., Mullinger, K. J., Geirsdottir, G. B., Bowtell, R. W. Physical modelling of pulse artefact sources in simultaneous EEG/fMRI. Hum Brain Mapp 2009;31:604–620. Anami, K., Mori, T., Tanaka, F. et al. Stepping stone sampling for retrieving artifactfree electoencephalogram during functional magnetic resonance imaging. Neuroimage 2003;19:281–295 (Part 1). Freyer, F., Becker, R., Anami, K., Curio, G., Villringer, A., Ritter, P. Ultra highfrequency EEG during fMRI: Pushing the limits of imaging-artifact correction. Neuroimage 2009;48:94–108. Mandelkow, H., Brandeis, D., Boesiger, P. Good practices in EEG-fMRI: The utility of retrospective synchronisation and PCA for the removal of MRI gradient artefacts. Neuroimage 2009;49:2287–2303. Gutberlet, I. BP_Press_Release_Issue_N3. Brain Products 2001;10. Moosmann, M., Schonfelder, V. H., Specht, K., Scheeringa, R., Nordby, H., Hugdahl, K. Realignment parameterinformed artefact correction for simultaneous EEG-fMRI recordings. Neuroimage 2009;45:1144–1150. Naizy, R. K., Bechmann, C. F., Iannetti, G. D., Brady, J. M., Smith, S. M. Removal of fMRI environment artifacts from EEG data using optimal basis sets. Neuroimage 2005;28:720–737. Debener, S., Strobel, A., Sorger, B. et al. Improved quality of auditory event-related potentials recorded simultaneously with 3T fMRI: Removal of the ballistocardiogram artefact. Neuroimage 2007;34:587–597. Iriarte, J., Urrestarazu, E., Valencia, M. et al. Independent component analysis as a tool to eliminate artifacts in EEG: A quantitative study. J Clin Neurophysiol 2003;20:249–257. Jung, T. P., Makeig, S., Humphries, C. et al. Removing electroencephalographic artifacts by blind source separation. Psychophysiology 2000;37:163–178. Makeig, S., Westernfield, M., Jung, T. P. et al. Functionally independent components of the late positive event-related potential during visual spatial attention. J Neurosci 1999;19:2665–2680.

39. Srivastava, G., Crottaz-Herbette, S., Lau, K. M., Glover, G. H., Menon, V. ICA-based procedures for removing ballistocariogram artifacts from EEG data acquired in MRI scanner. Neuroimage 2005;24:50–60. 40. Mantini, D., Perrucci, M. G., Cugini, S., Ferretti, A., Romani, G. L., Del Gratta, C. Complete artifact removal for EEG recorded during continuous fMRI using independent component analysis. Neuroimage 2007;34:598–607. 41. Assecondi, S., Hallez, H., Staelens, S., Bianchi, A. M., Huiskamp, G. M., Lemahieu, I. Removal of the ballistocardiographic artifact from EEG-fMRI data: A canonical correlation approach. Phys Med Biol 2009;54:1673–1689. 42. Sun, L., Rieger, J., Hinrichs, H. Maximum noise fraction (MNF) transforms to remove ballistocariographic artifacts in EEG signals recorded during fMRI scanning. Neuroimage 2009;46:144–153. 43. Van Veen, B. D., van Drongelen, W., Yuchtman, M., Suzuki, A. Localization of brain electrical activity via linearly constrained minimum variance spatial filtering. IEEE Trans Biomed Eng 1997;44:867–880. 44. Brookes, M. J., Mullinger, K. J., Stevenson, C. M., Morris, P. G., Bowtell, R. W. Simultaneous EEG source localisation and artifact rejection during concurrent fMRI by means of spatial filtering. Neuroimage 2008;40:1090–1104. 45. Brookes, M. J., Vrba, J., Mullinger, K. J. et al. Source localisation in concurrent EEG/fMRI: Applications at 7t. Neuroimage 2009;45:440–452. 46. Dunseath, W. J. R. Interference reduction apparatus for electroencephalography (EEG) measurement, has one or more of reference electrodes arranged to be in close physical proximity but not in direct electrical contact with subject. 47. Mullinger, K. J., Debener, S., Coxon, R., Bowtell, R. W. Effects of simultaneous EEG recording on MRI data quality at 1.5, 3 and 7 tesla. Int J Psychophysiol 2008;67:178–188. 48. Scarff, C. J., Reynolds, A., Goodyear, B. G., Ponton, C. W., Dort, J. C., Eggermont, J. J. Simultaneous 3-T fMRI and high-density recording of human auditory evoked potentials. Neuroimage 2004;23:1129–1142. 49. Yarnykh, V. L. Actual flip angle imaging in the pulsed steady state: A method for rapid three-dimensional mapping of the transmitted radio frequency field. Magn Reson Med Sci 2007;51:192–200. 50. Krakow, K., Allen, P. J., Symms, M. R., Lemieux, L., Josephs, O., Fish, D. R. EEG

Combining EEG and fMRI

51.

52.

53.

54.

55. 56.

57.

58.

59.

60.

61.

62. 63.

recording during fMRI experiments: Image quality. Hum Brain Mapp 2000;10:10–15. Stevens, T. K., Ives, J. R., Bartha, R. Energy Coupling between Electric Fields and Conductive Wires: Image Artifacts and Heating. In Joint Annual meeting ISMRMESMRMB; 2007 19–25 May; Berlin; 2007. Stevens, T. K., Ives, J. R., Klassen, L. M., Bartha, R. MR compatibility of EEG scalp electrodes at 4 tesla. J Magn Reson Imaging 2007;25:872–877. Vasios, C. E., Angelone, L. M., Purdon, P. L., Ahveninen, J., Belliveau, J., Bonmassar, G. EEG/(f)MRI measurements at 7 tesla using a new EEG cap (“InkCAP”). Neuroimage 2006;33:1082–1092. Adriany, G., Van de Moortele, P. F., Wiesinger, F. et al. Transmit and receive transmission line arrays for 7 tesla parallel imaging. Magn Reson Med Sci 2005;53:434–445. Katscher, U., Bornert, P. Parallel RF transmission in MRI. NMR Biomed 2006;19:393–400. Lemieux, L., Allen, P. J., Franconi, F., Symms, M. R., Fish, D. R. Recording of EEG during fMRI experiments: Patient safety. Magn Reson Med Sci 1997;38: 943–952. Lazeyras, F., Zimine, I., Blanke, O., Perrig, S. H., Seeck, M. Functional MRI with simultaneous EEG recording: Feasibility and application to motor and visual activation. J Magn Reson Imaging 2001;13:943–948. Angelone, L. M., Potthast, A., Segonne, F., Iwaki, S., Belliveau, J. W., Bonmassar, G. Metallic electrodes and leads in simultaneous EEG-MRI: Specific absorption rate (SAR) simulation studies. Bioelectromagnetics 2004;25:285–295. Mullinger, K. J., Brookes, M. J., Stevenson, C. M., Morgan, P. S., Bowtell, R. W. Exploring the feasibility of simultaneous EEG/fMRI at 7 T. Magn Reson Imag 2008;26:607–616. Agency, H. P. Protection of Patients and Volunteers Undergoing MRI Procedures. In: Radiation CaEH, ed. Health Protection Agency; 2008;90. ICNIRP Guidelines for limiting exposure to time-varying electric, magnetic and electromagnetic fields (up to 300 GHz). Health Phys 1998;74:494–522. ICNIRP Medical magnetic resoance (MR) procedures: Protection of patients. Health Phys 2004;87:197–216. Mandelkow, H., Halder, P., Boesiger, P., Brandeis, D. Synchronisation facilitates removal of MRI artefacts from concurrent

64.

65. 66.

67.

68.

69.

70.

71.

72.

73.

74.

325

EEG recordings and increases usable bandwidth. Neuroimage 2006;32:1120–1126. Mullinger, K. J., Morgan, P. S., Bowtell, R. W. Improved artefact correction for combined electroencephalography/functional MRI by means of synchronization and use of VCG recordings. J Magn Reson Imaging 2008;27:607–616. Kilner, J. M., Mattout, J., Henson, R., Friston, K. J. Hemodynamic correlates of EEG: A heuristic. Neuroimage 2005;28:280–286. Goldman, R. I., Stern, J. M., Engel, J., Cohen, M. S. Simultaneous EEG and fMRI of the alpha rhythm. NeuroReport 2002;13:2487–2492. Moosmann, M., Ritter, P., Krastel, I. et al. Correlates of alpha rhythm in functional magnetic resonance imaging and near infrared spectroscopy. Neuroimage 2003;20:145–158. Koch, S. P., Steinbrink, J., Villringer, A., Obrig, H. Synchronisation between background activity adn visually evoked potential is not mirrored by focal hyperoxygenation: Implications for the interpretation of vascular brain imaging. J Neurosci 2006;26:4940–4948. Scheeringa, R., Petersson, K. M., Oostenveld, R., Norris, D. G., Hagoort, P., Bastiaansen, M. C. M. Trial-by-trial coupling between EEG and BOLD identifies networks related to alpha and theta EEG power increases during working memory maintance. Neuroimage 2009;44:1224–1238. Eichele, T., Specht, K., Moosmann, M. et al. Assessing the spatiotemporal evolution of neuronal activation with singletrial event-related potential and functional MRI. Proc Natl Acad Sci USA 2005;102:17789–17803. Mulert, C., Jager, L., Schmitt, R. et al. Integration of fMRI and simultaneous EEG: Towards a comprehensive understanding of localization and time-course of brain activity in target detection. Neuroimage 2004;22:83–94. Phillips, C., Rugg, M. D., Friston, K. J. Anatomically informed basis functions for EEG source localization: Combining functional and anatomical constraints. Neuroimage 2002;16:678–695. Wan, X., Sekiguchi, A., Yokoyama, S. R., Riera, J., Kawashima, R. Electromagnetic source imaging: Backus-Gilbert resolution spread function-constrained and funtional MRI-guided spatial filtering. Hum Brain Mapp 2008;29:627–643. Ostwald, D., Porcaro, C., Bagshaw, A. P. An information theoretic approach to

326

75.

76.

77.

78.

Mullinger and Bowtell EEG-fMRI integration of visually evoked responses. Neuroimage 2010;49:498–516. Sotero, R. C., Trujillo-Barreto, N. J. Biophysical model for integrating neuronal activity, EEG, fMRI and metabolism. Neuroimage 2008;39:290–309. Salek-Haddadi, A., Friston, K. J., Lemieux, L., Fish, D. R. Review: Studying spontaneous EEG activity with fMRI. Brain Res Rev 2003;43:110–133. Bates, A. T., Keiehl, K. A., Laurens, K. R., Liddle, P. F. Low-frequency EEG oscillations associated with information processing in schizophrenia. Schizophr Res 2009;115:220–230. Czisch, M., Wetter, T. C., Kaufmann, C., Pollmacher, T., Holsboer, F., Auer, D. P. Altered processing of acoustic stimuli during sleep: Reduced auditory activa-

tion and visual deactivation detected by a combined fMRI/EEG study. Neuroimage 2002;16:251–258. 79. Wacker, J., Dillon, D. G., Pissagalli, D. A. The role of the nucleus accumbens and rostal anterior cingulate cortex in anhedonia: Integration of resting EEG, FMRI and volumetric techniques. Neuroimage 2009;46: 327–337. 80. Logothetis, N. K. The neural basis of the blood-oxygen-level-dependent functional magnetic resonance imaging signal. Philos Trans R Soc Lond B Biol Sci 2002;357:1003–1037. 81. Nir, Y., Fisch, L., Mukamel, R. et al. Coupling between neuronal firing rate, gamma LFP, and BOLD MRI is related to interneuronal correlations. Curr Biol 2007;17:1275–1285.

Chapter 16 MR Angiography and Arterial Spin Labelling David Thomas and Jack Wells Abstract MRI offers the ability to visualise and measure blood flow in the human body non-invasively. MR angiography (MRA) provides images of the arterial blood vessels within the body and allows measurement of blood velocities along these arteries. Arterial spin labelling (ASL) is a method for measuring the perfusion of blood into tissue (i.e. blood flow at the capillary level). This provides a key indicator of nutrient supply to the tissue. In this chapter, we have described the technical basis and practical implementation of these methods, emphasising their non-invasive (no contrast agents required) and quantitative nature. Key words: Perfusion, blood flow, arterial spin labelling, angiography, MRI, vasculature, quantification, ischaemia, imaging.

1. Introduction MR images have an inherent sensitivity to motion. While this is often seen as a problem which needs to be mitigated to avoid unwanted image artefacts, it can also be used positively to highlight tissue in the body that is in motion. In particular, a range of MR methods have been developed to visualise the vascular system and the supply of arterial blood to the various organs of the body. In this chapter, we have described the conceptual basis of this range of methods. We have focused exclusively on non-invasive methods which rely on endogenous mechanisms of contrast, and therefore do not require the administration of exogenous contrast agents. 1.1. Blood Flow and Perfusion

Firstly, it is important to clarify the difference between bulk blood flow and perfusion in the context of MRI. The former term relates to the flow of blood through the large vessels of the body

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_16, © Springer Science+Business Media, LLC 2011

327

328

Thomas and Wells

(arteries and veins), and in general can be thought of as laminar flow through these vessels. It is characterised in terms of the mean velocity of the flowing blood, with the standard units of velocity (e.g. cm/s) and is visualised by angiographic methods. Physiologically, this is useful because it allows identification of full or partial blockages in the vessels (caused by embolism or stenosis) and arteriovenous malformations, which may compromise the supply of blood to organs downstream. In contrast, perfusion is defined as the volume of blood delivered to the capillary bed of a block of tissue in a given period of time. As such, it is measured in units of millilitre of blood per 100 gram of tissue per minute (usually abbreviated to ml/100 g/min). This is distinct from bulk blood flow through the arteries and veins because it directly reflects the flow at the capillary level, where exchange of nutrients (such as oxygen and glucose) between the blood and tissue occurs. This nutritive supply is essential to maintain healthy tissue and is an important indicator of viability and function; therefore, accurate measurement is of great utility in the study and diagnosis of acute and chronic brain disorder. It is worth noting that the term ‘blood flow’ is often used to denote perfusion when associated with a particular organ or tissue structure, e.g. cerebral blood flow (CBF) in the brain or myocardial blood flow (MBF) in the heart. Tissue perfusion can be measured using MRI, either by dynamic susceptibility contrast agent bolus tracking or by arterial spin labelling (ASL), which uses blood water as an endogenous tracer and is described in detail below. Presently, the methods for MR angiography are established clinical sequences and will be found as standard options on any clinical MR scanner. On the other hand, ASL is a relative newcomer to the clinical environment and is currently still considered a ‘specialist’ sequence. Consequently, in this chapter, while we endeavour to provide a full description of both techniques, significantly more practical detail is provided for the implementation of ASL. 1.2. Non-invasive MR Angiography Methods

MR angiography (MRA) can be split into the following two main categories: • Flow-dependent techniques, which use the effect of the MR pulse sequence on the endogenous blood water to generate the desired image contrast • Contrast agent techniques, which use the effect of exogenous paramagnetic contrast agents on the relaxation properties of blood water to generate the desired image contrast The aim of both these approaches is to make vessels with flowing blood appear hyper-intense against background tissue, and they are therefore known as ‘bright blood’ sequences. Both use rapid acquisition gradient echo imaging techniques, with short TR (∼10 ms) and fairly large excitation pulse flip angle (in the

MR Angiography and Arterial Spin Labelling

329

range of 20–90◦ ) to provide heavy T1 weighting and short TE (∼1–4 ms) to minimise T2 ∗ weighting. In this chapter, we have described the use and implementation of the first category (noninvasive methods that do not require contrast agents), which can be further broken down into the following two distinct approaches: time-of-flight and phase contrast. The fundamental difference between these two approaches is that the former makes use of the effect of flow on the longitudinal magnetisation to generate flow-related contrast in the resulting images, whereas the latter is based on the effect of flow on the excited transverse magnetisation. 1.2.1. Time-of-Flight MRA

In a gradient echo acquisition with high flip angle and short TR, the longitudinal magnetisation of static tissue rapidly reduces, as a significant proportion is tipped into the transverse plane by each excitation pulse and minimal T1 recovery occurs between pulses. After the first RF pulse, the excited signal is large, but after subsequent RF pulses the signal is accordingly lower, until a low-level steady state is reached. Therefore, if the centre of k-space (which determines overall image contrast) is acquired halfway through this type of acquisition, static tissue will have relatively low signal intensity in the resulting image. However, if ‘fresh’ tissue moves into the imaging slice (which has not experienced the effect of previous RF pulses), its magnetisation will be fully relaxed. This will cause an apparently much more rapid recovery of the tissue’s longitudinal magnetisation – not due to an increased T1 relaxation rate, but due to replacement of partially saturated magnetisation with new fully relaxed magnetisation (see Fig. 16.1). This is known as the time-of-flight (TOF) effect (1, 2), and it is straightforward to see how TOF can be used to enhance the signal from flowing blood relative to adjacent static tissue. Since arterial blood flows rapidly, it will experience only a limited number of RF pulses before it washes out of the imaging slice, and so arteries will appear bright in TOF images. Conversely, static tissue experiences the whole train of RF pulses and so will have relatively low signal intensity (see Fig. 16.1).

1.2.2. Phase-Contrast MRA

Following excitation of spin magnetisation into the transverse plane by an RF pulse, the application of a magnetic field gradient will cause a phase shift by an amount dependent on the position along the direction of the applied gradient. If a second gradient is applied immediately after, with equal magnitude but opposite polarity, exactly the opposite phase shift will be imparted for stationary spins, and thus the effect of the first gradient will be completely undone by the second gradient (i.e. no overall phase shift). However, if the spins are in motion, they will be in a different position during the two gradients and phase cancellation will not occur. A residual phase shift will remain, which is proportional

330

Thomas and Wells

Fig. 16.1. Schematic representation of the time of flight (TOF) MRA technique. See text for details.

to the velocity along the direction of the gradient. Measurement of this phase shift therefore provides quantitative information on blood flow (3). 1.3. Arterial Spin Labelling (ASL)

ASL harnesses blood water as an endogenous contrast agent. Blood water is magnetically inverted as it flows through the feeding vessels that supply the brain (we focus here on applications of ASL for imaging cerebral perfusion; however, please note that ASL can also be applied to other organs of the body). When the labelled spins arrive in the brain they reduce the overall longitudinal magnetisation. An image is acquired where the measured signal is attenuated according to the volume of labelled water present in each voxel (the ‘labelled’ acquisition). Another image is also acquired; this time with no inversion of the inflowing blood (the ‘control acquisition’). A simple subtraction of the ‘labelled’ and ‘control’ scans yields a perfusion-weighted difference image ( M). The decrease in signal caused by the labelled water that has been delivered to the brain is relatively small; a few percent of the total brain signal. This reflects perfusion itself which is about 1 ml/100 g/s in the humans (the units of perfusion are ml/100 g/s or ml/100 g/min, which represents the volume of blood (in ml) which flows through 100 g of tissue in the specified time interval). Therefore, if blood is labelled for a second, it will contribute to around 1% of the measured signal. Here lies the principal limitation of ASL – the low SNR of the measurements.

MR Angiography and Arterial Spin Labelling

331

However, sequence and hardware developments have helped to dramatically enhance ASL’s utility in research and clinical application (see (4) for a detailed review of recent innovations in the field). Indeed a recent multi-centre viability study has provided convincing evidence for the reproducibility of ASL for CBF measurement (5). ASL is used in a wide variety of fields from functional studies to diagnosis of vascular disorder. From a methodological point of view, ASL can be broken down into two phases: the labelling/control phase and the image acquisition. 1.3.1. The Labelling and Control Phase

There are two main approaches to spin labelling: pulsed and continuous. In continuous ASL (CASL), blood is inverted for a few seconds at a single plane that intersects the main feeding arteries inferior to the brain using flow-induced adiabatic inversion. In pulsed ASL (PASL), a brief RF pulse inverts a large slab surrounding or adjacent to the imaging region. In both cases, the extent of inversion should be optimised since the measured M signal is directly proportional to the degree of labelling. The increased volume of blood water that is labelled in CASL means that a greater number of tagged spins contribute to the M image, yielding improved SNR in comparison to PASL. In particular, CASL offers a distinct SNR advantage in rodents in comparison to humans due to the marked difference in transit time from the labelling plane to the brain (∼0.25 s in rats and ∼1 s in humans). However, hardware and SAR restrictions have limited the availability of CASL and often PASL methods are used as a pragmatic alternative. Pseudo-CASL (pCASL) is a relatively new innovation that offers some of the benefits of both methods (6). The novel approach to flow-driven adiabatic inversion employs repeated rather than continuous RF pulses. The idea is to achieve the same effect as CASL without the need for a continuous RF output from the amplifier. Continuous labelling can be achieved using a separate coil positioned proximal to the neck (e.g. see (7, 8)). The labelling pulse can then be localised to the feeding arteries so that the RF does not penetrate the brain. In this way, negligible magnetisation transfer (MT) effects are induced in the imaging region. MT acts to reduce the T1 of the labelled water that has exchanged into the tissue, and hence without MT the tagged spins will have relaxed to a lesser extent upon image acquisition, providing superior contrast per labelled spin. The removal of MT also renders multi-slice imaging considerably more straightforward (see below). However, this dedicated labelling coil requires a separate channel for transmission which is not always readily available on MRI systems. Unless a separate labelling coil is used, multi-slice or 3D perfusion imaging requires a special control pulse that induces identical MT effects in the imaging volume whilst minimising any inadvertent tagging. The earliest example was a sinusoidal modulation of the RF labelling waveform during the control phase that maintained the frequency offset of the labelling phase (9). The

332

Thomas and Wells

application of such an RF pulse together with a gradient will continuously invert two planes at the same time, theoretically leaving the net magnetization of the arterial blood unaltered. The power of the amplitude modulated control is configured to produce an identical MT profile in the brain during the tag and control imaging stages, permitting multi-slice or 3D acquisitions. pCASL is a recent and more promising approach which also provides this multi-slice capability. The labelling phase of the sequence is summarised as follows: • Achieving efficient inversion is important as the SNR of the perfusion measurement is proportional to the effective labelling efficiency (i.e. the difference between the longitudinal magnetisations of blood in feeding vessels during tagging and during control). • A separate labelling coil positioned proximal to the neck can be used to continuously label blood whilst inducing negligible MT into the imaging region. • CASL or pCASL techniques can yield improved SNR in comparison to PASL methods. • Care is needed to ensure matching of MT effects between the labelling and control pulses for multi-slice imaging (when no dedicated labelling coil is used). 1.3.2. The Image Acquisition Phase

ASL requires fast imaging sequences to capture the labelled blood that has travelled to the brain before it ‘loses its tag’ due to T1 relaxation. Rapid imaging also allows labelled and controlled acquisitions to be acquired frequently in an interleaved manner to reduce the effect of physiological and scanner drift on the low SNR measurements. This makes rapid single-shot interleaved tagged and control imaging methods, such as EPI (10, 11) or spiral imaging (12), highly advantageous. Alternative readout methods based on fast-spin echo (13) or snapshot FLASH (fast low-angle shot) (14) methods have been suggested as viable alternatives. These have important applications in imaging outside the brain. A single-shot 3D sequence has recently been designed that was shown to increase the SNR of the perfusion weighted images by a factor of 2.8 in comparison to 2D EPI at the same nominal resolution. This sequence is known as 3D GRASE (15) and is becoming increasingly popular in ASL applications.

2. Materials 2.1. MRI Hardware

The techniques described in this chapter use standard MRI system hardware. An advantage of these approaches is that they do not require any additional material, i.e. no exogenous contrast agents

MR Angiography and Arterial Spin Labelling

333

need to be used and no additional pieces of equipment need to be added. Because of this, the techniques can easily be included as part of any MRI imaging session. Higher field strength systems (e.g. 3 T and above for clinical scanning and 7 T and above for experimental scanning) offer the advantage of higher SNR for both MR angiography and ASL. For ASL, it is important for the RF transmitter coil to have good coverage of the tagging region as well as the imaging region. This is usually something which is not taken into consideration in coil design (e.g. for a standard head coil, the B1 field will be optimised for the homogeneity across the brain, but will not extend far inferiorly). For pulsed ASL in particular, this can cause problems because the tagging volume must be sufficient to ensure that the labelled bolus has a large enough temporal width to generate a measurable perfusion signal difference. For the specific ASL setup, we have described below for the acquisition of CBF maps in the rat brain, a horizontal bore 2.35 T MR scanner was used with a body coil transmitter and a 1-cm diameter single-loop surface coil for reception. Halothane was used as a general anaesthetic.

3. Methods 3.1. Time-of-Flight (TOF) MRA

In order to ensure a flow-related increase in signal intensity, it is important to choose the following MR sequence parameters carefully: 1. Short TR and TE are essential to give good sensitivity to flow, suppression of static signal and maximise SNR. Typical values for TR and TE are 10–40 ms and <10 ms respectively. In addition, the TE can be chosen to correspond to the time at which fat and water are 180◦ out of phase, which causes the signals to (at least partially) cancel out and therefore reduces the appearance of fat in the images (the specific timings required depend on the Larmor frequencies of water and fat and therefore the field strength used). This will most likely increase the TE, so the merit of doing it will depend on the region of the body being scanned. 2. Positioning and orientation of the imaging slab are optimal if the shortest dimension (usually slice select) is perpendicular to the direction of flow. This will allow maximal inflow of fresh (unsaturated) blood and ensure good flow contrast (see Note 1). 3. Flow-compensated gradients should be used for slice selection and frequency encoding to fully refocus the transverse magnetisation and maximise intraluminal signal intensity (see Note 2).

334

Thomas and Wells

4. TOF can be performed as a multi-slice 2D or volume 3D acquisition. The advantages of a 2D acquisition are as follows: (i) a higher flip angle can be used, which causes better suppression of stationary tissue and therefore better contrast between this and flowing tissue, (ii) short acquisition times, (iii) and good sensitivity to slow flow. Following are the advantages of a 3D acquisition: (i) higher intrinsic SNR and (ii) thinner slices, allowing better visualisation of small vessels. For 3D TOF, the flip angle should be 15–25◦ . For 2D TOF, typical flip angles are in the range of 50–70◦ . 5. The main disadvantage of 3D acquisition techniques comes from the saturation effect on blood signal as it moves through the imaging slab. After the blood has seen several RF excitation pulses, its longitudinal magnetisation will be reduced such that contrast between the blood signal and the signal from stationary tissue will be significantly reduced. This effect is known as progressive saturation. As a compromise between 2D and 3D acquisition schemes (to retain good flow contrast while achieving high spatial resolution in the slice select direction), the hybrid technique known as multiple overlapping thin-slab acquisition (MOTSA) is often used (16). This method reduces the progressive saturation effect associated with 3D acquisitions by reducing the thickness of the 3D slabs and uses multiple slabs to keep good volume coverage. To overcome slice profile effects degrading image quality at the edges, the slabs are made to overlap during the acquisition and this overlap is accounted for in the image reconstruction. 6. As well as choosing the basic pulse sequence parameters to optimise flow contrast, there are several additional measures that can be taken to improve the MRA image quality further. To selectively suppress the signal from stationary tissue, offresonance magnetisation transfer (MT) pulses can be applied (17, 18). In the brain, the signal from grey and white matter is selectively reduced by MT pulses, whereas blood signal is affected much less. However, this comes at the cost of increased power deposition and an increased minimum TR. The effect of progressive saturation (see point 5 above) can be reduced by using ‘ramped’ RF pulses with spatially varying excitation flip angles, also known as tilted optimised nonsaturating excitation (TONE) pulses (19). The flip angle of these pulses varies linearly in the slab-select direction, with lower flip angles used where the flowing spins enter the imaging slab (to reduce saturation effects) and higher flip angles are used at the opposite side (to increase static signal saturation and compensate for the progressive saturation of flowing signal).

MR Angiography and Arterial Spin Labelling

335

Fig. 16.2. Maximum intensity projection (MIP) image of the normal human brain using TOF MRA (MOTSA acquisition scheme, transverse projection).

7. TOF MRA image volumes are often viewed as maximum intensity projections (MIPs) (Fig. 16.2). These allow excellent visualisation of 3D vascular anatomy in a 2D image. 3.2. Phase-Contrast MRA

In phase-contrast MRA (PC MRA), bipolar magnetic field gradients are added to a flow-compensated sequence to encode the velocity of moving water in the phase of its transverse magnetisation (3). These bipolar gradients have no effect on the signal from stationary spins and give flowing tissue a phase which is directly proportional to its velocity along the direction of the gradient. Quantification of blood velocity is therefore possible. 1. The key parameter in phase-contrast MRA is VENC (velocity encoding). This parameter determines the relationship between the flow velocity and the induced phase shift. It is important to choose this parameter carefully: if VENC is chosen to be too high, then the SNR and accuracy of the flow measurement will be compromised; if VENC is chosen to be too low, then phase aliasing will occur, which results in image artefacts and erroneous flow measurements. VENC is defined as the velocity which causes a 180◦ -phase shift and is therefore the maximum velocity that can be unambiguously measured. If a velocity slightly higher than VENC exists, the induced phase shift will be indistinguishable from that from

336

Thomas and Wells

a velocity slightly lower than negative VENC. For example, if VENC is set to be 180 cm/s and the system contains flowing water with a maximum velocity of 200 cm/s, this will be incorrectly assigned a value of –160 cm/s (i.e. 160 cm/s in the opposite direction; see Note 3). 2. In order to fully characterise flow in three dimensions, four acquisitions are required for PC MRA: one without flowencoding gradients and one with flow-encoding gradients along each of the x, y and z directions. As a result, PC MRA takes longer to acquire than TOF MRA (see Note 4). Subtraction of the phase images acquired by flow-encoding along the three directions from the non-flow-encoded image allows quantification of blood velocity (see Note 5). 3. Stationary tissue should always be well suppressed in PC MRA (Fig. 16.3), and the limitations associated with progressive saturation of intravascular signal which plague TOF MRA do not exist. Therefore, the imaging slab can be chosen with arbitrary thickness and orientation, and this will not affect the flow contrast. Additional measures, such as the use of MT pulses or MOTSA-type acquisition schemes (see above), are not required. 4. In order to measure the blood velocity at the same point in the cardiac cycle, ECG gating should be used. Multiple images can be acquired at several points during the cardiac cycle using a cine sequence. In this case, it can be advantageous to use different VENC values for different cardiac phases, to match the expected variations in blood velocity and so maximise the accuracy of all the measurements (20).

Fig. 16.3. Transverse a structural and b PC MRA image of the human neck. In (b), signal intensity is proportional to velocity: grey = stationary tissue; white is high velocity towards the head; black is high velocity in the opposite direction (towards the body). The carotid and vertebral arteries taking blood to the brain are easily visible as high-signal intensity, as are the jugular veins which appear dark.

MR Angiography and Arterial Spin Labelling

337

3.3. Arterial Spin Labelling 3.3.1. Preliminary Validation

It is imperative for accurate CBF quantification that any coherent signal in the subtracted, M, image is due to labelled spins. As previously mentioned, unless a separate, small, labelling coil is used, the labelling/control pulses will induce magnetization transfer (MT) effects within the imaging plane. Magnetisation transfer is an ever-present consideration in ASL. With naive application, MT can significantly influence the measured M signal and confound CBF quantification. Phantom experiments using agar for example (which is susceptible to MT effects) should be performed prior to in vivo data collection to check that the labelled and control images cancel, particularly when implementing multi-slice or 3D sequences. A valuable in vivo experiment to check the accuracy of the method can be performed by acquiring ASL data following intravenous injection of gadolinium. The exogenous contrast agent will markedly reduce the T1 1 of the blood which will then rapidly relax back to equilibrium in a few milliseconds following labelling. The subtracted image should then be made up of noise with no coherent signal.

3.3.2. ASL to Measure CBF in the Rat Brain

Here, we have described the specific implementation of ASL in our lab, to cover the methodological aspects in detail. 1. Anaesthesia is induced using 3% halothane in 100% O2 and is maintained via a nose cone at 2% halothane in 100% O2 , whilst the animal is placed on a custom-designed Perspex probe (see Note 6). Temperature is monitored using a rectal probe and should be carefully maintained throughout the procedures. There is considerable evidence that changes in core temperature can markedly influence CBF (21). Breathing rate is monitored using ECG electrodes. 2. A volume coil is used for spin labelling and to apply the imaging RF pulses. A surface coil is used for signal reception to maximise SNR and is positioned immediately adjacent to the top of the animal’s head (see Note 7). 3. A single-slice CASL sequence is used. This gives the best SNR and measurement accuracy, but obviously provides limited spatial coverage. The slice of interest should be positioned in the centre of the magnet. The probe is then tightly secured within the scanner and halothane concentration is reduced to 1.25% in 60% N2 O and 40% O2 . 4. A slice-selective shim is performed on the imaging slice to ensure good EPI image quality (see Note 8). The flowinduced adiabatic inversion process is relatively insensitive to off-resonance effects and so a poor shim in the labelling region can be tolerated.

338

Thomas and Wells

5. In these experiments, a volume coil is used to transmit the tagging and control pulses. An alternating adiabatic spin tagging pulse (total duration 3 s) is applied to minimise eddy current effects (see Note 9). The offset frequency of the labelling pulse oscillates in accordance with the sliceselect gradient to maintain a constant tagging plane. The labelling pulse is applied 2 mm caudal to the cerebellum, perpendicular to the carotid and vertebral arteries, to ensure efficient spin tagging. An axial scout image is acquired in order to determine the correct offset frequency for accurate positioning of the labelling plane (2 mm caudal to the cerebellum). 6. Knowledge of several parameters is required for CBF quantification (it should be noted that this will depend on the quantification model used. In this example, we have used the model described in (22)). Some can be estimated from the acquired data and others must be taken from the literature (see methods point 12 below). At this stage, the experimenter will implement a single- or multi-post-labelling delay (PLD) approach (see Note 10). In this example, we acquire images at six different PLDs (0.05, 0.1, 0.2, 0.3, 0.5 and 0.8 s) (see Note 11). 7. Single-slice images are acquired using single-shot EPI. Acquisition parameters are as follows: slice thickness = 2 mm, image matrix size = 128×64, field of view = 40 × 20 mm2 , TE = 36 ms, and inter-experiment delay = 2 s. 8. Diffusion gradients should be included in the sequence to reduce the effect of vascular contamination in the perfusion-weighted images (see Note 12). 9. It is crucial that tagged and control acquisitions be acquired successively and should be interleaved with one another. This is in order to reduce the effect of physiological and scanner drift on the low-SNR M images. For example, if one acquires 10 averages (at the six PLD times specified above), the control acquisition should follow the tagged acquisition at a PLD of 0.05 s, followed by an additional control and then tagged EPI acquisition, both at a PLD of 0.1 s. This process is repeated at increasing PLD. Once tagged and control images have been taken at a PLD time of 0.8 s, then the whole process is repeated for a total of 10 averages. 10. Single-average images should be acquired and saved for post-processing, rather than combined and averaged on the console (see Note 13). 11. The individual average control and tagged images are masked and de-noised prior to CBF quantification using

MR Angiography and Arterial Spin Labelling

339

independent component analysis techniques to improve the precision of the CBF estimates (23). This postacquisition pre-processing step is particularly effective when applied to ASL images acquired at multiple PLD prior to model fitting. 12. CBF is quantified according to the model proposed in Ref. (22) and is shown in Eq [1] and [2] below. The list that follows defines each of the terms in the equation and how they are determined. f = −λ(Mtag − Mctrl ) 2αM0b 1 C

C(T1 ns , T1s , T1a , δ, δa)

[1]

=T

1 ns exp(−δ/T1a )

  exp(min(δ − w, 0)/T1 ns ) − exp(−w/T1 ns ) 1−TT11sns

+ T1a [exp(min(δa − w, 0) − δa)/T1a ) − exp((min(δ − w, 0) − δ)/T1a )]

[2]

f = estimated CBF in ml/min/100 g. λ= blood:brain partition coefficient, commonly assumed to be 0.9 in grey matter and 0.8 in white matter (24). Mtag –Mctrl = subtracted perfusion weighted signal ( M). Mb 0 = equilibrium magnetisation of the brain tissue. This can be estimated from the recovery of the control magnetisation as a function of post-labelling delay time (w) if multi-PLD ASL images are acquired. Otherwise this should be calculated from a separate inversion recovery experiment. α = labelling efficiency. T1s = T1 relaxation constant of the tissue during application of the labelling/control pulse. This can be estimated from the recovery of the control magnetisation as a function of post-labelling delay time (w) if multi-PLD ASL images are acquired. Otherwise this should be calculated from a separate experiment, where a control pulse is applied and images acquired at a range of times after the end of the pulse. T1 ns = T1 relaxation constant of the tissue following application of the labelling/control pulse. This can be estimated from the recovery of the control magnetisation as a function of post-labelling delay time (w) if multi-PLD ASL images are acquired. Otherwise this should be calculated from a separate inversion recovery experiment (see Note 14). T1a = longitudinal relaxation constant of the blood. Ideally this should be measured for each system using the technique described in (25) but may be taken from the literature. w = PLD time. This is defined in the imaging protocol.

340

Thomas and Wells

Fig. 16.4. CASL M images acquired with a range of post-labelling delays in the rat brain (coronal slice). The signal intensity in these images is directly proportional to CBF. Number of averages = 30.

δ = tissue transit time. The tissue transit time affects the PLDdependant M signal in a way that is difficult to distinguish from CBF changes (Fig. 16.4). Therefore, given the low SNR of the M signal, in practice it is difficult to estimate δ, even if multi-PLD data are acquired, as a small degree of noise can introduce marked imprecision into the δ and CBF estimates. For this reason, δ is often a fixed parameter or is measured using diffusion sensitised ASL measurements (26, 27). There is evidence for fast rapid exchange of labelled water into the tissue in rats (7, 28–30). Therefore, assuming instantaneous exchange of labelled water into the tissue (i.e. setting δ = δa) is a sensible approach in rodent studies. δa = arterial transit time. This can be simultaneously estimated with CBF by fitting ASL images acquired at multiple delay times to the model. We have measured δa to be approximately 0.2–0.3 s in the rat cortex.

4. Notes 1. The amount of contrast between flowing and stationary tissue depends on several factors: tissue-specific parameters, such as T1 and T2 , sequence-specific parameters, such as flip angle and TE/TR, and geometric parameters, such as slice/slab thickness. Some of these parameters will be determined by the requirements of the scan (e.g. volume coverage required, maximum total scan time permissible), some by the hardware limits (minimum TE/TR); the remaining parameters can then be optimised to get the best contrast for a particular study. For example, for a multi-slice 2D TOF acquisition acquired with axial slice orientation, one can choose whether to acquire the slices in the inferior to superior direction or in superior to inferior. In the former case, the most inferior slices will demonstrate the best flow contrast and vice versa. This choice will depend on which region of the imaging volume is of most interest for that particular study.

MR Angiography and Arterial Spin Labelling

341

2. Typically, first-order (velocity) flow compensation is used. Higher order (acceleration) flow compensation can be used, but it results in a significant increase in minimum TE/TR, and the benefit must be balanced against the loss in SNR and contrast caused by this. 3. In practice, it is always prudent to choose a VENC value slightly higher than the largest expected velocity. VENC depends on the bipolar gradient amplitude and duration (33). 4. In order to fit the bipolar velocity-encoding gradients into the sequence, the minimum echo time has to be extended compared to a TOF MRA acquisition. However, it is important to keep TE as short as possible to minimise the development of phase due to off-resonance velocityindependent effects. 5. An alternative to acquiring images with velocity-encoding gradients on and off is to acquire a pair of images with bipolar gradients of opposite polarity. Doing this reduces the required amplitude of the gradients (or allows their duration to be reduced) but also means that pairs of images must be acquired for each direction of velocity encoding, therefore increasing the total scan time. 6. ASL is extremely sensitive to subject movement; even small alterations of the brain position within the FOV can have serious implications for accurate CBF measurement. Much can be done at this early stage in the imaging protocol to minimise this confounding factor. Rigid motion prevention through the use of ear bars can significantly improve stability, though these should be applied cautiously and carefully to minimise any possible discomfort to the animal. Optimal positioning of ear bars can be a tricky procedure, but perseverance will be rewarded with reduced motion artefacts during imaging. For this reason, the choice of anaesthetic should be carefully considered in order to achieve a sufficient depth of anaesthesia throughout. 7. To improve the SNR, a surface coil should be centred on the slice of interest and should not just rest on the fur, but should be positioned as close to the head as possible without causing any discomfort to the animal. 8. Achieving a good shim can be important to maintain anatomical accuracy and reduce drop-out affects, particularly for EPI at high field strength (e.g. 9.4 T). 9. The application of a sustained gradient during the labelling and control phase (typically 1–4 s for CASL) can induce eddy currents which may affect the acquired image as a function of the PLD time. In our experiments, these eddy

342

Thomas and Wells

currents caused a non-rigid warping of the brain in the phase-encoded direction. A straightforward way of reducing these effects is to use an alternating labelling/control pulse that switches gradient polarity and resonant frequency in sync to maintain a constant tagging or control plane. Spoiler gradients at the end of the tagging pulse can also help reduce eddy current effects. 10. The post-labelling delay (PLD) time is the time between the end of the tagging pulse and image acquisition. ASL images can be acquired over a range of PLD values to generate dynamic data that reflect the delivery and progression of the tagged bolus to and within the brain tissue. This improves the accuracy of the CBF measurements and also allows estimation of the transit time. The transit time is the time taken for the labelled blood to travel from the labelling plane to the tissue of interest and is often altered in cerebral vascular disease. However, the multi-PLD approach reduces the sensitivity and temporal resolution of the technique in comparison to repeated measurements at a single PLD time. For the single PLD approach, the delay time should be chosen to be greater than the transit time to the tissue of interest (22). For example, if the range of transit times in the normal cortex is 0.2–0.5 s, then the PLD should be 0.5 s. The choice of single- or multi-PLD measurements depends on the application. Single-PLD measurements are performed in fMRI experiments because of the demand for high temporal resolution and sensitivity. Multi-PLD measurements may be necessary when there is a range of transit times across the imaging volume (in an animal model of cerebral ischaemia for example). 11. Multi-PLD can be optimised to improve the precision of CBF and transit time estimation depending on the transit time within the tissue of interest (31). However, if there are a range of transit times within the imaging region, then the value of this approach may be limited. If a multi-PLD approach is employed (with an ASL sequence that induces MT in the imaging region), it may be possible to use the recovery of the control images as a function of PLD to measure T1n and M0 (parameters required for CBF quantification). However, it is important to consider that the optimal PLD times for model fitting the M signal for precise CBF and δa quantification may be very different to those for model fitting the recovery of the control images to calculate T1n and M0 with a high degree of certainty. It is well known in the MR literature that it is necessary to sample the asymptote of the longitudinal relaxation

MR Angiography and Arterial Spin Labelling

343

curve for precise T1 estimation. Therefore, whilst measurements at long PLD may not be very useful for CBF and δa measurement (particularly given the low SNR of the M images and the rapid decay of the labelled water), such a measurement may be necessary for precise T1 and M0 measurement. 12. This is not so crucial in the rat brain where the labelled water rapidly exchanges into the tissue after short PLD times (7, 28, 30). However, the proportion of intra- to extravascular ASL signal tends to be significantly greater in humans, and cerebral perfusion can be significantly overestimated without the inclusion of vascular crusher gradients in the ASL sequence. Silva et al. (28) and Wang et al. (32) have measured the ASL signal with diffusion gradients over a range of b values in rats and humans respectively. These studies provide evidence that the signal from the vascular compartment is effectively suppressed using diffusion b-values of greater than 20 s/mm2 . However, one should bear in mind that the extent of vascular suppression is likely to be dependent on the time between the diffusion gradients as labelled blood in the microvasculature is more likely to have flowed parallel to the direction of the diffusion gradient with longer diffusion time. 13. By acquiring single average data (rather than averaging the images acquired under the same conditions on the system), it is possible to remove images collected over the course of the experiment that have been significantly influenced by systematic errors (outlier images). Ideally, this may not be necessary if such errors (due to subject movement or scanner instability) have been minimised in the experimental protocol. 14. Experimenters should take note that constant T1 and M0 of the tissue is not a valid assumption in some applications (in animal models of stroke, for example, where T1 has been shown to undergo marked changes following ischaemia). General Note. Background suppression has been shown to significantly reduce physiological noise, yielding more precise CBF estimates. An inversion pulse is applied to the imaging slab before the tagging pulse. An image is acquired at the null point of the tissue. Two inversion pulses can be used with careful sequence timing to null two different spin populations (with different T1 values, e.g. grey matter and CSF). However, the experimenter should consider that the signal intensity of the low SNR images (as we acquire at the null point of the tissue) is likely to increase due to noise rectification if the magnitude of the images is taken. The increase will be greater in the labelled images in comparison to the control signal and this may bias CBF estimates.

344

Thomas and Wells

General Note. Difficulties can arise when implementing pulsed ASL techniques due to the limited spatial coverage of the applied RF labelling slabs. For example, the control phase of FAIR should be a global inversion. However, if the inversion is not truly global (e.g. not the entire body), then the fully relaxed spins may flow into the imaging volume during the inflow time (IT) due to the high blood flow rates in rodent vasculature. This can confound accurate CBF measurement. References 1. Ruggieri, P. M., Laub, G. A., Masaryk, T. J., Modic, M. T. Intracranial circulation: Pulse-sequence considerations in three-dimensional (volume) MR angiography. Radiology 1989;171(3):785–791. 2. Keller, P. J., Drayer, B. P., Fram, E. K., Williams, K. D., Dumoulin, C. L., Souza, S. P. MR angiography with two-dimensional acquisition and threedimensional display. Work in progress. Radiology 1989;173(2):527–532. 3. Dumoulin, C. L., Hart, H. R., Jr. Magnetic resonance angiography. Radiology 1986;161(3):717–720. 4. Wolf, R. L., Detre, J. A. Clinical neuroimaging using arterial spin-labeled perfusion magnetic resonance imaging. Neurotherapeutics 2007;4(3):346–359. 5. Petersen, E. T., Golay, X. (2008) The QUASAR Reproducibility Study. Is arterial spin labeling ready for the prime time? Preliminary results from the QUASAR reproducibility study. Proceedings of 16th Meeting of ISMRM 191. 6. Dai, W., Garcia, D., de Bazelaire, C., Alsop, D. C. Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields. Magn Reson Med 2008;60:1488–1497. 7. Silva, A. C., Zhang, W. G., Williams, D. S., Koretsky, A. P. Estimation of water extraction fractions in rat brain using magnetic resonance measurement of perfusion with arterial spin labeling. Magn Reson Med 1997;37: 58–68. 8. Zaharchuk, G., Ledden, P. J., Kwong, K. K., Reese, T. G., Rosen, B. R., Wald, L. L. Multislice perfusion and perfusion territory imaging in humans with separate label and image coils. Magn Reson Med 1999; 41:1093–1098. 9. Alsop, D. C., Detre, J. A. Multisection cerebral blood flow MR imaging with continuous arterial spin labeling. Radiology 1998;208:410–416.

10. Edelman, R. R., Siewert, B., Darby, D. G., Thangaraj, V., Nobre, A. C., Mesulam, M. M. et al. Qualitative mapping of cerebral blood-flow and functional localization with echo-planar MR imaging and signal targeting with alternating radio-frequency. Radiology 1994;192:513–520. 11. Kwong, K. K., Chesler, D. A., Weisskoff, R. M., Donahue, K. M., Davis, T. L., Ostergaard, L. et al. MR perfusion studies with T1 weighted echo-planar imaging. Magn Reson Med 1995;34:878–887. 12. Yang, Y. H., Frank, J. A., Hou, L., Ye, F. Q., Mclaughlin, A. C., Duyn, J. H. Multislice imaging of quantitative cerebral perfusion with pulsed arterial spin labeling. Magn Reson Med 1998;39:825–832. 13. Chen, Q., Siewert, B., Bly, B. M., Warach, S., Edelman, R. R. STAR-HASTE: Perfusion imaging without magnetic susceptibility artifact. Magn Reson Med 1997;38: 404–408. 14. Calamante, F., Lythgoe, M. F., Pell, G. S., Thomas, D. L., King, M. D., Busza, A. L. et al. Early changes in water diffusion, perfusion, T1 and T2 during focal ischaemia in the rat studied at 8.5t. Magn Reson Med 1999;41:479–485. 15. Günther, M., Oshio, K., Feinberg, D. A. Single-shot 3d imaging techniques improve arterial spin labeling perfusion measurements. Magn Reson Med 2005;54: 491–498. 16. Parker, D. L., Yuan, C., Blatter, D. D. MR angiography by multiple thin slab 3d acquisition. Magn Reson Med 1991;17:434–451. 17. Edelman, R. R., Ahn, S. S., Chien, D., Li, W., Goldmann, A., Mantello, M. et al. Improved time-of-flight MR angiography of the brain with magnetization transfer contrast. Radiology 1992;184(2):395–399. 18. Pike, G. B., Hu, B. S., Glover, G. H., Enzmann, D. R. Magnetization transfer timeof-flight magnetic resonance angiography. Magn Reson Med 1992;25:372–379.

MR Angiography and Arterial Spin Labelling 19. Atkinson, D., Brant-Zawadzki, M., Gillan, G., Purdy, D., Laub, G. Improved MR angiography: Magnetization transfer suppression with variable flip angle excitation and increased resolution. Radiology 1994;190(3):890–894. 20. Swan, J. S., Weber, D. M., Grist, T. M., Wojtowycz, M. M., Korosec, F. R., Mistretta, C. A. Peripheral MR angiography with variable velocity encoding. Work in progress. Radiology 1992;184(3):813–817. 21. Li, M., Miao, P., Yu, J., Qiu, Y., Zhu, Y., Tong, S. Influences of hypothermia on the cortical blood supply by laser speckle imaging. IEEE Trans Neural Syst Rehabil Eng 2009;17(2):128–134 (Advance online publication). 22. Alsop, D. C., Detre, J. A. Reduced transittime sensitivity in noninvasive magnetic resonance imaging of human cerebral blood flow. J Cereb Blood Flow Metab 1996; 16:1236–1249. 23. Wells, J. A., Thomas, D. L., Connelly, A., King, M. D., Lythgoe, M. F., Calamante, F. (2007) Reduction of Random and Systematic Errors in ASL Quantitative Perfusion Maps Using Image Denoising. Proceedings of 15th Meeting of ISMRM 3484. 24. Herscovitch, P., Raichle, M. E. What is the correct value for the blood-brain partition coefficient for water? J Cereb Blood Flow Metab 1985;5:65–69. 25. Thomas, D. L., Lythgoe, M. F., Gadian, D. G., Ordidge, R. J. In vivo measurement of the longitudinal relaxation time of arterial blood (T1a) in the mouse using a pulsed arterial spin labeling approach. Magn Reson Med 2006;55:943–947. 26. Wang, J., Alsop, D. C., Song, H. K., Maldjian, J. A., Tang, K., Salvucci, A. E. et al. Arterial transit time imaging with

27.

28.

29.

30.

31.

32.

33.

345

flow encoding arterial spin tagging (FEAST). Magn Reson Med 2003;50:599–607. Petersen, E. T., Lim, T., Golay, X. Modelfree arterial spin labeling quantification approach for perfusion MRI. Magn Reson Med 2006;55:219–232. Silva, A. C., Williams, D. S., Koretsky, A. P. Evidence for the exchange of arterial spin labeled water with tissue water in rat brain from diffusion-sensitized measurements of perfusion. Magn Reson Med 1997;38: 232–237. Zaharchuk, G., Bogdanov, A. A., Jr., Marota, J. J. A., Shimizu-Sasamata, M., Weisskoff, R. M., Kwong, K. K. et al. Continuous assessment of perfusion by tagging including volume and water extraction (CAPTIVE): A steady-state contrast agent technique for measuring blood flow, relative blood volume fraction and the water extraction fraction. Magn Reson Med 1998;40:666–678. Wells, J. A., Lythgoe, M. F., Choy, M., Gadian, D. G., Ordidge, R. J., Thomas, D. L. Characterizing the origin of the arterial spin labelling signal in MRI using a multiecho acquisition approach. J Cereb Blood Flow Metab 2009;29(11):1836–1845 (Advance online publication). Xie, J., Gallichan, D., Gunn, R. N., Jezzard, P. Optimal design of pulsed arterial spin labeling MRI experiments. Magn Reson Med 2008;59:826–834. Wang, J., Fernandez-Seara, M. A., Wang, S., St Lawrence, K. S. When perfusion meets diffusion: In vivo measurement of water permeability in human brain. J Cereb Blood Flow Metab 2006;27(4):839–849. McRobbie, D. W., Moore, E. A., Graves, M. J., Prince, M. R. MRI from Picture to Proton, 2nd ed. Cambridge: Cambridge University Press; 2007, 265–270, Chapter 13 in.

wwwwwww

Section III Specific Applications

wwwwwww

Chapter 17 MRI Phenotyping of Genetically Altered Mice Jason P. Lerch, John G. Sled, and R. Mark Henkelman Abstract The laboratory mouse, with its genetic similarity to humans and rich set of tools for manipulating its genome, has emerged as one of the key models for experimental investigation of the genotype/phenotype relationships in mammals. Recent innovations have made MRI an increasingly popular tool for examining the phenotype of genetically altered mice. Advances in field strengths, mouse handling, image analysis and statistics have contributed greatly in this regard. In this chapter, we illustrate the methods necessary to achieve high-throughput phenotyping of genetically altered mice using multiple-mouse MRI combined with advanced image analysis techniques and statistics. Key words: Magnetic resonance imaging, phenotyping, mouse models, genetics.

1. Introduction The completion of the human genome sequence in the century just past presents a major challenge for biomedical research in the twenty-first century: that of understanding the relationship of detailed genetic sequences both to the normal development of the human individual and to the propensity for disease development. From a biomedical perspective, this genotype/phenotype relationship needs to be understood for the human species. However, the limited ability to do genetic experiments in the human means that much of this relationship will be worked out instead in the mouse. Mice have very similar genetics to that of the human with a 99.5% ability to recognize equivalent genes from one species to the other (1). Mice show many of the symptoms of both the rare and common human diseases. Existing inbred strains of mice provide homogeneous genetic backgrounds making genetic M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_17, © Springer Science+Business Media, LLC 2011

349

350

Lerch, Sled, and Henkelman

modifications readily identifiable. There exist remarkably sophisticated genetic tools for manipulating genetics in the mouse. Finally, the rapid reproduction time and the comparatively low housing costs make the mouse an ideal model for experimental investigation of the genotype/phenotype relationships in the mammal. Magnetic resonance imaging (MRI) is an excellent modality for investigating genetically altered animals. It is capable of whole brain coverage, can be used in vivo and provides multiple contrast mechanisms for investigating different aspects of neuroanatomy and physiology. The advent of high-field scanners along with the ability to scan multiple mice simultaneously (2–4) allows for rapid phenotyping of novel mutations in ways heretofore not possible.

2. Materials 2.1. Mice

2.2. MRI Hardware

Groups of genetically or environmentally modified mice are otherwise identical to control animals not having the modification. Typically, inbred background stains are chosen due to their highly conserved neuroanatomy (5). 1. An MR imager of 7 T or higher field strength with as high a performance gradient system as can be obtained. 2. An array of closely coupled radiofrequency coils and an array of spectrometers for multiple mouse imaging (e.g. Varian Direct Drive) (4). The coils need to be isolated from each other and operate in both transmit and receive. Spatial uniformity of coils is important for automated computer image processing (see Note 1). 3. Mouse holders incorporating physiological monitoring (cardiac, respiration, temperature, blood oxygenation, etc.), as well as gaseous anaesthetic and oxygen (6).

2.3. Computer Hardware

1. A cluster of Linux workstations containing at least 2 GB of memory per processor (see Note 2). 2. Networked disk storage with at least 3 GB of storage capacity per brain to be analysed. 3. Visualization workstations to analyse the data (see Note 3).

2.4. Software

1. A modern Linux distribution with a full set of C/C++ compilers and scripting languages (Bash, Perl, Python). 2. Sun Grid Engine (http://gridengine.sunsource.net) for distributing work across the Linux cluster.

MRI Phenotyping of Genetically Altered Mice

351

3. The MINC toolkit (http://www.bic.mni.mcgill.ca/ software) and ANIMAL registration software (7, 8) (see Note 4).

3. Methods Mouse phenotyping using MRI can be divided into four subparts (1): mouse preparation (2), MR acquisition (see Figs. 17.1 and 17.2) (3), image processing (see Fig. 17.3) and (4) statistical analysis. The goal is to acquire the best data possible within the available time, to accurately define the anatomical correspondence between the different images and to identify any statistically significant differences between the groups of mice. Differences may exist in MRI signal intensities, derived measures such as fractional anisotropy or metrics of brain shape. Image registration, the task of finding anatomical correspondence between the image, can be divided into three stages: an initial rigid body alignment of all images to a pre-existing target

Fig. 17.1. The in vivo MR imaging setup consists of a custom-moulded sled (a), upon which the mouse is placed and enclosed within a test tube (b). Up to 16 mice can be scanned at once (c). Once inside the scanner, gas anaesthesia is continuously administered and key bodily functions monitored (d). Three slices through an example image are shown in (e).

352

Lerch, Sled, and Henkelman

Fig. 17.2. The ex vivo consists of 16 independent solenoid coils assembled into an array (a). Two example slices of an overnight acquisition of a fixed brain are shown in (b).

establishing the pose and coordinate system of the analysis, pairwise 12-parameter registration to create a groupwise atlas of all images in the dataset being analysed and a series of iterative nonlinear registrations resulting in a model-independent atlas of that dataset. The points below will address these steps in greater detail; it is important to note, however, that in practice they are all incorporated into a single pipeline environment and each step is not separately executed by the user (see Note 5). 3.1. Mouse Preparation

1. Live mice. Live mice are anaesthetized with 3% isoflourane in warm air, injected with 0.5 ml of saline to maintain hydration, shaved along the chest to provide contact with the ECG electrodes, placed on a moulded sled to provide uniform posture, restrained with a Velcro strap over the head and inserted into 50 ml centrifuge tubes with a hole drilled in the end to provide anaesthetic. After insertion into the RF coils in the magnet, 1% isoflourane mixed with oxygen is provided to maintain the anaesthesia. Warmed air is blown into the magnet to maintain an ambient temperature of 27 ◦ C. 2. Fixed Brains. To provide better resolution and SNR and to avoid motion artefacts, mouse brains are fixed and scanned overnight. Mice are anaesthetized with a combination of Ketamine (Pfizer, Kirkland, QC) (100 mg/kg) and Rompun

MRI Phenotyping of Genetically Altered Mice

353

Fig. 17.3. This figure illustrates the derivation of local volume changes used throughout the study. Two line drawings of the brain are shown on top, one with an increased cortical volume and the other with decreased cortical volume. These images are then aligned to create a consensus average, shown in the middle with the deformation grid superimposed. The deformations are then inverted to map back to the individual brains; the distortions in the grid can be seen in the bottom of the figure, with shaded areas indicating local volume increase and decrease inside the brain.

(Bayer Inc., Toronto, ON) (20 mg/kg) via intraperitoneal injection. Thoracic cavities are opened and animals are perfused through the left cardiac ventricle with 30 mL of phosphate-buffered saline (PBS) (pH 7.4) at room temperature (25 ◦ C) at a rate of approximately 100 mL/h (see Note 6). This is followed by infusion with 30 mL of iced

354

Lerch, Sled, and Henkelman

4% paraformaldehyde (PFA) plus 2 mM ProHance in PBS at the same rate. Following perfusion, the heads are removed along with the skin, lower jaw, ears, and the cartilaginous nose tip. The remaining skull structures are allowed to postfix in 4% PFA plus 2 mM ProHance at 4 ◦ C for 12 h. The skulls are then transferred to solution containing 1X PBS + 0.02% sodium azide + 2 mM ProHance for 4 days at 15◦ C. MRIs can then be acquired no sooner than 4 days and not longer than 2.5 months postfixation. 3.2. MR Acquisition

1. MR images are acquired using pulse sequences that provide appropriate contrast for the phenotype that is being investigated. For anatomical images and analyses, fast spin echo (turbo spin echo) or gradient echo scans are efficient and typically used. Isotropic voxels from 3D acquisitions are preferable to multi-slice imaging when computer analysis is being used. Whatever the acquisition, excess SNR should be spent on improved resolution until the SNR is approximately 20, the value which provides an optimal tradeoff for computer registration (9). Representative examples of scan quality can be seen in Figs. 17.1 and 17.2, respectively, for in vivo and ex vivo protocols (see Note 7). 2. Calibration and compensation for geometric distortion may be required to obtain geometric accuracy comparable to the morphological differences of interest. The presence of geometric distortions is readily assessed by scanning an object of known geometry immersed in solution. Postprocessing for geometric distortion is typically available from the MRI scanner vendor or can be implemented by identifying known landmarks on a test object (see Note 8).

3.3. Image Processing

1. Six-parameter registration: the goal is to rigidly align all scans in a study into the same coordinate space. 1.1. Create a series of Gaussian blurred representations of the input data and the atlas (see Note 9). 1.2. Align each blurred representation to the equivalently blurred version of the atlas using six degrees of freedom (three rotations, three scales; referred to as lsq6 from hereon). The first blur can be initialized with either a centre-of-gravity estimation or an identity transform (see Note 9); each subsequent registration is initialized with the previous registration transform. 1.3. Resample the data with the final transform from step 1.2. 1.4. Remove any intensity inhomogeneities using the N3 algorithm (10).

MRI Phenotyping of Genetically Altered Mice

355

2. At this point, all the scans are in the same space and orientation. The next step is to create the best possible linear model of all the brains contained in this dataset. 2.1. Create a series of Gaussian blurred representations of the lsq6-resampled datasets using FWHM 0.3, 0.2 and 0.15 mm kernels. Take the gradient of the blur for the FWHM 0.2 mm dataset. 2.2. Align each scan towards every other scan in the dataset. That is, assuming five scans in the study, scan 1 will be separately aligned to scans 2, 3, 4 and 5. For each pair, three separate registrations are computed, starting with the FWHM 0.3 mm kernel initialized by an identity transform, followed by the gradient of the FWHM 0.2 mm kernel initialized by the previous registration and finalized by the FWHM 0.15 mm kernel. Each registration has 12 degrees of freedom (three rotations, three translation, three scales and three shears, referred to as lsq12 from hereon). The average transform for all pairs is then computed, i.e. staying with the above cited example, scan 1 was separately aligned towards scans 2, 3, 4 and 5; the four resulting transforms are then averaged (see Note 10). 2.3. Resample each scan with the transform resulting from step 2.2. 2.4. Average all of the resampled images from step 2.3 to create a study-population-specific atlas (lsq12 atlas). 3. The third step is to refine the registration by locally deforming the scans to bring them into exact correspondence. This is done using an iterative process, each time aligning the scans towards the model created by the previous iteration. 3.1. Create a series of blurred representations of the individual lsq12 resampled images, as well as the lsq12 atlas. 3.2. Deform each lsq12 resampled image towards the lsq12 atlas. 3.3. Resample each image with the transform from step 3.2. 3.4. Average all of the resampled images from step 3.3 to create the first non-linear version of the studypopulation-specific atlas. 3.5. Repeat steps 3.1 through 3.4, this time substituting the first non-linear resampled images and atlas for the lsq12 images and atlas. This process is repeated six times with the registration parameters taken from Table 17.1.

356

Lerch, Sled, and Henkelman

Table 17.1 Parameters used for non-linear registration Generation

Blurring kernel (mm)

Step size (mm)

Registration target

Iterations

1

0.3

0.7

LSQ12

20

2

0.2

0.6

Generation 1

6

3

0.2

0.5

Generation 2

8

4

0.2

0.24

Generation 3

8

5

0.1

0.12

Generation 4

8

6

0.06

0.06

Generation 5

8

4. The fourth step is to prepare data for statistical analysis. The end result of this step is a series of parametric maps that define the shape of the brain as accurately and comprehensively as possible. 4.1. Find any remaining linear components of the final non-linear transform that maps the lsq12 aligned images to the non-linear average. This is accomplished by selecting 20,000 randomly placed points in the brain and extracting their spatial coordinates before and after non-linear resampling. The best possible 12-parameter fit is then estimated from this point set and subtracted from the final nonlinear transform, resulting in a displacement field containing the pure non-linear component of the registration. 4.2. These pure non-linear displacement fields are then centred with respect to the population mean by computing the mean displacement vector at each voxel and subtracting it from every scan’s displacement field. 4.3. The centred displacement fields are then smoothed using a series of Gaussian blurring kernels (see Note 11). 4.4. Next, derive two metrics from each of the smoothed displacement fields: the Jacobian determinant (11) a measure of local volume expansion and contraction, as well as the magnitude of the deformation at each voxel (see Note 12). 4.5. Multiply the Jacobian determinants with the scale factor determined from the 12 parameter registration. This results in two parametric maps per smoothed displacement field: the normalized Jacobian determinant, which describes local volume differences after

MRI Phenotyping of Genetically Altered Mice

357

overall brain size has been normalized, and the absolute Jacobian determinant, which incorporates both the local volume change and brain size differences at every voxel. 4.6. Align a segmented atlas dividing the brain into 62 separate structures (114 if bilateral structures are separated into left and right) onto the final non-linear study-population-specific atlas (12). 4.7. Using the smallest smoothed displacement vector, integrate the Jacobian determinant multiplied by the voxel size across each structure to give a structure volume in cubic millimetre. 4.8. Use the transform taking each scan from native (scanner) space to the non-linear average to resample any associated data with the scan, i.e. this might include blood flow maps derived from Arterial Spin Labelling, fractional anisotropy or related measures from Diffusion Tensor Imaging, etc. 3.4. Statistical Analysis

The registration pipeline described above results in (1) a table of anatomical structure volumes (2), measures of local expansion and contraction at every voxel (3), deformation vectors at every voxel, and (4) associated data resampled into the space of the nonlinear average. The next set of steps determines whether there are any statistically significant differences in these measures between genotypes or other natural groupings of the animals (environment, gender, etc.). 1. Create a spreadsheet containing information about each scan (genotype, bodyweight, age of the animal, etc.). 2. Load the spreadsheet in software capable of performing statistical analyses across imaging volumes – we use RMINC for this purpose (see Note 13). 3. All subsequent tests are performed using variants of linear models (see Note 14). 3.1. Comparisons of differences in overall brain volume and volumes of individual brain structures. 3.2. Analyses of the same model at every voxel using the log-transformed Jacobian determinants of the blurred displacement fields. 3.3. Analyses of co-registered data (perfusion maps, FA maps, etc.) 4. Comparisons of the deformation vectors themselves are carried out using Hotelling’s T2 statistic – though this is only applicable for the simple two-group comparison.

358

Lerch, Sled, and Henkelman

5. Once all statistical tests are done, collect all of the resulting p-values and threshold them using the False Discovery Rate (FDR) (13) in order to control the number of false positives due to multiple comparisons (see Note 15). 6. Imaging data are typically analysed in an exploratory manner through a continuous process of refining statistical models. After the first five steps, the key is to now investigate individual regions that were found significant or, as importantly, were hypothesized to be significant, but did not survive multiple comparisons, and to examine the distributions of the data at specific voxels. In particular, one should check for 6.1. Outliers: outliers can significantly affect data analyses. We have found two main reasons for outliers in our data: (1) misregistration, which happens rarely, but still needs to be checked; and (2) biological; at times a whole litter of mice will have considerably smaller brains for no discernible reasons. 6.2. Necessary covariates: confounding factors can easily spoil an analysis; for example, one of the most difficult confounds we encounter is that of bodyweight/brain volume covariation. If the brains in a particular mutation are smaller, one needs to rule out that the whole mouse is not smaller with the implication that there is nothing unique about the findings in the brain. Other important checks are for gender balance, consistent ages, etc.

4. Notes 1. Considerable SNR advantages can be gained by construction of application-specific close-fitting solenoid coils for specimen scanning. An array of custom-built 14-mm diameter solenoid coils with a length of 18.3 mm and over wound ends provides for convenient scanning of multiple brain-in-skull specimens (see Fig. 17.2). 2. The software that we use runs just as happily on OS X and can be made to work on Windows as well. In our experience, however, a Linux cluster is the easiest to maintain for these types of computational problems. 3. If using Linux, we highly recommend using NVidia graphics cards with the NVidia OpenGL drivers; for all the system administration headaches that these proprietary drivers

MRI Phenotyping of Genetically Altered Mice

359

can cause, they still do provide the best performance for 3D rendering on Linux. 4. The recipe given herein for phenotyping genetically altered mice is quite obviously based on our work. Needless to say, the overall approach that we are using is clearly compatible with other software toolkits. 5. Our pipeline environment, named MICe-build-model (https://wiki.phenogenomics.ca/display/MICePub/MICebuild-model), is freely available to the research community; interested users should contact the authors of this chapter in order to obtain the software. 6. Leakage of CSF from fixed brains must be carefully avoided, since it leads to air bubbles in the sample with associated susceptibility artefacts. Inadvertent bubbles can be removed by immersing samples in phosphate-buffered saline with the brainstem vertical and applying multiple cycles of vacuum pumping. 7. Scan parameters for the images shown in Fig. 17.1 were a T2-weighted, 3D fast spin echo sequence with eight echoes, TR/TE = 900/12 ms, two averages, field-of-view 24 × 24 × 40 mm3 and matrix size = 240 × 240 × 400 giving an image with 100 μm isotropic voxel and a total scan time of 2.8 h; and those shown in Fig. 17.2 were a T2-weighted, 3D fast spin echo sequence with six echoes, TR/TE = 325/32 ms, four averages, field-of-view 14 × 14 × 25 mm3 and matrix size = 432 × 432 × 780 giving an image with 32 μm isotropic voxel and total imaging time of 11.3 h. Long scans such as the latter should be implemented to be insensitive to drift in the magnet and DC offsets in the RF receivers. 8. Tensor cubic b-spline approximation is a robust means of estimating a non-linear distortion field based on landmarks. An implementation can be found at https://wiki.phenogenomics.ca. 9. The blurring kernel varies by the type of input data; in the common case of scans originating from the same scanner with care taken to ensure that the head is in a similar position then two blurs of FWHM 0.5 mm and 0.3 mm are sufficient. If the data are more disparate, which can occur with fixed sample scans and/or if the data originate from different scanners, then an initial centre-of-gravity estimation followed by a larger set of kernels (FWHM 5, 1, 0.5, 0.3 mm) is needed. 10. We found that using these pairwise registration results in a considerably improved representation of the data than

360

Lerch, Sled, and Henkelman

to continue aligning each scan towards the atlas. We are indebted to Andrew Janke and Paul Thompson for the matrix averaging code/mathematics. 11. As with most brain imaging analysis, the choice of blurring kernels is a bit of a black art. We tend to blur with 0.1, 0.2, 0.5 and 1.0 mm kernels and rely on the 0.5 mm for the bulk of our analyses. The larger kernel is useful for looking at very diffuse changes and the smaller ones for very focal differences. 12. We define displacement and Jacobian determinant such that the identity transformation corresponds to 0 displacement and a Jacobian determinant of 1. 13. Find out more about RMINC at http://launchpad.net/ rminc. Many other software packages are capable of carrying out the same analyses. 14. The main variation on the linear model employed is the mixed-effects linear model, which correctly models interrelationships in the data that occur by, for example, having multiple time points for some of the animals in the study. 15. We tend to use both FDR and, on occasion, more stringent Family Wise Error Rate (FWER) correction techniques (14) such as permutation testing (15) and Random Field Theory (16). In our experience with phenotyping, especially when dealing with novel mutations, the role of MRI is primarily to generate hypotheses which will often be followed up using other techniques, including immunohistochemistry, electrophysiology. The more lenient FDR is therefore well suited for this task, whereas FWER’s more stringent threshold risks leave too many false negatives.

Acknowledgments Many of the present and previous members of the Mouse Imaging Centre (MICe) in Toronto have contributed to the development of these techniques. Many developmental biologists and mouse geneticists from around the world have provided mutant mice for imaging and improved the analysis methods. Funding from the Canadian Foundation for Innovation, the Ontario Research and Development Challenge Fund, the US National Institutes of Health and the Canadian Institutes of Health Research is gratefully acknowledged.

MRI Phenotyping of Genetically Altered Mice

361

References 1. Waterston, R. H., Lindblad-Toh, K., Birney, E. et al. Initial sequencing and comparative analysis of the mouse genome. Nature 2002;420:520–562. 2. Henkelman, R., Dazai, J., Lifschitz, N. et al. High throughput microimaging of the mouse brain. Proc Int Soc Magn Reson Med 2006. 3. Nieman, B., Bishop, J., Dazai, J. et al. MR technology for biological studies in mice. NMR Biomed 2007;20:291–303. 4. Bock, N., Nieman, B., Bishop, J., Mark Henkelman, R. In vivo multiple-mouse MRI at 7 tesla. Magn Reson Med 2005;54: 1311–1316. 5. Spring, S., Lerch, J., Henkelman, R. Sexual dimorphism revealed in the structure of the mouse brain using three-dimensional magnetic resonance imaging. Neuroimage 2007;35:1424–1433. 6. Dazai, J., Bock, N. A., Nieman, B. J., Davidson, L. M., Henkelman, R. M., Chen, X. J. Multiple mouse biological loading and monitoring system for MRI. Magn Reson Med 2004;52:709–715. 7. Collins, D. L., Neelin, P., Peters, T. M., Evans, A. C. Automatic 3d intersubject registration of MR volumetric data in standardized talairach space. J Comput Assist Tomogr 1994;18:192–205. 8. Collins, D., Holmes, C., Peters, T. M., Evans, A. Automatic 3-D model-based neuroanatomical segmentation. Hum Brain Mapp 1995;3:190–208.

9. Kale, S. C., Lerch, J., Henkelman, R. M., Chen, X. J. Optimization of the SNRresolution tradeoff for registration of magnetic resonance images. Hum Brain Mapp 2008;29:1147–1158. 10. Sled, J. G., Zijdenbos, A. P., Evans, A. C. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 1998;17:87–97. 11. Chung, M. K., Worsley, K. J., Paus, T. et al. A unified statistical approach to deformation-based morphometry. Neuroimage 2001;14:595–606. 12. Dorr, A., Lerch, J., Spring, S., Kabani, N., Henkelman, R. High resolution threedimensional brain atlas using an average magnetic resonance image of 40 adult C57Bl/6j mice. Neuroimage 2008;42:60–69. 13. Genovese, C. R., Lazar, N. A., Nichols, T. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 2002;15:870–878. 14. Nichols, T., Hayasaka, S. Controlling the familywise error rate in functional neuroimaging: A comparative review. Stat Methods Med Res 2003;12:419–446. 15. Nichols, T. E., Holmes, A. P. Nonparametric permutation tests for functional neuroimaging: A primer with examples. Hum Brain Mapp 2002;15:1–25. 16. Worsley, K. J., Taylor, J. E., Tomaiuolo, F., Lerch, J. Unified univariate and multivariate random field theory. Neuroimage 2004;23(Suppl 1):S189–S195.

wwwwwww

Chapter 18 Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications Philip K. Liu and Christina H. Liu Abstract Gene action plays a role in neural cell migration, learning processes, stress response, drug addiction, cancer, mental health, psychiatric and neurological disorders, as well as neurodegenerative diseases. Studies also show that upregulation of certain gene activities in neurons may contribute to the development of Alzheimer’s disease and other progressive cognitive disorders many decades after the alteration itself occurs. Endogenous, environmental stress-related, or drug-induced chemical imbalances in the brain affect the homeostasis of gene activities in neurons in specific brain regions and contribute to the comorbidity of mental illness and substance dependence. On the other hand, altered gene activities are also a necessary part of repair processes after brain injury. Our general well-being is governed by the highly regulated gene activities in our brains. A better understanding of gene activities and their relationship to the progression of neurological disease can help the research and medical communities develop necessary measures for early intervention, as well as plan more appropriate interventions or new therapeutic approaches that can benefit a broad spectrum of patients who will be or have been affected by brain diseases. We developed a non-invasive imaging technique that allows real-time assessment of gene transcription profiles in live brains. This imaging method has the potential to provide first-hand information about the progression of neurological disorders by gene targeting and cell typing, and it could elucidate a surrogate marker for therapeutic efficacy for future planning of treatments for human diseases. We have established a workable and reproducible MRI technique in live rodent brains. Key words: Amphetamine, antisense, cell typing, cerebral ischemia, drug abuse, gene targeting, heart arrest, molecular imaging, MRI, nanoparticles, prognosis, siDNA.

1. Introduction Procedures to evaluate gene activity in the brain at the transcription level are not performed for clinical purposes, because current techniques to detect altered gene transcription rely on the use of M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_18, © Springer Science+Business Media, LLC 2011

363

364

Liu and Liu

biopsy samples. The difficulty in obtaining biopsies of brain tissue severely limits the utility of these methods to monitor gene activities in vivo. Commonly brain biopsies are only permitted in patients with late-stage brain tumor. Currently, the diagnosis of progressive cognitive disorders, such as Alzheimer’s disease, must rely on the assessment of clinically defined symptoms, anatomical hallmarks in the brain, and peripheral biomarkers obtained from blood, urine, or cerebral spinal fluid samples. These biomarkers are typically only evident after irreversible damage to the brain has already occurred. For example, the use of blood tests to detect key peripheral biomarkers expressed as a result of neuronal death is too delayed to be useful to plan a therapeutic intervention. Moreover, it cannot provide an indication as to where cell death occurs in the brain. However, imaging tools have shown great promise as alternative approaches to these more conventional methods. Magnetic resonance (MR) spectroscopy and several MR imaging (MRI) methods – tools that are routinely used for non-invasive detection of abnormal function and structure in patients suffering from neurological disorders, neurodegenerative diseases, and mental illness – are emerging as powerful tools for detection of gene action in brains. Ligand (isotope)-guided positron emission tomography imaging also has great utility for detecting changes in the distribution of brain receptors associated with chronic drug abuse and mental illness as well as neurodegenerative diseases. These imaging techniques have helped tremendously to advance neuroscience research by enabling direct or indirect measurement of gene actions, so as to decipher gene transcription events that occurred before symptoms emerge. However, there remains a gap between our scientific understanding of gene activity in in vitro experiments compared to what takes place in the living brain during the evolution of diseases or disorders. Our technique uses a gene targeting MR contrast agent for the detection of intracellular mRNA in live brains using MRI. This technique combines two well-established research platforms, namely mRNA targeting by oligonucleic acids and iron-based contrast-enhanced MRI for in vivo applications. Specifically, this design creates a targeting MR contrast conjugate consisting of a modular short nucleic acid probe and a MR contrast agent that is covalently connected via a biotin–avidin linkage (Fig. 18.1). To afford contrast-enhanced MRI, we use dextran-coated superparamagnetic iron oxide nanoparticles (SPION, a T2 susceptibility agent). This modular probe construct can be modified from antisense DNA (ODN) to various RNA platforms (siRNA, microRNA, ribozyme, peptide DNA, etc.) for use with different imaging modalities (fluorescence probes, isotopes, and MR contrast agents) and for different targets by changing the sequences

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

365

Fig. 18.1. Molecular basis of MRI for gene targeting: nucleic acids (sODN) with a charged backbone are taken up via membrane-bound receptors, the sequence in the sODN determines its retention in the cells. The same mechanism is proposed for SPION-sODN.

of the nucleic acids (1).1 Our design using nucleic acid and an iron oxide-based MR visible probe was independently reproduced using short interfering RNA (2). This approach improves mRNA imaging, showing measurable improvement for a range of imaging applications, from biopsy or postmortem samples (as shown in Fig. 18.2) to living organs (as shown in Fig. 18.3). Still, many hurdles of working with nucleic acids as brain probes in vivo remain, and we have focused on improving our methods through rigorous research. We expect to see modifications of our methods as investigation in this area progress. The nucleic acid-based MR probe used in MRI contains a sequence that can bind to a specific target mRNA, forming a hybrid that lasts long enough to create a transient window for imaging. In addition, the dose should not block translation if gene knockdown is not the objective; such a gene knockdown is often seen in methods that use antisense DNA or short interfering RNA. Because gene transcript targeting and reporting are based on specific binding of the reporter probe to its target, the probe must have the highest reporting sensitivity when its loading capacity is one; that is, maximum reporting sensitivity is achieved with one targeting ODN to one contrast agent. Four targeting ODNs per contrast agent (loading capacity of 4) will reduce reporting sensitivity by 75%. This concept of probe design for gene targeting MRI is somewhat

366

Liu and Liu

Fig. 18.2. In vivo hybridization of FITC-sODN-c-fos in the neuronal formation of the dentate gyrus (arrow) demonstrates DNA transfection in vivo via ICV delivery (by passing BBB), target hybridization, clearance of nonspecific sODN-mRNA hybrids in live brain and detection in postmortem samples.

opposite to those used in current molecular imaging which may call for a maximum loading capacity. Since the first description of our methods, a few modifications for ODN-based probes in SPION MR contrast conjugate have been described (3). Additional considerations and modifications are being made to achieve maximal MR signal changes to improve signal to noise and to delineate mechanism of this novel technique. One suggestion that has been made, and which we have carried out, is to validate our methods in neurons grown in culture. Transfection of ODN probes to cultured primary neurons in the resting state, unlike primary astroglia, has been very difficult as the primary neurons may not express adequate mRNA transcripts after adaptation to the culture condition. Neuronal transfection by fluorescent-labeled short DNA in vivo has been detected in postmortem samples (4, 5). For this reason, we have always validated sODN specificity by transfection in vivo (6). Below, we present our current methods for making our gene

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

367

Fig. 18.3. DNA transfection in vivo: FITC-sODN or SPION-sODN is delivered to cerebrospinal fluid, the probe is distributed through the Virchow–Robin space, and uptake and retention for MRI in live brains, or postmortem histology (Copyright: J Neurosci, 2009, Reproduced with permission).

targeting probes with examples of conjugations between sODN and SPION to achieve a sufficient contrast-to-noise ratio for gene targeted MRI.

2. Materials All magnetic resonance contrast agents are characterized (e.g., particle sizes and magnetic relaxivities) before use. All enzymes were tested before use for optimal concentration, time of incubation, and temperature according to vendor’s specification. 1. Superparamagnetic iron oxide nanoparticles (SPION) with a core diameter of 10 nanometers (nm) and a hydrodiameter of 18–30 nm or less were homemade, but the same material can be purchased from Biophysics Assay Laboratory, Inc. [Molday ION (No. CL-30Q02-2), colloidal size of 30 nm, Worchester, MA].2 2. Sodium citrate, sodium hydroxide, BupHTM Phosphate Buffered Saline (PBS), AmionLink Reductant sodium cyanoborohydride, and NeutrAvidinTM Biotin-binding

368

Liu and Liu

protein were purchased from Thermo Scientific (St Louis, MO). Prepared BupH-PBS in 500 ml distilled water (0.1 M, pH 7.2). 3. 2-Chloroethylamine hydrochloride (98%) ACROS ORGANICS (Geel, Belgium).

was

from

4. Centrifugal filter devices (Amico Ultra-4, Ulturacel-100 K) were from Millipore (Billerica, MA). 5. Biotinylated single-stranded phosphorothioate-modified oligodeoxynucleotides (sODN) labeled with fluorescein isothiocyanate (FITC), Cy3, or rhodamine are homemade.3 We have ordered the same material from commercial sources such as ThermoScientific, Invitrogen, Sigma, or Amitof (Cambridge, MA). 6. Antisense DNA design: One antisense sODN was used as example – sODN-fosB (5 -CCTTAG complementary to CGGATGTTGACCCTGG-3 , sequence no. 1,925–1,946 of FosB mRNA of the mouse [mmFosB, accession no. X14897]) can be obtained from the mouse genome library. The sequences of sODNc-fos, sODN-actin, and sODN-Ran have been reported previously (7, 8). Single-stranded ODNs were synthesized with protection from the use of non-specific nucleases by phosphorothioate modification of all nucleotide bridges, and the resulting sODNs were purified by polyacrylamide gel electrophoresis. 7. Sense DNA (UPS, 5 -GATCGCCGAGCTGCAAAAAG3 ), 146 nucleotides upstream of sODN-fosB, from the same mouse FosB mRNA, was used to determine the specificity of antisense sODN-fosB in the polymerase chain reaction (9). 8. Mouse brain cDNA library is homemade, but can be purchased from Stratagene (La Jolla, CA). Polymerase chain reaction buffer and Taq polymerase were also purchased from Stratagene. 9. Antibodies were from Abcam (Cambridge, MA). 10. All solutions are sterilized by filtration using Nalgene filter units (Nalge Nunc International Corp., Rochester, NY). 11. Inhalation anesthesia for in vivo MRI of rodents: halothane (2-bromo-2-chloro-1,1,1-trifluoroethane, Sigma-Aldrich, Inc.) or forane (isoflurane, USP, Baxter Healthcare Corp., Deerfield, IL). 12. Toothpaste (preferably alcohol free to minimize repeat irritation in the case of chronic survival studies) to fill rodents’ ear canals during MRI to reduce artifacts that occurs at the air–tissue interface.

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

369

3. Methods 3.1. Preparation of the SPION-NA (Functionalization)

1. Attaching functional groups to SPION: Incubate 20 g of Molday ION in a 15 ml of 2 N sodium hydroxide with 2 M 2-chloroethylamine with gentle stirring overnight at room temperature. The reaction generates hydrochloride and should be performed in a well-ventilated hood. 2. Dialysis filter the solution in 10× the volume of 25 mM filter-sterilized sodium citrate (pH 8.0), then continue dialysis filtering in additional 10× the volume of 0.1 M BupHPBS using an Ulturacel-30 K filter (30 KD cutoff). 3. Incubate NeutrAvidin (NA, 1–20 mg) overnight in the presence of sodium cyanoborohydride (1 M) in BupH-PBS (0.1 M, pH 7.2) in an amber bottle (avoid light) at a total volume of 10 ml at room temperature.1 The resulting covalently linked product, SPION-NA, is filtered and dialyzed against a 20X volume of sodium citrate buffer solution (25 mM, pH 8.0). 4. Concentrate the volume to 5–6 ml or less. 5. Iron concentration is measured in hydrogen peroxide and the optical density is measured at 410 nm. 6. Store the activated SPION (SPION-NA) at 4◦ C in an amber-colored bottle with rubber stopper to minimize oxygen contact, at a concentration of 3–4 mg iron per milliliter sodium citrate buffer (25 mM, pH 8.0).

3.2. Testing of sODN-fosB Targeting Specificity in Total Cerebral cDNA

1. Mix 10 μg of total cDNA, FITC-sODN-fosB, upstream sense DNA (10 pmol each), and polymerase reaction buffer in 40 μl volume. 2. Prepared polymerase and four deoxynucleotide triphosphates (dNTP, 20 mM) in 10 μl. 3. The reaction mix is incubated at 95◦ C for 30 s; the mix is then maintained at 70◦ C when a bolus of 10 μl of polymerase and four dNTPs is added. 4. Polymerase chain reaction is carried out for 25 cycles at 90◦ C (30 s), 39◦ C (30 s), and 65◦ C (45 s) followed by 68◦ C for 2 min. 5. The resulting product is resolved by electrophoresis on agarose gel (1%) for 100 V-h (100 V in 1 h or 50 V in 2 h). 6. A good antisense sODN should yield only one product of 146 base pairs according to the mouse fosB mRNA sequence.

370

Liu and Liu

3.3. Saturation Binding Assay

1. We mix 10 μl of SPION-NA (~30 μg, duplicate tubes) to increasing concentration of FITC-sODN-fosB-biotin, from 0, 1, 2, 9, 15, or 30 pmol of FITC-sODN-biotin per microgram of SPION (lanes 1–6, lane 7 free FITC-sODN-fosB, Fig. 18.4a), incubate 30 min at room temperature (or overnight at 4◦ C). The coupling reaction is resolved by electrophoresis on agarose gel (1%) at room temperature for 75 V-h. 2. A successful coupling reaction is shown by a fragment upshift of FITC signal: Lane 7 shows unbound sODN (arrow) that traveled 4 cm from the loading well, SPION-sODNFITC hardly traveled (or only a short distance from the loading well), as shown in lanes 3–6. Lane 2 with a duplicate sample shows saturation binding at 30 pmol of FITCsODN-biotin per microgram of SPION, especially when

Fig. 18.4. Binding capacity of SPION-NA by FITC-sODN-biotin. Upon binding between NeutrAvidin (NA) on SPION and biotinylated-sODN-FITC, there is an upshift from low molecular weight sODN to high molecular sODN (Panel a, 50 V-h). The mobility of sODN, as little as 1 pmol of sODN per microgram of SPION-NA (Fe) and increasing with binding at 30 pmol per microgram Fe (lanes 2 and 3), is capable of pulling bound SPION (dark spot below the well) to the opposite direction of SPION-NA (lane 1, Panel a). This pulling mobility is used to detect the binding capacity of SPION-NA by increasing the amount of FITC-sODN-biotin (Panel b, 110 V-h). The presence of unbound FITC-sODN (arrows) indicates the saturation of binding. FITC-sODN is pulling the SPION along during electrophoresis.

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

371

free (unbound) FITC-ODN appears as excess FITC-sODNfosB-biotin. Lane 3 shows near saturation because there is no unbound FITC-sODN-fosB-biotin. 3. We observed the pulling of sODN on SPION with high sODN binding (lane 2). 4. To narrow the saturation binding, we incubated FITCsODN-gfap-biotin and SPION-NA at a ratio of 28, 26, 24, 22, 20, 18, 16 pmol per microgram SPION. Figure 18.4b shows all mixtures contain increasing unbound FITC-sODN-fosB-biotin, from 16 to 28 pmol/μg SPION. Indeed, free FITC-sODN-gfap appeared at high ratio (28 pmol/μg), at the highest intensity, and reduces the intensity as binding capacity is lowered to 16 pmol FITC-sODN-fosB-biotin per microgram iron oxide.4 5. Each NA has four binding sites for biotin; we calculated no more than 4 pmol of NA per microgram of iron oxide in this batch of SPION-NA. 3.4. Conjugation of Biotinylated sODN to SPION-NA

1. SPION-NA and biotinylated sODN are mixed and incubated at 4◦ C overnight (Section 3.4 describes how the ratio is determined). Generally, we add 3–4 pmol per microgram SPION-NA (16 pmol divided by 4, as there are four biotin binding sites per molecule of avidin).5 2. The optimal MR contrast is achieved when one molecule of SPION reports one copy of mRNA. This can be achieved by reducing the sODN-biotin to a constant amount of SPIONNA and testing the signal reduction frequency (R2 ∗ value) in vivo (Section 2.4.8). 3. Take four microvials (0.5 ml capacity) and add 30 μg of SPION-NA to each vial, then add 120, 60, 30, and 15 pmol of biotin-sODN-gfap to each vial to achieve four different sODN to SPION conjugation ratios (pmol sODN per microgram SPION). 4. Mix well by tapping and then briefly centrifuge in a benchtop microfuge (Fisher Scientific). 5. Add sodium citrate (25 mM, pH 7.8) to a final volume of 60 μl, so that the iron oxide is 0.5 μg Fe/μl. 6. Incubate overnight at 4◦ C. 7. Deliver SPION-fosB (1.6–2 μl, ICV) to mice (40 μg/kg for 20–25 g of mice); generally, two mice per dose or n=8. 8. Acquire MRI R2 ∗ maps at 9.4 T 4 h after delivery (according to Section 2.9) and select the best conjugate ratio (described in Section 2.4.3) which results in maximal R2 ∗ increase in most brain regions (Section 2.4.3) from each batch of SPION-NA.

372

Liu and Liu

3.5. Stability of FITC-sODN in Serum and in Cerebrospinal Fluid

1. Animals are anesthetized using either isoflurane or halothane (for drug stimulation studies). 2. Five milliliters of arterial blood is withdrawn from the carotid artery with a 26 g needle on a 10-ml syringe; blood plasma is collected after centrifugation at 500×g for 10 min. 3. We incubated SPION-sODN-FITC (200 pmol) in 0.1 ml of saline (9 or 45 μg Fe/μl), plasma (9 μg Fe/μl), or cerebrospinal fluid (45 μg Fe/μl) for 24 h at 37◦ C, and the resultants of reaction were resolved by electrophoresis agarose gel (0.8%) (120 V-h). We found that SPION-sODN remained linked in both fluids for the duration of the study (Fig. 18.5).

Fig. 18.5. Stability of SPION-sODN in body fluid: SPION-sODN or sODN was incubated in body fluids [serum or cerebrospinal fluid (CSF)] at 37◦ C for various hours and then the degradation of sODN was determined in agarose gel (1%) by electrophoresis. The sODN in SPION-sODN is protected from binding to serum proteins (compare lanes 3–9 to lanes 11 and 1) and is stable for 24 h at physiological temperature without significant degradation as seen in lanes 2 and 10. We observed the same result of SPION-sODN in the CSF for the same duration (not shown).

3.6. Delivery of SPION-sODN: Intracerebroventricular Delivery (ICV)

1. Animals are anesthetized with either isoflurane or halothane (for drug stimulation studies) and placed in a stereotactic device (Stoelting, Wood Dale, IL) for ventricle infusion, via Hamilton syringe, to bypass the blood–brain barrier. 2. The mouse is placed in a supine position; the skin is cut open on the midline of the head and the skull is exposed. The needle is placed on bregma and the needle tip is adjusted to stereotactic coordinates from bregma: LR 1.0 mm, AP –0.2 mm.

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

373

3. The coordinate is marked by a surgical marker and the skull on the mark is drilled to make the skull thinner so the needle can puncture through the skull. 4. The needle is lowered to the surface of the thin skull (zero point), and the needle is lowered by 3.0 mm (DV–3.0 mm). 5. We deliver 2 μl of SPION-sODN to the lateral third ventricle of the mouse at the rate of 0.5 μl/min. The coordinates can change from strain to strain. For example, those for C57black6 mice may differ by 0.05–0.1 mm on the AP plane between 18 and 25 g. Furthermore, C57black6 mice from Jackson Laboratory may show some differences from mice of the same strain from Taconic Farm, NY. 6. The burr hole on the skull created by the infusion procedure is sealed with bone wax before the incision is closed with suture. The infusion dose of SPION-sODN we used in mice was 0.04 mg Fe/kg body weight in 2 μl sodium citrate solution (25 mM). 3.7. Delivery of SPION-sODN: Intraperitoneal Delivery (i.p.)

1. We conjugate SPION-NA and FITC-sODN-biotin at the optimal mixture (0.5 μg Fe/μl at 4◦ C overnight). For one mouse, we generally deliver 0.2 ml (0.5 μg Fe/μl) to a 25 g mouse or 4 mg/kg. 2. Immediately before delivery, we add 2 μl lipofectamine2000 (1 mg/ml, Invitrogen Lifesciences) per 200 μl of SPIONsODN solution. Mix well by tapping and then briefly centrifuge in a bench-top centrifuge. We acquire baseline MRI before delivery of the probe (pre-delivery), then one or more images after delivery in a 9.4 T magnet at various time points (generally at 2, 4, 6, and 24 h after i.p).

3.8. Delivery of SPION-sODN: Ophthalmic Route of Delivery (OTRD)

We prepared SPION-sODN (3 μg Fe/μl) and lipofectamine as in step 2.7. After general anesthesia [ketamine (100 mg/kg) and xylazine (10 mg/kg) anesthesia i.p.] we delivered SPION-sODNFITC via micropipette and pipette tips as eye drops to the eye sac at a rate of 10 μl every 15 min (33.3 μl, OTRD, 3.996 mg Fe/kg).

3.9. MRI Acquisition (Live Brain Imaging)

The following steps refer to MRI acquisition with a 9.4 T Bruker/Magnex horizontal bore (21 cm) animal MR scanner. A custom-made cradle equipped with a tooth bar and nose cone complete with gas input and exhaust lines is used for scanning. The MR parameters described below are suitable for a surface coil with transmit and receive capability. 1. Animals are anesthetized during the entire MRI session to minimize motion. Inhalation anesthesia such as halothane or isoflurane is used for most survival MRI studies, as

374

Liu and Liu

it has maintenance advantages over injectable anesthetics. 2% isoflurane or halothane in pure O2 is used to ensure proper oxygenation of the brain throughout the entire MRI session. Whilst isoflurane is a more acceptable anesthetic regimen for small animal MRI, halothane is recommended for drug stimulation studies, because isoflurane has been shown to block or reduce the cerebral hemodynamic changes associated with pharmacological stimuli in rodents (depicted in functional MRI studies). Interestingly, we also observed in animals under isoflurane anesthesia reduced SPION-sODN signal changes associated with amphetamine stimulation. 2. Animals are placed in a prone position and secured with the tooth bar. To reduce air–tissue interface artifact, toothpaste was injected into the ear canals, using a flexible catheter, without rupturing the ear drums. It is important that ear drums are not punctured to prevent the leakage of toothpaste into the trachea, which can result in suffocation. A surface coil is then placed on top of the animal’s head and the whole setup is inserted into the MRI scanner. 3. A standard tune and match procedure is performed followed by automatic or manual shimming before each MRI scan series to ensure maximal homogeneity in the field strength inside the gradient for signal sensitivity and reproducibility. 4. The MRI protocols include as follows: (1) A general localization sequence (for example, RARE Tripilot) (2) T2 anatomical imaging for slice positions, TR/TE=7,000/25 ms, 117×117 μm2 , 20 0.5 mm contiguous slices, RARE factor 8, number of average (NA)=2 (2) Serial, 2D gradient echo fast imaging (GEFI) with TR/TE=500/3, 4, 6, 8, 10 ms, flip angle=30, NA=2 with the same geometry as in (2). The entire scan series lasts for less than 30 min. 5. For voxel-wise and region-of-interest comparison, images should first undergo automatic and manual alignment procedures using any standard image processing software with registration capability. Fine-tuning of alignment can be performed by visual comparison to template images, focusing on obvious landmarks such as the corpus callosum and outlines of the ventricles. R2 ∗ maps are constructed from the aligned images (with incremental TEs). R2 ∗ (inverse of T2 ∗ ) maps are calculated using pixel-wise linear fitting from the set of images with the same TR and incremental TEs based on equation M=M0 × exp(–TE/T2 ∗ ). Elevated R2 ∗ (or reduced T2 ∗ ) is theoretically caused by the presence of SPION in tissue. All ROIs are outlined according to ‘The Mouse Brain in Stereotaxic Coordinates’ (10). Averaged R2 ∗ values within ROIs are extracted from each animal and

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

375

we calculate the group mean and standard error of the mean (SEM) in each group for statistical analysis. 3.10. Histological Localization of sODN Under Fluorescent Microscope

1. Postmortem brain preparation: animals are anesthetized and brains are quickly removed from the skull and flash frozen in liquid nitrogen. Thin coronal tissue sections can be prepared using Cryostat and mounted on glass slides; thick tissue sections can be prepared using a Vibratom system and stored as floating sections in 0.1% sodium azide/PBS solution. 2. Frozen sections were incubated in 3.7% freshly prepared paraformaldehyde for 10 min and rinsed in fresh doubledistilled water. Fluorescent signals can be directly observed under a fluorescent microscope.6

3.11. Histological Localization of SPION Under Light Microscope5

1. Postmortem sample preparation: animals are anesthetized and transcardially perfused with 15 ml heparinized saline at a rate of 10 ml/min and then with 10 ml of freshly prepared 4% paraformaldehyde (PFA) in 0.1 M PB, pH 7.4. The brains are stored overnight in PFA solution at 4◦ C and stored at 4◦ C in 20% sucrose/PBS solution to chase out PFA. Samples are ready when the brains drop to the bottom of the sample container, usually within 24 h. Brain samples collected using this method can also be used for MRI microscopy (see Section 2.12 for detailed steps). Samples are embedded in paraffin and thick coronal tissue slices (20–100 μm thickness) are prepared and mounted on glass slides. 2. After removing paraffin from the glass slides, tissue samples can be stained for Prussian blue (PB) with 2% potassium ferrocyanide in 2% HCl (Perl’s method) and counterstained with nuclear fast red (NFR) (Fisher Scientific, Houston, TX). Iron stains can be viewed under a light microscope.

3.12. MRI Acquisition (Postmortem Brain Imaging)

The following steps refer to a 14.1 T Bruker/Magnex vertical bore (8.9 cm) MR spectrometer. A volume coil with inner diameter of 1 cm was used. 1. The entire brain can be immersed in a 1 cm diameter NMR tube in perfluoro compound FC-40 (Fluoroinert FC-40, Sigma) to eliminate background proton signal. Care should be taken to minimize air bubbles as the brain is inserted into the tube. 2. A standard tune and match procedure is performed followed by automatic or manual shimming before each MRI scan series to ensure maximal homogeneity in the field strength inside the gradient, for signal sensitivity and reproducibility.

376

Liu and Liu

3. The MRI protocols include as follows: (1) A general localization sequence (for example, RARE Tripilot) (2), a threedimensional (3D) fast low-angle shot (FLASH) gradient sequence is to acquire T2 ∗ –weighted images (TR/TE = 50/18 ms, 40 × 40 × 40 μm3 , flip angle=20, NA=24). The entire scan lasts for approximately 11 h.

4. Notes 1. Notes to nomenclature of MR probes: We use the terms oligodeoxynucleotide (ODN) and oligoribonucleotide (ORN) for synthetic nucleic acids with antisense sequence to mRNA targets because they may function differently from naturally occurring dicer-dependent and dicereindipendent micro RNA that exists endogenously from de novo synthesis. For consistency with practice of using capital letters in the names of proteins, we use lower case characters for our mRNA-targeting sODN linked to MR-visible agents (SPION-cfos, SPION-fosB, sODN-fos or sODN-fosB), and reserve upper case for future antigen-targeting probes (SPION-ACTIN or sODN-ACTIN). We use SPION-Ractin or SPION-Rgfap for sORN-labeled SPION. We use abbreviations for probes with no target (SPION-NA), or for DNA with random sequence (SPION-Ran). Phosphorothioatemodified nucleic acid is abbreviated as sODN or sORN, and peptide-modified nucleic acid is referred to as peptide ODN or peptide ORN. 2. Generally, we ask BIOPAL to send us a small 0.1 ml sample of freshly made Molday ION (CL30Q02-2, 30 nm in diameter) to validate the core size and the ability to be functionalized to accept NeutrAvidin, a key step in probe construction. We generally start with 20 mg of SPION made within 3 weeks. The ratio of NA to SPION is essential as more NA on one SPION will reduce the sensitivity of SPION-NA. We have been working with 1–20 mg ratio. SPION-NA has a shelf-life of 3 months at 4◦ C (1). 3. For histological localization of intracellular sODN, we label the sODNs with FITC, Cy3, or rhodamine. 4. SPION-NA (dark brown band) often migrates to the opposite direction of FITC-sODN in agarose gel of 1%. We observed the pulling of sODN on SPION because the dark brown band (SPION) in SPION-sODN with high sODN binding (lane 2) will migrate slower than the SPION in lane 1 (no sODN). This is evidence that the charge on

Gene Targeting MRI: Nucleic Acid-Based Imaging and Applications

377

FITC-sODN, that is different from peptide DNA, affects the migration and transfection of SPION (Fig. 18.4a). 5. Saturation binding on SPION-NA will reduce probe sensitivity, as more mRNA transcripts will bind to antisense sODN per SPION. We found 1 mg or less NA with 20 mg SPION ratio to be most suitable for our work (1). 6. Permeabilization of tissue samples for immunohistochemistry should be minimized to retain sODN probes. If perfusion is needed, try the two-buffer system (4).

Acknowledgments We thank Drs. Charng-ming Liu and Jia Q. Ren for technical assistance and Ms. N. Eusemann for excellent editing. This project was supported by grants from NIH (DA024235 and DA026108 to CHL, NS057556 and DA29889 to PKL), American Heart Association (09GRNT2060416 to CHL), and funds from the Stanley Medical Research Institute through the Stanley Center for Psychiatric Research at the Broad Institute. Athinoula A Martinos Center for Biomedical Imaging is partially supported by RR14075. References 1. Liu, C. H., Kim, Y. R., Ren, J. Q., Eichler, F., Rosen, B. R., Liu, P. K. Imaging cerebral gene transcripts in live animals. J Neurosci 2007;27:713–722. 2. Medarova, Z., Pham, W., Farrar, C., Petkova, V., Moore, A. In vivo imaging of siRNA delivery and silencing in tumors. Nat Med 2007;13:372–377. 3. Liu, P. K., Mandeville, J. B., Guangping, D., Jenkins, B. G., Kim, Y. R., Liu, C. H. Transcription MRI: A new view of the living brain. Neuroscientist 2008;14:503–520. 4. Cui, J. K., Hsu, C. Y., Liu, P. K. Suppression of postischemic hippocampal nerve growth factor expression by a c-fos antisense oligodeoxynucleotide. J Neurosci 1999;19:1335–1344. 5. Hecker, J. G., Hall, L. L., Irion, V. R. Nonviral gene delivery to the lateral ventricles in rat brain: Initial evidence for widespread distribution and expression in the central nervous system. Mol Ther 2001;3:375–384. 6. Liu, C. H., Ren, J. Q., Yang, J., Liu, C. M., Mandeville, J. B., Rosen, B. R., Bhide, P. G.,

7.

8.

9.

10.

Yanagawa, Y., Liu, P. K. DNA-based MRI probes for specific detection of chronic exposure to amphetamine in living brains. J Neurosci 2009b;29:10663–10670. Liu, C. H., Huang, S., Cui, J., Kim, Y. R., Farrar, C. T., Moskowitz, M. A., Rosen, B. R., Liu, P. K. MR contrast probes that trace gene transcripts for cerebral ischemia in live animals. FASEB J 2007c;21: 3004–3015. Liu, C. H., Huang, S., Kim, Y. R., Rosen, B. R., Liu, P. K. Forebrain ischemiareperfusion simulating cardiac arrest in mice induces edema and DNA fragmentation in the brain. Mol Imaging 2007b;6:156–170. Liu, C. H., Huang, S., Cui, J., Kim, Y. R., Farrar, C. T., Moskowitz, M. A., Rosen, B. R., Liu, P. K. MR contrast probes that trace gene transcripts for cerebral ischemia in live animals. FASEB J 2007;21: 3004–3015. Paxinos, G., Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates. London: Academic Press Limited; 2001.

Chapter 19 Molecular MRI Approaches to the Detection of CNS Inflammation Nicola R. Sibson, Daniel C. Anthony, Sander van Kasteren, Alex Dickens, Francisco Perez-Balderas, Martina A. McAteer, Robin P. Choudhury, and Benjamin G. Davis Abstract Inflammation is a key component of many neurological diseases, yet our understanding of the contribution of these processes to tissue damage remains poor. For many such diseases, magnetic resonance imaging (MRI) has become the method of choice for clinical diagnosis. However, many of the MRI parameters that enable disease detection, such as passive contrast enhancement across a compromised blood–brain barrier, are weighted towards late-stage disease. Moreover, whilst these methods may report on disease severity, they are not able to provide information on either disease activity or the underlying molecular processes. There is a need, therefore, to develop methods that enable earlier disease detection, potentially long before clinical symptoms become apparent, together with identification of specific molecular processes that may guide specific therapy. This chapter describes the methodology for the synthesis and validation of two novel, functional MRI-detectable probes, based on microparticles of iron oxide (MPIO), which target endothelial adhesion molecules. These contrast agents enable the detection of acute brain inflammation in vivo, at a time when pathology is undetectable by conventional MRI. Such molecular MRI methods are opening new vistas for the acute diagnosis of CNS disease, together with the possibility for individually tailored therapy and earlier, more sensitive assessment of the efficacy of novel therapies. Key words: Microparticles of iron oxide, MPIO, MRI, inflammation, brain, molecular imaging, vascular cell adhesion molecule-1, VCAM-1, sialyl-LewisX , sLeX .

1. Introduction The inflammatory response is essential for survival as part of the host defense system to injury and infection. However, inflammation can be inappropriate or excessive, and as such has M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_19, © Springer Science+Business Media, LLC 2011

379

380

Sibson et al.

the capacity to exacerbate tissue injury. This so-called bystander damage is particularly an issue in the brain where the capacity for tissue regeneration is minimal. It has become increasingly clear that inflammation contributes not only to the archetypal inflammatory disease of the brain, multiple sclerosis, but also to a spectrum of acute and chronic neurological diseases, including stroke, Alzheimer’s disease, prion disease and HIV-related dementia. Yet relatively little is known of the effects of inflammatory processes within the CNS. For many of these diseases, MRI is now routinely used for clinical diagnosis and assessment, but the contribution of inflammatory processes to those MRI data is not clear. Inflammation can induce a range of different processes all of which are potentially identifiable using MRI, including alterations in cerebral perfusion, permeability of the blood–brain barrier, changes in tissue water diffusion related to energetic compromise, oedema formation and demyelination. However, a lack of specificity in many of these parameters, together with weighting towards late-stage disease, has driven recent interest in molecular imaging techniques that can accurately identify markers of early inflammation, and hence disease, in the brain. To this end, we have developed a number of targeted MRI-detectable contrast agents that enable the detection of specific molecular epitopes; of particular interest as targets are the endothelial adhesion molecules (e.g. vascular cell adhesion molecule-1 (VCAM-1) (1), E-selectin (2–4)) which are upregulated early in the inflammatory cascade and are expressed on the luminal surface of the endothelium. Consequently, these molecules provide an accessible tag for the detection of acute inflammatory events within the brain, but without the requirement for translocation across the blood–brain barrier (BBB) or accumulation of the agent within the brain itself. We have recently used similar MPIO-based constructs to image adhesion molecules in atherosclerosis (5) and activated platelets in mouse models of cerebral malaria (6) and atherothrombosis (7, 8). The contrast element of these agents is a core of microparticles of iron oxide (MPIO). These are superparamagnetic particles consisting of a magnetite (Fe2 O3 ) and/or maghemite (Fe3 O4 ) core surrounded by a polymer coat. MPIO possess several characteristics that are potentially useful for imaging endovascular molecular targets: (i) a high iron content that is orders of magnitude greater than that contained in ultrasmall particles of iron oxide (USPIO) commonly used for MRI contrast; (ii) potent negative contrast effects on T2 ∗ -weighted images that extend to a distance roughly 50 times the physical diameter of the MPIO owing to their high iron content; (iii) reduced susceptibility to nonspecific uptake by endothelial cells owing to their size, compared to USPIO, and therefore specificity for endovascular molecular targets (9); (iv) rapid clearance from the circulation again owing to their size (USPIO have relatively long blood half-lives of ∼6 h),

Molecular MRI Approaches to the Detection of CNS Inflammation

381

resulting in low background contrast effects and enabling rapid imaging post-administration; and (v) synthetic and commercial availability with a range of reactive surface groups, providing the opportunity for covalent conjugation of protein, antibodies, small peptides or carbohydrate ligands. Although the precise details are beyond the scope of this chapter, MPIO can be readily synthesised using a variety of procedures (4, 10) based on the in situ formation of magnetite in the presence of dextran. Such methods use solutions of FeCl3 , FeCl2 and dextran T10 in water, often under controllable inert atmospheres (e.g. argon). Variation of pH through addition of NH4 OH solution, coupled with controlled mechanical stirring and warming, allows access to a variety of particle compositions. Key factors include ratio of dextran to iron (II) and iron (III), stirring speed and composition of dextran employed. The dextran layer can subsequently be crosslinked by reaction with epichloridrin followed by ammonium hydroxide to increase the stability of the microparticles and concomitantly functionalise them with amino groups. We have demonstrated the use of such contrast agents, targeted with either antibodies or carbohydrate moieties, for the detection of inflammation in the CNS. We have further shown that these reveal endothelial upregulation of their targets at times when other techniques show no evidence of abnormality (1, 2, 4). Surface-functionalised MPIO (1 μm diameter) are used for conjugation to either mouse monoclonal antibodies (e.g. VCAM1) or carbohydrate oligosaccharides (e.g. sialyl-LewisX ). Activated mouse endothelial cells, stimulated with tumour necrosis factor-α (TNF-α), are used to test the capacity of the targeted MPIO constructs for specific and quantitative binding in vitro. For proof-ofprinciple in vivo studies, acute endothelial activation is induced by stereotaxic injection of interleukin-1β (IL-1β) into the left striatum of the mouse or rat. To date, the VCAM–MPIO has been used in mice only, given that it is a mouse monoclonal antibody. However, we have recently demonstrated the efficacy of the antibody for rat VCAM-1 and believe, therefore, that this agent should also work in rats. In contrast, the sLeX –MPIO agent has been tested in rat models only to date. However, the advantage of the carbohydrate-targeting ligand is the lack of species specificity, such that the same molecule is likely to be efficacious across a range of species from mouse to man. Targeted MPIO are administered intravenously, and in vivo MRI of the brain is performed at 7 T using a T2 ∗ -weighted 3D gradient echo sequence, with a final isotropic resolution of 90–120 μm. Focal hypointense contrast effects that delineate the architecture of activated cerebral blood vessels, with minimal background contrast, are induced where the MPIO bind to their target molecules. This protocol is readily adaptable to other endothelial-specific targets through modification of the targeting ligand.

382

Sibson et al.

2. Materials 2.1. Conjugation of VCAM-1 Antibody to MyOneTM Tosylactivated MPIO

1. MyOneTM ‘Tosylactivated’ superparamagnetic polystyrene R (1.08 μm diameter) (Invitrogen, Paisley, UK). Dynabeads Store at 4◦ C (see Note 1). R -S magnetic particle concentrator (magnet) 2. Dynal MPC (Invitrogen).

3. Pre-washing and coating buffer: 0.1 M sodium borate, pH 9.5. Store at 4◦ C (see Note 2). 4. 3 M ammonium sulphate. 5. Antibody ligand (see Notes 3 and 4): purified monoclonal rat anti-mouse CD106/VCAM-1 antibody (clone M/K2) (Cambridge Bioscience, Cambridge, UK). Store at 4◦ C (see Note 5). Purified isotype negative control IgG-1 antibody (clone Lo-DNP-1) (Serotec, Oxford, UK). Store at 4◦ C. 6. Blocking buffer: phosphate-buffered saline (PBS), pH 7.4, 0.5% bovine serum albumin (BSA), 0.05% Tween-20. Store at 4◦ C. 7. Washing and storage buffer: PBS, pH 7.4, 0.1% BSA, 0.05% Tween-20. Store at 4◦ C. 2.2. Synthesis of Sialyl-LewisX Conjugated MPIO

1. 1-Thio-S-cyanomethyl-modified carbohydrate, e.g. 1-thioS-cyanomethyl-sialyl-LewisX 2. Methanol – HPLC grade. 3. Sodium methoxide solution in methanol (25% wt in methanol, Sigma-Aldrich 156256). 4. Amine-terminated superparamagnetic iron oxide particles (1 μm diameter) (see Note 6). 5. Magnet, such as Dynal DynaMag-2 magnet or Miltenyi OctoMACS. 6. Phosphate-buffered saline (PBS).

2.3. In Vitro VCAM–MPIO Binding to TNF-α -Stimulated sEND-1 Cells

1. Dulbecco’s Modified Eagle’s Medium (DMEM) is supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 U penicillin and 0.1 mg/ml streptomycin. Store at 4◦ C. 2. Trypsin/EDTA solution. Store at 4◦ C. 3. Round microscope coverslips (19 mm). 4. Murine recombinant tumour necrosis factor-alpha (TNF-α, R&D systems, Abingdon, UK). Reconstitute with sterile PBS and store aliquots at –20◦ C. Reconstituted TNF-α is stable at –20◦ C for 3 months.

Molecular MRI Approaches to the Detection of CNS Inflammation

383

5. Recombinant mouse VCAM-1 Fc chimera (Fc–VCAM1) and recombinant mouse ICAM-1 Fc chimera (Fc– ICAM-1) (R&D systems). Reconstitute with sterile water (50 μg/ml) and store aliquots at –20◦ C. Once reconstituted, Fc–VCAM-1 and Fc–ICAM-1 are stable at –20◦ C for 4 weeks. 6. Paraformaldehyde. Prepare a 1% (w/v) solution fresh for each experiment. 7. Purified monoclonal rat anti-mouse CD106/VCAM-1 antibody (clone M/K2) (Cambridge Bioscience). Store at 4◦ C. 8. Alexa Fluor 488 conjugated rabbit secondary antibody to rat IgG (Vector Laboratories, Peterborough, UK). Store at 4◦ C. 9. Vectashield mounting media containing 4 ,6-diamidino-2phenylindole (DAPI) nuclear stain (Vector Laboratories). Store at 4◦ C. 10. Fluorescence microscope (Nikon, Kingston upon Thames, UK). 11. Laser scanning confocal microscope (Zeiss LSM150). 2.4. In Vitro sLeX –MPIO Binding to TNF-α -Stimulated sEND-1 Cells

1. Dulbecco’s Modified Eagle’s Medium (DMEM) is supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 U penicillin and 0.1 mg/ml streptomycin. Store at 4◦ C. 2. Trypsin/EDTA solution. Store at 4◦ C. 3. Round microscope coverslips (19 mm). 4. Murine recombinant tumour necrosis factor-alpha (TNFα, R&D systems). Reconstitute with sterile PBS and store aliquots at –20◦ C. Reconstituted TNF-α is stable at –20◦ C for 3 months. 5. Paraformaldehyde. Prepare a 1% (w/v) solution fresh for each experiment. 6. Laser scanning confocal microscope (Zeiss LSM150).

2.5. In Vivo Mouse Protocol

1. Adult male NMRI mice (30–35 g). 2. Isofluorane (1.8–2.5%). 3. Nitrous oxide compressed gas (BOC Gases, Guildford, UK). 4. Oxygen compressed gas (BOC Gases). 5. Wild M650 operating microscope (Leica Microsystems, Milton Keynes, UK). 6. Stereotaxic frame (David Kopf, Tujunga, LA, USA).

384

Sibson et al.

7. Glass 1 μl graduated haematocrit tubes pulled with Narishige forge (Narishige International, London, UK) to a tip diameter <50 μm. 8. 1 mg/ml mouse recombinant IL-1β (R&D systems). Reconstitute with sterile PBS and store in aliquots at –20◦ C. Once reconstituted, IL-1β is stable for 3 months at –20◦ C. 9. Low endotoxin saline containing 0.1% BSA. 2.6. In Vivo Rat Protocol

1. Adult male Wistar rats (200–250 g). 2. Isofluorane (1.8–2.5%). 3. Nitrous oxide compressed gas (BOC Gases). 4. Oxygen compressed gas (BOC Gases). 5. Wild M650 operating microscope (Leica Microsystems). 6. Stereotaxic frame (David Kopf). 7. Glass 1 μl graduated haematocrit tubes pulled with Narishige forge (Narishige International) to a tip diameter <50 μm. 8. 1 mg/ml rat recombinant IL-1β (R&D systems). Reconstitute with sterile PBS and store in aliquots at –20◦ C. Once reconstituted, IL-1β is stable for 3 months at –20◦ C. 9. Low endotoxin saline containing 0.1% BSA.

2.7. Magnetic Resonance Imaging

1. Quadrature birdcage coil with an in-built stereotaxic frame. 2. 7 T horizontal bore magnet with a Varian Inova spectrometer (Varian, Inc., Palo Alto, CA, USA). 3. Physiological monitoring and regulation: subcutaneous ECG electrodes (in-house); circulating warm-water system and rectal probe (Harvard Apparatus).

2.8. MRI Data Analysis

1. ImagePro Plus Image analysis software (Media Cybernetics, Marlow, UK). 2. 3D Constructor plug-in for ImagePro Plus (Media Cybernetics).

3. Methods For the synthesis of VCAM–MPIO, mouse monoclonal VCAM-1 antibodies are covalently conjugated to MyOneTM Tosylactivated MPIO (1 μm diameter). Tosylactivated MPIO do not require surface activation, unlike, for example, iron oxide particles with reactive carboxylic acid surface groups, which require activation prior to conjugation. Moreover, the hydrophobic properties of

Molecular MRI Approaches to the Detection of CNS Inflammation

385

tosylactivated MPIO facilitate optimal antibody orientation since Fc regions of the antibody are generally more hydrophobic than the Fab portion and will adsorb to the hydrophobic surface of MPIO followed by rapid covalent bond formation. This exposes the Fab regions of the antibody, maximising the binding potential of antibody-conjugated MPIO to the target protein. The antibody-conjugated MPIO are stable at 4◦ C for several months. For the synthesis of sLeX –MPIO, sialyl-LewisX (sLeX ) is conjugated to the MPIO (1 μm diameter) via surface amine groups. Since this requires a linker molecule on the sLeX , commercially available sLeX is not suitable and instead we synthesise this compound in-house. The capacity of the targeted MPIO constructs for specific and quantitative binding in vitro should be validated prior to in vivo studies. For this purpose, VCAM–MPIO or sLeX – MPIO is incubated with activated mouse endothelial cells, stimulated with graded doses of TNF-α. After extensive washing to remove unbound MPIO, specific MPIO binding to cells can be visualised and quantified using differential interference confocal microscopy. To further validate binding specificity, binding can be pre-blocked, for example, with a soluble chimeric protein containing the extracellular domain of VCAM-1 (Fc– VCAM-1) or with negative control soluble extracellular ICAM-1 (Fc–ICAM-1). Fc-blocked VCAM–MPIO can be incubated with activated endothelial cells as above and subsequently VCAM1 demonstrated using immunofluorescence. Appropriate control MPIO, for example IgG–MPIO, are included to test for non-specific MPIO retention. Cells can be assessed by confocal microscopy for VCAM–MPIO or sLeX –MPIO binding. In VCAM-1 blocking experiments, VCAM-1 expression can be confirmed by immunofluorescence. For in vivo MRI studies, acute brain inflammation is induced by unilateral stereotaxic injection of IL-1β into the left striatum of either mice or rats. After 3 h, VCAM–MPIO or sLeX –MPIO is injected intravenously via a tail vein and allowed to circulate for ca. 1 h prior to MRI. This allows time for specific MPIO binding in the brain and clearance of unbound MPIO from the blood. Control groups of animals may undergo identical treatments with substitution of IgG–MPIO or MPIO conjugated to equivalent, but biologically inactive, carbohydrate molecules. Additional control groups may undergo blocking experiments, for example, VCAM-1 antibody (0.2 mg/kg body weight) can be injected 15 min prior to VCAM–MPIO administration to block VCAM-1 binding sites in vivo. MR images are acquired using a T2 ∗ weighted 3D gradient echo sequence, with a final isotropic resolution of 90–120 μm. For MR image analysis, hypointense signal areas are segmented using an automated histogram-based tool using ImagePro Plus and rendered to create a three-dimensional

386

Sibson et al.

Fig. 19.1. VCAM–MPIO binding pattern in the brain following IL-1β injection. T2 ∗ -weighted 3D gradient echo MR images acquired from mouse brain 1–2 h post-MPIO injection, using a 7 T magnet, approximately 90 μm isotropic resolution. Each animal was stereotaxically injected with IL-1β into the left striatum (1 ng in 1 μl saline), 3 h prior to intravenous injection of either a VCAM–MPIO or b IgG–MPIO (4 × 108 MPIO). a Intense low signal areas in the left hemisphere reflect specific VCAM–MPIO binding on acutely activated vascular endothelium with virtually absent contrast effect in the contralateral control hemisphere. b No contrast effects were observed in control IgG–MPIO-injected animals. c and d Low signal contrast effects were segmented in 41 contiguous MR slices using an automated signal intensity histogram tool using ImagePro Plus. Masks of low signal areas were merged and 3D reconstructed to create a 3D volumetric map of low signal voxels. c VCAM–MPIO contrast effects delineated the architecture of inflamed cerebral vasculature in the IL-1β-stimulated hemisphere (image left) with almost total absence of binding on the contralateral, non-activated side. The midlines are indicated by vertical sections. d No MPIO contrast effects were seen in animals intravenously injected with IgG–MPIO.

volumetric map of MPIO binding in the brain (see Figs. 19.1 and 19.2). 3.1. Conjugation of VCAM-1 Antibody to MyOneTM Tosylactivated MPIO

1. MyOneTM Tosylactivated MPIO (5 mg, 5 × 109 MPIO) are transferred into a 1.5 ml microcentrifuge tube (see Note 7). The tube is placed in a Dynal MPC-S magnet (see Note 8) until MPIO have formed a pellet at the side of the tube and the liquid is clear. The supernatant is discarded. 2. The tube is removed from the magnet and MPIO resuspended in 1 ml of pre-washing and coating buffer (0.1 M sodium borate buffer, pH 9.5). The tube is placed in the magnet to pellet MPIO. The supernatant is removed and this step is repeated once more. 3. The MPIO pellet is resuspended in 200 μg antibody (see Notes 9 and 10). 4. Ammonium sulphate (3 M) is immediately added to give a concentration of 1 M in the final coating solution.

Molecular MRI Approaches to the Detection of CNS Inflammation

387

Fig. 19.2. sLex –MPIO binding pattern in the brain following IL-1β injection. T2 ∗ -weighted 3D gradient echo MR images acquired from rat brain 1–2 h post-MPIO injection, using a 7 T magnet, approximately 120 μm isotropic resolution. Each animal was stereotaxically injected with IL-1β into the left striatum (100 ng in 1 μl saline), 3 h prior to intravenous injection of either a sLeX –MPIO or b untargeted MPIO (4 × 108 MPIO). a Intense low signal areas in the left hemisphere reflect specific sLeX –MPIO binding with little retention in the contralateral control hemisphere. b No contrast effects were observed in control MPIO-injected animals. c and d Low signal contrast effects were segmented in contiguous MR slices using an automated signal intensity histogram tool using ImagePro Plus. Masks of low signal areas were merged and 3D reconstructed to create a 3D volumetric map of low signal voxels for sLeX –MPIO (c) or control MPIO (d). The brain has been surface rendered to aid spatial orientation of MPIO binding within the brain.

5. The tube is placed in a rotating wheel and incubated, with constant head-over-head rotation, at 37◦ C for 20 h. 6. MPIO are pelleted and the supernatant discarded to remove unbound antibody. 7. Blocking buffer is added at the same volume used for coating the MPIO, i.e. in this example 600 μl. 8. The tube is placed in a rotating wheel and incubated, with constant head-over-head rotation, at 37◦ C overnight to block any remaining unbound active tosyl sites. 9. MPIO are pelleted using the magnet and the supernatant discarded. 10. The tube is removed from the magnet and MPIO resuspended in Washing and Storage buffer (1 ml). The tube is placed in a rotating wheel and incubated, with constant head-over-head rotation, at 4◦ C for 5 min. MPIO are pelleted using the magnet and the supernatant discarded. This step is repeated three times.

388

Sibson et al.

11. Antibody-conjugated MPIO are stored in Washing and Storage buffer at concentration of 2.5 × 1010 MPIO/mL. The solution is stable at 4◦ C for several months without loss of antigen binding (see Note 11). 3.2. Conjugation of Sialyl-LewisX to MPIO

1. A suspension of amine-terminated particles (see Note 6) in water (∼40–50 mg/mL, 1 mL) is placed in a microfuge tube in the Dynamag-2 magnet, and the particles are allowed to concentrate against the side of the tube. 2. The supernatant is discarded and the particles resuspended in water (1 mL). 3. Procedures 1 and 2 are repeated two further times. 4. A suspension of these amine-terminated particles in HPLCgrade methanol is placed in a microfuge tube in the Dynamag-2 magnet, and the particles are allowed to concentrate against the side of the tube. 5. The supernatant is discarded and the particles resuspended in HPLC-grade methanol (1 mL). 6. Procedures 4 and 5 are repeated two further times. 7. A 1-thio-S-cyanomethyl-modified carbohydrate, e.g. 1-thio-S-cyanomethyl-sialyl-LewisX reagent [thio-S-cyanomethyl-(5-acetimido-3,5-dideoxy-D-glycero-α-D-galacto2-nonulo pyranosylonic acid-(2→3)-β-D-galactopyranosyl -(1→4)-(1→3)-(α-L-fucosyl)-2-acetimido-2-deoxy-β-Dglucopyranoside [sLeX –SCM] (4 mg, 5 μmol, ∼100 M equiv.) is dissolved in 1 mL of dry HPLC-grade methanol. 8. This solution is added to the washed particles from step 6. 9. Sodium methoxide solution (5 μL of 25% w/w solution in MeOH) is added to this mixture, and the resulting mixture is shaken at <1,000 rpm on an orbital shaker for 2 days. 10. The particles are washed as in steps 4–6 with HPLC-grade methanol. 11. The particles are washed as in steps 1–3 with water. 12. The particles are placed in a microfuge tube in the Dynamag-2 magnet, and the particles are allowed to concentrate against the side of the tube. 13. The supernatant is discarded and the particles resuspended in sterile PBS (1 mL). 14. Procedures 12 and 13 are repeated two further times. 15. The particles are resuspended in sterile PBS (at ∼20– 40 mg/mL particle) (see Note 15).

Molecular MRI Approaches to the Detection of CNS Inflammation

3.3. In Vitro VCAM–MPIO Binding to TNF- α -Stimulated sEND-1 Cells

389

1. Cells of a mouse endothelial cell line, sEND-1, are passaged when approaching confluency with trypsin/EDTA. Cells are plated at a density of 8 × 10 (5) per 35 mm well in a six-well plate, each well containing a sterile 19 mm round coverslip. 2. Cells are stimulated for 20 h at 37◦ C with graded doses of mouse recombinant TNF-α (0–10 ng/ml DMEM). 3. The TNF-α media is then removed by aspiration. 4. Stimulated cells are incubated with VCAM–MPIO or IgG– MPIO (2.5 × 107 MPIO in 2 ml DMEM) in duplicate. The cell plate is immediately placed onto a sample rocker to avoid sedimentation of MPIO and incubated for 30 min at room temperature with continual mixing. 5. The media is removed by aspiration and unbound MPIO removed by extensive washing with PBS. 6. 1% paraformaldehyde (2 ml) is added for 30 min at room temperature to fix cells. 7. The cell coverslip is mounted onto a glass slide by slowly inverting the coverslip onto mounting medium (13 μl) on a microscope slide (see Note 16). Nail varnish is used to seal the sample. 8. MPIO binding to cells is viewed using differential interference contrast microscopy. Four fields of view are acquired per sample. The number of bound MPIO per field is quantified using ImagePro Plus. 9. For blocking experiments, VCAM–MPIO is pre-blocked with 5 μg Fc–VCAM-1 or Fc–ICAM-1 per μg MPIO for 1 h at room temperature. 10. Fc-blocked MPIO (2.5 × 107 MPIO in 2 ml DMEM) are incubated with cells stimulated with 50 ng/ml TNF-α or unstimulated cells, extensively washed with PBS and fixed as described above (see steps 4–6). Experiments are performed in triplicate. 11. Fc-blocked MPIO binding to cells is viewed using an inverted microscope (×20 objective). Four fields of view are acquired per sample. The number of bound MPIO per field is quantified using ImagePro Plus. For immunofluorescent VCAM-1 staining, MPIO-bound cells are incubated with rat anti-mouse VCAM-1 (5 μg per ml PBS) for 1 h at room temperature. 12. Primary antibody is removed and the cells washed three times for 5 min each with PBS.

390

Sibson et al.

13. Cells are incubated with secondary Alexa Fluor 488 conjugated rabbit antibody to rat IgG (1:100) for 1 h at room temperature (see Note 17). 14. The secondary antibody is removed and the cells washed three times for 5 min each with PBS. 15. The cell coverslip is carefully mounted onto a glass slide as described above (see step 7) using mounting media containing DAPI nuclear stain. Cells are kept in the dark at 4◦ C until imaged using confocal microscopy. Slides are viewed on the same day of preparation. 16. Cells are assessed for VCAM–MPIO binding (see Note 18) and VCAM-1 immunofluorescent staining using confocal microscopy (see Note 19). 3.4. In Vitro sLeX –MPIO Binding to TNF- α -Stimulated sEND-1 Cells

1. Cells of a mouse endothelial cell line, sEND-1, are passaged when approaching confluency with trypsin/EDTA. Cells are plated at a density of 8 × 105 per 35 mm well in a six-well plate, each well containing a sterile 19 mm round coverslip. 2. Cells are stimulated for 20 h at 37◦ C with graded doses of mouse recombinant TNF-α (0–10 ng/ml DMEM). 3. The TNF-α media is then removed by aspiration. 4. Stimulated cells are incubated with sLeX –MPIO (2.5 × 107 MPIO in 2 ml DMEM) in duplicate. The cell plate is immediately placed onto a sample rocker to avoid sedimentation of MPIO and incubated for 30 min at room temperature with continual mixing. 5. The media is removed by aspiration and unbound MPIO removed by extensive washing with PBS. 6. 1% paraformaldehyde (2 ml) is added for 30 min at room temperature to fix cells. 7. The cell coverslip is mounted onto a glass slide by slowly inverting the coverslip onto mounting medium (13 μl) on a microscope slide (see Note 16). Nail varnish is used to seal the sample. 8. MPIO binding to cells is viewed using differential interference contrast microscopy. Four fields of view are acquired per sample. The number of bound MPIO per field is quantified using ImagePro Plus.

3.5. In Vivo Mouse Protocol for VCAM–MPIO

1. Mice are anaesthetised using 2.0–2.5% isofluorane in 70% N2 O:30% O2 . 2. Mice are positioned in a stereotaxic frame under the operating microscope.

Molecular MRI Approaches to the Detection of CNS Inflammation

391

3. A midline incision in the scalp is made and a burr hole drilled in the left hemisphere of the skull centred 0.5 mm anterior and 2 mm lateral to Bregma. 4. Using a glass pipette with a tip <50 μm, mouse recombinant IL-1β (1 ng in 1 μl of low endotoxin saline containing 0.1% BSA) is stereotaxically injected into the left striatum, 0.5 mm anterior and 2 mm lateral to Bregma, at a depth of 2.5 mm, over a 10-min period. 5. The wound is sutured. 6. After 3 h, a cannula is inserted into the tail vein for administration of MPIO (4 × 108 MPIO in 100 μl low endotoxin saline containing 0.1% BSA) (see Note 20). VCAM– MPIO (4 × 108 ) is also administered to control mice that receive intracerebral injections of saline or no intracerebral injections. 7. To block VCAM binding sites in vivo, a further group of mice are injected with VCAM-1 antibody (0.2 mg/kg) 3 h after IL-1β injection and VCAM–MPIO is administered 15 min later. 8. Following MPIO injection, mice are positioned in a quadrature birdcage coil with an in-built stereotaxic frame. All mice are closely monitored for any signs of ill-health or toxicity (see Note 21). 3.6. In Vivo Rat Protocol for sLeX –MPIO

1. Rats are anaesthetised using 2.0–2.5% isofluorane in 70% N2 O:30% O2 . 2. Rats are positioned in a stereotaxic frame under the operating microscope. 3. A midline incision in the scalp is made and a burr hole drilled in the left hemisphere of the skull centred 1 mm anterior and 3 mm lateral to Bregma. 4. Using a glass pipette with a tip <50 μm, rat recombinant IL-1β (1 ng in 1 μl of low endotoxin saline containing 0.1% BSA) is stereotaxically injected into the left striatum, 1 mm anterior and 3 mm lateral to Bregma, at a depth of 4 mm, over a 10-min period. 5. The wound is sutured. 6. After 3 h, a cannula is inserted into the tail vein for administration of MPIO (4 × 109 MPIO in 200 μl low endotoxin saline containing 0.1% BSA) (see Note 20). sLeX –MPIO (4 × 109 ) is also administered to control rats that receive equivalent intracerebral injections of saline or no intracerebral injections. 7. Following MPIO injection, rats are positioned in a quadrature birdcage coil with an in-built stereotaxic frame. All rats

392

Sibson et al.

are closely monitored for any signs of ill-health or toxicity (see Note 21). 3.7. Magnetic Resonance Imaging in Mice

1. MRI is performed using a 7 T horizontal bore magnet with a Varian Inova spectrometer. During MRI, anaesthesia is maintained with 1.0–1.5% isofluorane in 70% N2 O:30% O2 , ECG is monitored via subcutaneous electrodes and body temperature maintained at 37◦ C by a circulating warm-water system. 2. The radiofrequency coil is tuned and matched to the proton frequency, and positioning of the mouse brain at the magnet isocentre confirmed with scout images (T2 -weighted fast spin echo sequence; repetition time (TR) = 3 s, echo time (TE) = 40 ms, field of view (FOV) 25 × 25 mm, matrix size 128 × 128, single average). 3. After shimming, a T2 ∗ -weighted 3D data set is acquired using a gradient echo sequence with the following parameters: flip angle 35◦ , repetition time (TR) = 50 ms, echo time (TE) = 5 ms, field of view (FOV) 22.5 × 22.5 × 31.6 mm, matrix size 192 × 192 × 360, two averages. The total acquisition time is approximately 1 h (see Note 22). MRI is started 1–1.5 h after intravenous injection of MPIO. 4. The data are zero-filled to 256 × 256 × 360 and reconstructed off-line, to give a final isotropic resolution of 88 μm3 .

3.8. Magnetic Resonance Imaging in Rats

1. MRI is performed using a 7 T horizontal bore magnet with a Varian Inova spectrometer. During MRI, anaesthesia is maintained with 1.5–1.8% isofluorane in 70% N2 O:30% O2 , ECG is monitored via subcutaneous electrodes and body temperature maintained at 37◦ C by a circulating warm-water system. 2. The radiofrequency coil is tuned and matched to the proton frequency, and positioning of the rat brain at the magnet isocentre confirmed with scout images (T2 -weighted fast spin echo sequence; repetition time (TR) = 3 s, echo time (TE) = 40 ms, field of view (FOV) 35 × 35 mm, matrix size 128 × 128, single average). 3. After shimming, a T2 ∗ -weighted 3D data set is acquired using a gradient echo sequence with the following parameters: flip angle 11◦ , repetition time (TR) = 25 ms, echo time (TE) = 10 ms, field of view (FOV) 30.7 × 30.7 × 42.0 mm, matrix size 192 × 192 × 350, two averages. The total acquisition time is approximately 1 h. MRI is started 1–1.5 h after intravenous injection of MPIO.

Molecular MRI Approaches to the Detection of CNS Inflammation

393

4. The data are zero-filled to 256 × 256 × 350 and reconstructed off-line, to give a final isotropic resolution of 120 μm. 3.9. MRI Data Analysis

1. Extra-cerebral structures in each MR image are manually masked using ImagePro Plus. 2. Low signal areas are segmented in ten evenly spaced slices per brain using the automated signal intensity histogrambased tool in ImagePro Plus to obtain the median low signal intensity value (see Note 23). 3. Low signal areas are segmented in 41 contiguous slices of the brain, spanning a depth of 3.6 mm from the dorsal hippocampus ventrally. To ensure true laterality, the left and right hemispheres are segmented simultaneously, 1 mm from the midline outwards. 4. The median signal intensity value is applied to the 41 slice sequence to correct for minor variations in absolute signal intensity between individual scans. 5. Masks of the segmented low signal areas in 41 contiguous slices are merged and reconstructed using the 3D Constructor plug-in for ImagePro Plus to visualise MPIO binding patterns in the inflamed (left) and non-inflamed (right) cerebral hemispheres. Examples of three-dimensional (3D) volumetric maps of VCAM–MPIO and sLeX –MPIO binding patterns are shown in Figs. 19.1 and 19.2, respectively.

4. Notes 1. MPIO should be kept in liquid suspension during storage at 4◦ C as drying will reduce the performance of MPIO. We found it useful to store the MPIO vial on a sample roller at 4◦ C in order to prevent sedimentation of MPIO. MPIO should never be frozen as this will cause irreversible aggregation. 2. The pre-washing and coating sodium borate buffer must not contain any protein or amino groups (e.g. glycine, Tris) as these will bind to MPIO surface, inhibiting specific antibody binding. A higher pH favours optimal antibody binding. 3. MyOneTM Tosylactivated MPIO can be conjugated to any ligand containing amino or sulphydryl groups (i.e. antibody, protein, peptide or glycoprotein). However, the antibody or protein must be purified since all proteins or amino groups will bind to the MPIO surface.

394

Sibson et al.

4. Preservatives, such as sodium azide, may disturb antibody conjugation to MPIO. In addition, sodium azide is cytotoxic and therefore not suitable for in vivo application. Therefore, antibodies free from stabilisers should be used, or else stabilisers removed from the antibody/protein solution prior to conjugation. 5. The VCAM-1 antibody is commercially supplied in sodium borate buffer, pH 8.0, free from stabilisers. We found this antibody to be excellent for immunofluorescence imaging. 6. 2.8 μm amine-functionalised MPIO are available commercially, but we synthesised our own in-house to maintain the 1 μm size that we have found to be optimal. However, we believe that the commercially available 2.8 μm MPIO would work well. 7. MPIO should be thoroughly resuspended prior to use by vortexing. 8. The Dynal MPC-S magnet can be used to prepare up to six microcentrifuge tubes of antibody-conjugated MPIO simultaneously. 9. For antibody conjugation, a concentration of 40 μg antibody per microgram of MPIO is optimal. Conjugating less than recommended amounts of antibody may cause aggregation of MPIO. We found it best to use a stock purified antibody concentration of 0.5–1 mg/ml in order to achieve a sufficient antibody coating concentration. More dilute antibodies may require methods to concentrate the antibody prior to conjugation with MPIO. 10. MPIO should be suspended in the antibody solution with very efficient mixing using a vortex and transferred immediately to the sample rotating wheel for incubation. MPIO should not be allowed to come out of suspension at any stage. 11. We store our antibody-conjugated MPIO, continually mixing on a sample roller at 4◦ C to avoid sedimentation and drying of MPIO, which would reduce their binding capacity. 12. All reactions involving organometallic or other moisturesensitive reagents were carried out under an argon atmosphere using glassware that was flame dried with a Bunsen burner and cooled under vacuum prior to use. 13. Dry solvents were purchased from Sigma-Aldrich and stored under an atmosphere of argon using an Oxford sure R . All other solvents and reagents were used as supseal cap plied by Sigma-Aldrich unless otherwise stated (analytical or HPLC grade) without prior purification.

Molecular MRI Approaches to the Detection of CNS Inflammation

395

R 14. Water was purified by an Elix UV-10 system and passed through a 22 μm filter for sterilisation.

15. MPIO are light sensitive and therefore should be kept in opaque containers. 16. Air bubbles are undesirable in the mounting medium. Therefore, the coverslip should be inverted slowly onto the mounting media using fine forceps. 17. We found Alexa Fluor 488 conjugated rabbit secondary antibody to be excellent for visualising VCAM-1 immunofluorescent staining by confocal microscopy, as it is not prone to bleaching. This is particularly important when constructing z-stacks. 18. MPIO autofluoresce under confocal microscopy due to their high iron content and can be viewed using either red or green emission. 19. Due to the diameter of the MPIO, it may be difficult to find a focal plane that is suitable for simultaneously visualising bound MPIO and immunofluorescent staining using confocal microscopy. We found it useful to create a z-stack of images throughout the depth of the MPIO and use the merged image to view co-localisation of immunofluorescence and MPIO. 20. For in vivo administration, we found it best to perform intravenous MPIO injections away from the MR magnet. We previously tried injecting MPIO into mice positioned inside the magnet. However, MPIO rapidly came out of solution when they were within the magnetic field and sedimented prior to administration. 21. We have found antibody-conjugated MPIO to be well tolerated in all mice, with no animals showing any signs of ill effect during close observation for up to 3 days post-injection. 22. We have serially imaged the same mouse and found maximal contrast at 1–2 h with diminution by 4 h. 23. Image analysis should be performed by an operator blinded to the origin of data.

Acknowledgments This work is funded by the Medical Research Council (NRS and DA), Cancer Research UK (NRS), the Wellcome Trust (RPC) and Glycoform Ltd (BD, NRS, DA).

396

Sibson et al.

References 1. McAteer, M. A., Sibson, N. R., von Zur Muhlen, C. et al. In vivo magnetic resonance imaging of acute brain inflammation using microparticles of iron oxide. Nat Med 2007;13(10):1253–1258. 2. Serres, S., Anthony, D. C., Jiang, Y. et al. Systemic inflammatory response reactivates immune-mediated lesions in rat brain. J Neurosci 2009;29(15):4820–4828. 3. Sibson, N. R., Blamire, A. M., BernadesSilva, M. et al. MRI detection of early endothelial activation in brain inflammation. Magn Reson Med 2004;51(2):248–252. 4. van Kasteren, S. I., Campbell, S. J., Serres, S., Anthony, D. C., Sibson, N. R., Davis, B. G. Glyconanoparticles allow pre-symptomatic in vivo imaging of brain disease. Proc Natl Acad Sci USA 2009;106(1):18–23. 5. McAteer, M. A., Schneider, J. E., Ali, Z. A. et al. Magnetic resonance imaging of endothelial adhesion molecules in mouse atherosclerosis using dual-targeted microparticles of iron oxide. Arterioscler Thromb Vasc Biol 2008;28(1):77–83. 6. von Zur Muhlen, C., Sibson, N. R., Peter, K. et al. A contrast agent recognizing acti-

7.

8.

9.

10.

vated platelets reveals murine cerebral malaria pathology undetectable by conventional MRI. J Clin Invest 2008;118(3):1198–1207. von zur Muhlen, C., Peter, K., Ali, Z. A. et al. Visualization of activated platelets by targeted magnetic resonance imaging utilizing conformation-specific antibodies against glycoprotein iib/iiia. J Vasc Res 2009;46(1): 6–14. von zur Muhlen, C., von Elverfeldt, D., Moeller, J. A. et al. Magnetic resonance imaging contrast agent targeted toward activated platelets allows in vivo detection of thrombosis and monitoring of thrombolysis. Circulation 2008;118(3):258–267. Briley-Saebo, K. C., Johansson, L. O., Hustvedt, S. O. et al. Clearance of iron oxide particles in rat liver: Effect of hydrated particle size and coating material on liver metabolism. Invest Radiol 2006;41(7): 560–571. Pittet, M. J., Swirski, F. K., Reynolds, F., Josephson, L., Weissleder, R. Labeling of immune cells for in vivo imaging using magnetofluorescent nanoparticles. Nat Protoc 2006;1(1):73–79.

wwwwwww

Chapter 20 Brain Redox Imaging Ken-ichiro Matsumoto, Fuminori Hyodo, Kazunori Anzai, Hideo Utsumi, James B. Mitchell, and Murali C. Krishna Abstract Nitroxyl contrast agents (nitroxyl radicals, also known as nitroxide) are paramagnetic species, which can react with reactive oxygen species (ROS) to lose paramagnetism to be diamagnetic species. The paramagnetic nitroxyl radical forms can be detected by using electron paramagnetic resonance imaging (EPRI), Overhauser MRI (OMRI), or MRI. The time course of in vivo image intensity induced by paramagnetic redox-sensitive contrast agent can give tissue redox information, which is the so-called redox imaging technique. The redox imaging technique employing a blood–brain barrier permeable nitroxyl contrast agent can be applied to analyze the pathophysiological functions in the brain. A brief theory of redox imaging techniques is described, and applications of redox imaging techniques to brain are introduced. Key words: Redox imaging, nitroxyl contrast agents, Overhauser MRI, electron paramagnetic resonance imaging, reactive oxygen species, hypoxia, hyperoxia.

1. Introduction 1.1. Redox-Sensitive Contrast Agent

Nitroxyl contrast agents (nitroxyl radicals, also known as nitroxide) are paramagnetic species, which have one unpaired electron. Structures of several commonly used nitroxyl radicals are shown in Fig. 20.1. Nitroxides belong to two main structural groups, i.e. five-member ring (pyrrolidine derivatives or so-called PROXYL derivatives) and six-member ring (piperidine derivatives or so-called TEMPO derivatives). Substitutional groups on the pyrrolidine ring of the PROXYL derivatives or the piperidine ring of the TEMPO derivatives determine the in vivo distribution of

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_20, © Springer Science+Business Media, LLC 2011

397

398

Matsumoto et al.

Fig. 20.1. Structures and common names of commercialized nitroxyl radicals. a PROXYL derivatives and b TEMPO derivatives.

these agents as well as their sub-cellular distribution. Nitroxyl contrast agents with appropriate substitutional groups on the ring can be directed to intracellular locations, such as the mitochondria, or restricted to extracellular spaces or even penetrate the blood–brain barrier (BBB) and localize intracellularly. The membrane permeability of nitroxyl contrast agents can be relatively easily modified by chemical substituents on pyrrolidine or piperidine rings of the molecule. Various nitroxyl contrast agents which have distinct membrane permeabilities have been reported. For example, the carboxy-PROXYL can not go through the BBB; however, methoxy-carbonyl-PROXYL (also known as MC-PROXYL or CxP-M), which is synthesized by esterification of carboxy-PROXYL, can readily go through the BBB (1, 2). The nitroxyl contrast agents are stable free radical species at room temperature in the solid form or even in solutions at physiological pH. However, the nitroxyl radical can be chemically and/or enzymatically reduced to the corresponding hydroxylamine form (Fig. 20.2). The nitroxyl radicals can also be oxidized to oxoammonium cation by reactive oxygen species (ROS), such as superoxide (·O2 – ) and hydroxyl radical (·OH). The oxoammonium cation can be reduced by superoxide by a oneelectron reduction back to the nitroxyl radical form or by two electrons to the corresponding hydroxylamine form by reduced glutathione (GSH), reduced nicotinamide adenine dinucleotide (NADH), and/or reduced nicotinamide adenine dinucleotide

Brain Redox Imaging

399

Fig. 20.2. Reactive oxygen species (ROS), such as superoxide (·O2 – ) and hydroxyl radical (·OH), undergo one electron reduction of nitroxyl radicals in the presence of hydrogen donors (H-donors), such as reduced nicotinamide adenine dinucleotide (NADH), and/or reduced nicotinamide adenine dinucleotide phosphate (NADPH). The nitroxyl radicals were once oxidized to an oxoammonium cation by ROS and then reduced to hydroxylamine by receiving a hydrogen atom from H-donors. Reduced glutathione (GSH) and oxoammonium cations may make another complex irreversibly. Therefore, the generation of total ROS can be estimated by a reduction of the nitroxyl radical under coexisting GSH.

phosphate (NADPH) (3, 4). The hydroxylamines can be again re-oxidized to the corresponding nitroxyl radicals under oxidative conditions (5). When a nitroxyl radical is administrated to an experimental animal, the nitroxyl radical is mainly metabolized to the corresponding hydroxylamine form (6). This intrinsic in vivo one-electron reduction of nitroxyl radical can be modified (amplified or retarded) by abnormal tissue redox conditions, such as hypoxia, hyperoxia, and/or several oxidative stress conditions accompanying excess ROS generation. In addition to the redox sensitivity of nitroxyl radicals, BBB permeability with suitable substituents makes them suitable for imaging redox reactions in the brain. 1.2. Redox Properties of Nitroxyl Contrast Agent

The reactivity of nitroxyl radicals with ROS is, in other words, that nitroxyl radicals can have antioxidant properties. In fact, several anti-oxidative properties of nitroxyl radicals have been reported (7–11). Leading such studies, the radio-protective activ-

400

Matsumoto et al.

ity of TEMPOL also has been studied prosperously (12–16). The radio-protective mechanism of TEMPOL has been considered that is based on its anti-oxidative property, i.e., the reaction of TEMPOL with ROS generated during irradiation of ionizing radiation, such as X-ray and/or γ-ray. In addition, the superoxide dismutase (SOD)-like activity of TEMPOL and its reaction mechanism has been reported (17–19). The currently known reaction mechanisms of nitroxyl radicals and ROS are summarized in Fig. 20.2. As described above, the process of one-electron reduction of nitroxyl radicals to corresponding hydroxylamines goes along with two steps, at first one-electron oxidation, then two-electron reduction. A recent study showed that the reaction of oxoammonium cation and GSH mainly gives a stable complex rather than hydroxyl radicals, although the structure of the complex has not been identified yet (20). However, the main in vivo reaction of nitroxyl radicals is the one-electron reduction to the hydroxylamine (6). Hydroxylamine can be re-oxidized to nitroxyl radical by some oxidants, such as ·OH and/or Fe3+ , in vivo. This redox cycle reactions support SOD mimetic catalytic activity, which may underlie its radio-protective effect of TEMPOL. The hydroxylamine form of TEMPOL, i.e., TEMPOL-H, does not have radio-protective activity for cell (12, 21). However, TEMPOL-H has a radio-protective effect, when TEMPOLH was administered to a mouse (22). This is because of in vivo re-oxidation of hydroxylamine to the radio-protective nitroxyl radical. The re-oxidation of hydroxylamine to the nitroxyl radical, which is an oxygen-dependent reaction, cannot occur in a hypoxic tumor tissue. Therefore no radioprotective effects of nitroxyl radicals (23, 24) in hypoxic tumors can be expected because of their facile reduction to hydroxylamines. In contrast, slight nitroxyl radical can be continuously available in normoxic normal tissues after dynamic equilibrium between reduction of nitroxyl radical and oxidation of hydroxylamine. Recently, a significantly fast decay of TEMPOL in tumor tissue compared with other normal tissues was reported using an MR-based redox imaging technique (24). It is suggested that nitroxyl radicals can be proposed as a normal tissue-selective radio-protector. The redox properties of nitroxyl radicals can give us double benefits which are (i) as redox-sensitive contrast agents and (ii) as a normal tissue-selective radio-protector. Both benefits are useful for a first diagnosis of a tumor and radiation therapy planning. These will give a radio-protective treatment for normal tissues in subsequent radiotherapy. Therefore, the magnetic resonancebased redox imaging techniques using nitroxyl contrast agents can have wide possibility.

Brain Redox Imaging

1.3. Applying Nitroxyl Contrast Agent to Brain Redox Imaging

401

Redox imaging techniques using electron paramagnetic resonance imaging (EPRI), Overhauser MRI (OMRI), or MRI (25, 26) can visualize and estimate the redox status in living experimental animals. The basis of redox imaging is to obtain a time course of image intensity, which is induced by paramagnetic redox-sensitive contrast agent. One suitable approach of redox imaging techniques is a brain image (27). The relation between ROS and several brain disorders, such as stroke, Parkinson’s disease, Alzheimer’s disease, is well established (28–30). Likewise, the relation between memory loss and radiation-induced oxidative injuries in brain/neuronal system has also been reported (31–33). Therefore, the redox imaging techniques employing a BBB permeable nitroxyl contrast agent can be another approach to analyze the pathophysiological function in the brain. A brief theory of redox imaging techniques is described, and applications of redox imaging techniques to the brain are introduced.

2. Redox Imaging in EPRI Paramagnetic nitroxyl radicals can be directly detected by EPR, while the hydroxylamines which are the reduction products of nitroxyl radicals are diamagnetic and therefore EPR silent species. Therefore, detecting a decrease in EPR signal intensity from nitroxyl radicals due to their reduction in biological/chemical samples can be used to monitor redox reactions in vivo. Nitroxyl radicals have been utilized as in vivo molecular probes to estimate redox status in the tissue of experimental animals using in vivo EPR (34, 35). Using low frequency (300–1,200 MHz) EPR spectrometers (36–39) or X-band EPR spectrometers with special sampling techniques for in vivo measurements (40–43) the signal decay and/or the spectral change of nitroxyl spin probe in live animals can be monitored. Combining with EPR imaging techniques using field gradients, the distribution and time course of nitroxyl spin probes in tissues can be visualized (44–47). Information related to tissue redox status using nitroxyl agents can be visualized by attaching some additional dimensions to an EPRI, such as time axis and/or spectral axis, to a simple image of spatial distribution of the probe molecule (Fig. 20.3). The time axis can be easily achieved by sequential measurement of several images of the spatial distribution of the spin probe. For example, when the image intensities

402

Matsumoto et al.

Fig. 20.3. A schematic drawing of the concept of dynamic imaging. The additional time dimension to the spatial mapping of the paramagnetic contrast agent, functional information, such as pharmacokinetics of the contrast agent, can be tagged on to the EPR image. The time axis is by sequential measurement of several EPR images. Consequently, pixel-wise decay rates of EPR image intensity can be obtained.

decrease/increase in a time-dependent manner, a mapping of pixel-wise decay/increment rates of image intensities can be observed and the slopes of the image intensity change can be used to generate the “redox map” of the region of interest in an in vivo experiment Most in vivo EPR spectrometers operate in a continuouswave (CW) modality, where the sample is irradiated continuously at low RF powers and the magnetic field is slowly swept to obtain an EPR spectrum. The temporal resolution of EPR spectroscopy and imaging is limited by the time of a relatively slow magnetic field scan. Recent developments in hardware and new data acquisition techniques achieved EPRI data acquisition capabilities in the range of seconds (48–50). Such a rapid data acquisition will improve the temporal resolution of EPRI in in vivo biological applications. However, the larger EPR linewidths of nitroxyl molecules and the available magnetic field gradients for imaging make the spatial resolution limited. In addition, no anatomical information is available by EPRI. To provide anatomic

Brain Redox Imaging

403

co-registration to EPR images of nitroxyl distribution in vivo, fusion techniques of EPRI and MRI have been developed by several groups (51–53).

3. Redox Imaging in OMRI OMRI, which is also known as proton electron double resonance imaging (PEDRI), is a double-resonance technique involving both EPR and NMR (54–56). At a given magnetic field, by virtue of their stronger magnetic moments, the electrons’ spin levels are polarized to a greater extent than the corresponding spin levels of nuclei. When an aqueous solution containing a paramagnetic agent, such as a nitroxyl radical, is placed in a magnetic field within a resonator assembly which is tuned to the resonant frequencies of both the electron and the nucleus, a significant enhancement of the NMR signal results via the Overhauser effect (57) when the NMR signal of 1 H of H2 O is collected after a period of irradiation at the EPR frequency. Comparing the images obtained with the EPR-irradiation (OMRI signal) and without irradiation (native MRI signal), the paramagnetic contrast agents used can be quantified based on intensity enhancement patterns. The imaging of nitroxyl contrast agents using OMRI was started by Lurie and colleagues in the late 1980s (54, 55). OMRI is a double resonance imaging technique. It relies on an EPR irradiation pulse followed by an MRI imaging sequence. In this technique, when a paramagnetic species, such as nitroxyls, are present in tissue, irradiating by an EPR pulse at the resonance frequency of the nitroxyl enhances the signal intensity of the surrounding water protons. This makes the image resolution dependent on MRI which is intrinsically a higher resolution technique compared to EPR imaging allowing the imaging of free radicals with ∼ millimeter resolution. In addition, OMRI techniques can obtain not only distribution but also EPR spectral information of nitroxyl contrast agents (58). Nitroxyl radicals give a doublet or triplet EPR spectrum by hyper-fine-splitting due to 15 N or 14 N nuclear spin. Each of these can be separately obtained in the form of OMRI signal by using a different frequency for EPR excitation of either the 14 N or the 15 N nuclei. Utsumi et al. (58) proposed a novel imaging technique to separate the reduction and oxidation reactions in an identical sample simultaneously using a nitroxyl contrast agent and its hydroxylamine form labeled by different isotopes (i.e., 14 N and 15 N). They also showed that simultaneous reactions occurred separately inside and outside of liposomes using membrane permeable and impermeable nitroxyls. OMRI is an EPR-dependent spectroscopic imaging technique that displays

404

Matsumoto et al.

a high temporal and spatial resolution. It hence affords the rapid imaging of in vivo metabolism of multiple nitroxyl radical probes simultaneously.

4. Trial of EPR-Based Brain Redox Imaging

Using a suitable BBB-permeable nitroxyl contrast agent, the time course of nitroxyl reduction rates in the brain can be visualized. CxP-M (also known as MC-PROXYL: 3-methoxy-carbonyl-2,2,5,5-tetramethyl-pyrrolidine-1-oxyl) (2, 59, 60) and CxP-AM (acetoxymethyl-2,2,5,5-tetramethylpyrrolidine-1-oxyl-3-carboxylate) were designed as BBBpermeable spin probe for EPR imaging (2). CxP-AM can be metabolized to BBB-impermeable CxP by esterases in the brain and can also be retained in the brain (2). CxP-AM has been commercialized and is currently available from LABOTEC Co. (Tokyo, Japan). Anzai et al. (59) reported the distribution of CxP-M in the rat brain using autoradiography with 14 C-labeled CxP-M and 3D EPRI. Figure 20.4a shows a scheme of chemical synthe-

Fig. 20.4. Autoradiograms of axial sections of rat head obtained after 3 min i.v. or 15 min after i.p. treatment with 14 C-labeled MC-PROXYL or 14 C-labeled carbamoyl-PROXYL. The black region shows high radioactivity and the white, no radioactivity, showing that MC-PROXYL, but not carbamoyl-PROXYL, can penetrate the BBB and is well distributed to brain tissue.

Brain Redox Imaging

405

sizing 14 C-labeled CxP-M. Figure 20.4b shows autoradiograms of coronal sections of rat head obtained after administration with 14 C-CxP-M or 14 C-carbamoyl-PROXYL. They suggested that CxP-M is a suitable paramagnetic probe for the study of free radical reactions in the brain using in vivo EPR imaging. Lee et al. (60) reported a faster nitroxyl decay in the brain of spontaneously hypertensive rats compared with Wistar rat using the in vivo EPRI technique with CxP-M. Miyake et al. (61) reported the development of other BBB-permeable nitroxyl radical probes designed for the purpose of oxygen mapping in the brain. Molecular oxygen with two unpaired electrons is paramagnetic and by collisional interaction with the paramagnetic nitroxyls, broadens their EPR spectra in a concentration-dependent manner via a T2 contrast process typical in MRI. The oxygendependent spectral broadening can be monitored by EPR imaging allowing the determination of tissue oxygen. Figure 20.5 shows an example of EPR-based brain redox imaging in the mouse brain (2). From two small ROIs selected in the brain area, the result of EPR imaging showed that the different parts of the brain have different decay rates of the nitroxyl

Fig. 20.5. Time-resolved 2D EPR images in the head of a mouse. A 100 mL volume of CxP-AM (25 mM) in PBS with 10% (v/v) of ethanol was injected into the tail veins of mice and the data for images were collected at various time points. a Data collected 0.9–9.9 min after injection. b Data collected 10.4–19.3 min after injection, c Data collected 19.8–28.7 min after injection. d A semilogarithmic plot of the imaging intensity of two different regions (ROI 1 and ROI 2) in the brain after intravenous injection of CxP-AM.

406

Matsumoto et al.

radical. However, anatomical information is not obtained from the EPR imaging experiment. The exact anatomical location of the ROIs in the brain cannot be estimated directly from the EPR image itself. Co-registering of EPR images of nitroxyl radical distribution with anatomical images from MRI is therefore necessary to reliably visualize the regions of interest where the redox reactions occur.

5. NMR-Based Redox Imaging If one needs to monitor the levels of nitroxyl probes, EPR spectral information is not required. In such applications, larger linewidths and spectral multiplicity of the EPR spectra would be interference in attempts to obtain the spatial distribution of nitroxyl radicals in vivo. Paramagnetic nitroxyl radicals, which have proton T1 shortening effect, can be detected using T1 -weighted MRI. The feasibility of nitroxyl radicals as T1 contrast agents in MRI has been examined in the early 1980s, before their use in the field of EPR imaging. However, nitroxyl radicals were found to be not optimal as MR contrast agents in 1980s due to their low relaxivity and rapid in vivo bio-reduction. Now, recent MRI scanners operating at higher magnetic field with better signal to noise ratio and timeefficient T1 -weighted pulse sequences once again make it useful to reconsider nitroxyl radicals as potential T1 contrast agents. Several problems (e.g., low spatial resolution, low temporal resolution, small sample volume) typical in EPR imaging experiments can be overcome using T1 contrast-based MRI. Two unique features distinguish nitroxyl radicals from Gd-based T1 contrast agents. (i) Being organic molecules of molecular weight ∼ 200 Da, they can be specifically directed to extra- or intracellular locations. (ii) Unlike Gd-complexes which do not undergo biological redox transformations, nitroxides not only undergo reversible biological redox transformations to diamagnetic species but also depend on cellular redox status (62, 63), oxygenation status (38, 64), focal inflammatory processes involving ROS (65). Figure 20.6 shows a simulation of T1 -weighted gradient echo MR contrast as a function of concentration of a nitroxyl radical. The relaxivity of the nitroxyl radical used in this calculation was 0.13 mM–1 s–1 at 37◦ C. At such a low relaxivity, the relationship between T1 enhancement and concentration of a contrast agent is almost linear especially in the lower concentration region below 2 mM. Figure 20.7 is a comparison of EPR imaging and T1 -weighted MR imaging of nitroxyl radicals undergoing chemical reduction at a rate determined by the concentration of the

Brain Redox Imaging

407

Fig. 20.6. Relationship between a paramagnetic nitroxyl contrast agent and an intensity change of T1 -weighted SPGR-based MRI. a Simulated M% increased with the concentration of paramagnetic nitroxyl contrast agent. b Low concentration region (<1.5 mM) in (a) (dotted rectangle area) was expanded. Fairly good linearity (R2 = 0.9995) was obtained between simulated M% and the concentration of the low concentration region. Values used for the simulation were as follows: r1 = 0.13 mM–1 s–1 , M0 = 1,000, TR = 75 ms, TE = 3 ms, FA = 45◦ , T1i = 2,350 ms, and T2 ∗ = 50 ms.

reductant, ascorbic acid (AsA). Solutions of a nitroxyl radical and ascorbic acid were loaded in individual syringes and mixed rapidly to initiate the reaction in a cylindrical reactor placed within the resonator. Although MRI has a higher spatial and temporal resolution than EPR imaging, the decay rate was similar to that from EPR imaging (66). This decay rate obtained from MR T1 -contrast can be treated as a pseudo-decay rate (67). In vivo nitroxyl decay rate imaging in SCC tumor implanted in a mouse leg was examined using MRI (66). Figure 20.8a indicates the location of a 2 mm axial slice, including the SCC tumor bearing leg and the normal leg. Figure 20.8b shows a time course of nitroxyl-induced T1 -weighted signal enhancement, M%, after the injection of the nitroxyl contrast agent, carbamoyl-PROXYL, and the T2 map of the corresponding slice. The T1 -weighted signal enhancement of a given pixel at (x, y) coordination, M%x,y is as follows:

408

Matsumoto et al.

Fig. 20.7. Comparison of EPR and MR redox imaging demonstrated using a phantom. Upper cartoon is a representation of the experimental device.

M %x,y = [ (Sx,y after CA injection)/ (Sx,y before CA injection) − 1] × 100

(1)

where Sx,y is an image intensity of pixel (x, y). Image intensity enhancement via T1 -contrast was enhanced in both normal and tumor tissues after the administration of carbamoyl-PROXYL and

Brain Redox Imaging

409

Fig. 20.8. In vivo MR redox imaging. a Position of the slice scanned. b A time course of T1 -weighted signal enhancement in the slice and a T2 -mapping of the identical slice as a scout image. c A comparison of the time course of T1 -weighted signal enhancement in the ROI-1 and ROI-2. d A redox map calculated based on the time course of nitroxylinduced tissue T1 -weighted signal enhancement.

then gradually decreased with time. Figure 20.8c shows the time course of average M% in ROI-1 and ROI-2. The tumor region showed faster decay rate than the normal region. The decay rate maps (Fig. 20.8d) showed clear difference between tumor region and normal tissues. There is a question whether the in vivo decay of MR T1 -contrast is a reduction of nitroxyl radical to the corresponding hydroxylamine or just a simple clearance of nitroxyl radical from the tissue. Chemically, the hydroxylamine can be oxidized to the corresponding nitroxyl radical by adding mild oxidant, such as potassium ferricyanide (FeK3 (CN)6 ) after which the total amount of the nitroxyl agent in the tissue samples (nitroxyl radical form plus hydroxylamine form) can be measured and quantified by EPR (Fig. 20.9). From a set of five animals, whose tissue (both normal and tumor) were analyzed for nitroxyl content

410

Matsumoto et al.

Fig. 20.9. Re-oxidation of a hydroxylamine to the corresponding nitroxyl radical by a chemical oxidant. Ferricyanide (FeK3 (CN)6 ) has been usually used to oxidize hydroxylamine to the nitroxyl radical.

Fig. 20.10. Time course of total amount (nitroxyl radical form + hydroxylamine form) of a contrast agent, carbamoyl-PROXYL, in normal and tumor tissues after i.v. injection.

Brain Redox Imaging

411

after chemical oxidation, it was found that the time course of the total amount of the nitroxyl in both tumor and normal tissues was stable during the time period used in the imaging experiment (Fig. 20.10). These observations suggest that the in vivo disappearance of T1 -contrast induced by a nitroxyl contrast agent as a function of time is due to a reduction (67, 68).

6. Recent Trial of NMR-Based Brain Redox Imaging

Figure 20.11 shows nitroxyl-induced T1 contrast in the head region of mice. These images show varying distributions of nitroxyl contrast agents in the brain depending on their membrane permeability. Highly membrane permeable TEMPOL and CxP-M showed significant T1 contrast induction in whole brain area. Carbamoyl-PROXYL (CmP), which has a limited membrane permeability, showed modest T1 contrast in the brain, while the membrane impermeable CxP displayed no induction of T1 contrast in the brain. These results show that a membrane permeable nitroxyl contrast agent, such as TEMPOL or CxP-M, in combination with their antioxidant properties and their ability to participate in intracellular redox reactions, can be useful with MRI to assess tissue redox status in vivo in various organs of interest, including the brain. Hyodo et al. (27) investigated the ability of CxP-M as a BBB-permeable MRI contrast agent in rats to monitor the production of reactive oxygen species (ROS) in the brain after reperfusion following defined periods of ischemia. In addition, they demonstrated fast T1 mapping acquired by a Look–Locker (LL) sequence (69). T1 -weighted signal enhancement was readily noticed throughout the brain after administering CxP-M. Maximum T1 -weighted signal enhancement occurred at about 3 min after injection of CxP-M and returned to baseline by 15 min (Fig. 20.12a, c). The BBB-impermeable CxP, however, showed no enhancement of the T1 -weighted signal in the brain region (Fig. 20.12b, d). Figure 20.12e shows the effect of cessation of blood flow on the reduction of CxP-M. When the blood flow was stopped by a KCl injection 40 s after administration of CxP-M, the respiration stopped, i.e., the rat died 15–20 s after KCl injection. MR signal intensity in the brain is still decreased after the death of a rat, except the reduction rate became slower than that found with normal cerebral perfusion. Disruption of blood perfusion and/or subsequent tissue injury, such as an ischemia-reperfusion (IR) model, can make abnormal redox status. For example, the reduction rate of CxP-M in the right hemisphere after reperfusion following ischemic periods was

412

Matsumoto et al.

Fig. 20.11. Nitroxyl-induced T1 contrast at the head part of mice. a TEMPOL, b CxP-M, c carbamoyl-PROXYL, and d CxP showed difference of distribution in brain region.

found to be significantly lower compared to that observed in the sham group (Fig. 20.12f) suggesting either a lower reduction rate of nitroxyls or a faster re-oxidation of the hydroxylamines to nitroxyls. While T1 is a quantitative value, the T1 -weighted signal enhancement ratio is not always quantitative especially in a tissue of relatively short T1 . Figure 20.13ashows a time course of T1 maps of the rat head region scanned before and after injection of CxP-M. T1 maps were obtained every 20 s with 13 slices

Brain Redox Imaging

413

Fig. 20.12. Time-course SPGR MR images of rat head region after injection of a CxP-M (cell-permeable) and b CxP (cellimpermeable). Contrast agents were injected 2 min after the MR scan was started. Sixty serial images were obtained in 20 min. The SPGR MR parameters were as follows: image resolution was 256 × 128 zero-filled to 256 × 256 (0.125 mm resolution), FOV = 3.2 × 3.2 cm, slice thickness = 2.0 and 0.2 mm gap. The number of slices was six. The green color shows % enhancement of MR signal intensity, M%. The time courses of intensity change of c CxP-M and d CxP in the ROI of cerebral cortex (red color) and thalamus (blue color) are shown. e Intensity change of CxP-M without blood flow is shown. KCl (2 mL) was injected 40 s after CxP-M injection and rats died within 20 s.

414

Matsumoto et al.

Fig. 20.13. MRI signal dynamic of a SLENU and b SLCNUgly in the brain and surrounding tissues after i.v. injection in mice (0.4 mmol/kg b.w.). Each image was obtained within 20-s intervals using a Gradient-echo T1 -weighted MRI. The red color in the images is the extraction of the signal between every single slide and the averaged baseline signal (first five slides – before injection). The red and black colors in the chart represent the MRI signal dynamic in the brain or entire area, respectively. A representative image from three independent experiments is shown in the figure.

(1.5 mm slice thickness) by the EPI-LL sequence. The T1 maps clearly showed the difference of T1 relaxation time between cerebral cortex and thalamus (Fig. 20.13a). The T1 relaxation times in cerebral cortex and thalamus before injection were 1,577 ± 19.9 and 1,315 ± 11.9 ms, respectively (Fig. 20.13b). The T1 values of these regions rapidly decreased after injection of CxPM, and the minimum T1 values of cerebral cortex and thalamus regions were 1,034 ± 70.4 and 779 ± 70.2 ms, respectively. The recovery of T1 in the cerebral cortex and thalamus showed similar trends. According to the T1 changes before and after injection of CxP-M and in vitro relaxivity of CxP-M (0.16 mM–1 s–1 ), the expected in vivo concentrations of CxP-M were calculated for all time points (Fig. 20.13c). The maximum concentrations in cerebral cortex and thalamus 30 s after injection were 1.9 ± 0.35 and 3.0 ± 0.50 mM, respectively. The CxP-M concentration

Brain Redox Imaging

415

calculated in the brain based on the T1 maps was in agreement with the results directly measured by ex-vivo EPR.

7. MR Imaging of BBB-Permeable Drugs

A recent study demonstrated the feasibility of using a combination of nitroxyl radical linked with a therapeutic drug and the use of T1 -weighted MRI to examine the BBB permeability of a nitroxyl labeled drug (70). This study demonstrates a BBB permeability for two TEMPO-labeled analogs (SLENU and SLCNUgly) of the anticancer drug Lomustine [1-(2-chloroethyl)-3-cyclohexyl1-nitrosourea] using MRI at 7 T. SLENU and SLCNUgly were rapidly transported and randomly distributed in the brain tissue, which indicated that the exchange of the cyclohexyl part of Lomustine with TEMPO radical did not suppress the BBB permeability of the anticancer drug. In drug delivery imaging, nitroxyl-labeled anti-tumor drugs can therefore work not only as a tracer but may also have a role as a diagnostic probe for tumor physiology.

8. Conclusion Brain redox imaging using a nitroxyl contrast agent as described in this chapter has a great potential to visualize the redox status in brain tissue. Non-invasive detection and visualization of the redox status in the brain will be an important contribution to the study of the pathophysiology of neurodegenerative diseases, such as stroke, Parkinson’s disease, and Alzheimer’s disease. Future therapeutic applications in these diseases can also be considered. Developing brain redox imaging techniques will be crucial to analyze redox phenomena and their influence on brain functions. References 1. Sano, H., Matsumoto, K., Utsumi, H. Synthesis and imaging of blood-brain-barrier permeable nitroxyl-probes for free radical reactions in brain of living mice. Biochem Mol Biol Int 1997;42:641–647. 2. Sano, H., Naruse, M., Matsumoto, K., Oi, T., Utsumi, H. A new nitroxyl-probe with high retention in the brain and its applica-

tion for brain imaging. Free Radic Biol Med 2000;28:959–969. 3. Samuni, A. M., DeGraff, W., Krishna, M. C., Mitchell, J. B. Nitroxides as antioxidants: Tempol protects against EO9 cytotoxicity. Mol Cell Biochem 2002;234/235:327–333. 4. Krishna, M. C., Grahame, D. A., Samuni, A., Bitchell, J. B., Russo, A. Oxoammonium

416

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

Matsumoto et al. cation intermediate in the nitroxide-catalyzed dismutation of superoxide. Proc Natl Acad Sci USA 1992;89:5537–5541. Ui, I., Okajo, A., Endo, K., Utsumi, H., Matsumoto, K. Effect of hydrogen peroxide in redox status estimation using nitroxyl spin probe. Free Radic Boil Med 2004;37:2012–2017. Takeshita, K., Utsumi, H., Hamada, A. Whole mouse measurement of paramagnetism-loss of nitroxide free radical in lung with a L-band ESR spectrometer. Biochem Mol Biol Int 1993;29:17–24. Miura, Y., Utsumi, H., Hamada, A. Antioxidant activity of nitroxide radicals in lipid peroxidation of rat liver microsomes. Arch Biochem Biophys 1993;300:148–156. Blonder, J. M., McCalden, T. A., Hsia, C. J., Billings, R. E. Polynitroxyl albumin plus tempol attenuates liver injury and inflammation after hepatic ischemia and reperfusion. Life Sci 2000;67:3231–3239. Samuni, A. M., DeGraff, W., Krishna, M. C., Mitchell, J. B. Nitroxides as antioxidants: Tempol protects against EO9 cytotoxicity. Mol Cell Biochem 2002;234–235:327–333. Herrling, T., Jung, K., Fuchs, J. Measurements of UV-generated free radicals/reactive oxygen species (ROS) in skin. Spectrochim Acta A 2006;63:840–845. Venditti, E., Scirè, A., Tanfani, F., Greci, L., Damiani, E. Nitroxides are more efficient inhibitors of oxidative damage to calf skin collagen than antioxidant vitamins. Biochim Biophys Acta 2008;1780:58–68. Mitchell, J. B., DeGraff, W., Kaufman, D., Krishna, M. C., Samuni, A., Finkelstein, E., Ahn, M. S., Hahn, S. M., Gamson, J., Russo, A. Inhibition of oxygen-dependent radiation-induced damage by the nitroxide superoxide dismutase mimic, tempol. Arch Biochem Biophys 1991;289:62–70. Goffman, T., Cuscela, D., Glass, J., Hahn, S., Krishna, C. M., Lupton, G., Mitchell, J. B. Topical application of nitroxide protects radiation-induced alopecia in guinea pigs. Int J Radiat Oncol Biol Phys 1992;22: 803–806. Miura, Y., Anzai, K., Ueda, J., Ozawa, T. Novel approach to in vivo screening for radioprotective activity in whole mice: In vivo electron spin resonance study probing the redox reaction of nitroxyl. J Radiat Res 2000;41:103–111. Vitolo, J. M., Cotrim, A. P., Sowers, A. L., Russo, A., Wellner, R. B., Pillemer, S. R., Mitchell, J. B., Baum, B. J. The stable nitroxide tempol facilitates salivary gland protection during head and neck irradia-

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

tion in a mouse model. Clin Cancer Res 2004;10:1807–1812. Anzai, K., Ueno, M., Yoshida, A., Furuse, M., Aung, W., Nakanishi, I., Moritake, T., Takeshita, K., Ikota, N. Comparison of stable nitroxide, 3-substituted 2,2,5,5tetramethylpyrrolidine-N-oxyls, with respect to protection from radiation, prevention of DNA damage, and distribution in mice. Free Radic Biol Med 2006;40:1170–1178. Samuni, A., Mitchell, J. B., DeGraff, W., Krishna, C. M., Samuni, U., Russo, A. Nitroxide SOD-mimics: Modes of action. Free Radic Res Commun 1991; 12–13:187–194. DeGraff, W. G., Krishna, M. C., Russo, A., Mitchell, J. B. Antimutagenicity of a low molecular weight superoxide dismutase mimic against oxidative mutagens. Environ Mol Mutagen 1992;19:21–26. Krishna, M. C., Russo, A., Mitchell, J. B., Goldstein, S., Dafni, H., Samuni, A. Do nitroxide antioxidants act as scavengers of O2 – or as SOD mimics? J Biol Chem 1996;271:26026–26031. Matsumoto, K., Okajo, A., Nagata, K., DeGraff, W. G., Nyui, M., Ueno, M., Nakanishi, I., Ozawa, T., Mitchell, J. B., Krishna, M. C., Yamamoto, H., Endo, K., Anzai, K. Detection of free radical reactions in an aqueous sample induced by low linearenergy-transfer irradiation. Biol Pharm Bull 2009;32:542–547. Hahn, S. M., Krishna, C. M., Samuni, A., DeGraff, W., Cuscela, D. O., Johnstone, P., Mitchell, J. B. Potential use of nitroxides in radiation oncology. Cancer Res 1994;54:2006s–2010s. Hahn, S. M., Krishna, M. C., DeLuca, A. M., Coffin, D., Mitchell, J. B. Evaluation of the hydroxylamine tempol-H as an in vivo radioprotector. Free Radic Biol Med 2000;28:953–958. Hahn, S. M., Sullivan, F. J., DeLuca, A. M., Krishna, C. M., Wersto, N., Venzon, D., Russo, A., Mitchell, J. B. Evaluation of tempol radioprotection in a murine tumor model. Free Radic Biol Med 1997;22:1211–1216. Cotrim, A. P., Hyodo, F., Matsumoto, K., Sowers, A. L., Cook, J. A., Baum, B. J., Krishna, M. C., Mitchell, J. B. Differential radiation protection of salivary glands versus tumor by tempol with accompanying tissue assessment of tempol by magnetic resonance imaging. Clin Cancer Res 2007;13:4928–4933. Subramanian, S., Matsumoto, K., Mitchell, J. B., Krishna, M. C. Radio frequency

Brain Redox Imaging

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

continuous-wave and time-domain EPR imaging and overhauser-enhanced magnetic resonance imaging of small animals: Instrumental developments and comparison of relative merits for functional imaging. NMR Biomed 2004;17:263–294. Matsumoto, K., Subramanian, S., Murugesan, R., Bitchell, J. B., Krishna, M. C. Spatially resolved biologic information from in vivo EPRI, OMRI, and MRI. Antioxid Redox Signal 2007;9:1125–1141. Hyodo, F., Chuang, K. H., Goloshevsky, A. G., Sulima, A., Gfiffiths, G. L., Mitchell, J. B., Koretsky, A. P., Krishna, M. C. Brain redox imaging using blood-brainbarrier-permeable nitroxide MRI contrast agent. J Cereb Blood Flow Metab 2008;28:1165–1174. Slemmer, J. E., Shacka, J. J., Sweeney, M. I., Weber, J. T. Antioxidants and free radical scavengers for the treatment of stroke, traumatic brain injury and aging. Curr Med Chem 2008;15:404–414. Yokoyama, H., Kuroiwa, H., Yano, R., Araki, T. Targeting reactive oxygen species, reactive nitrogen species and inflammation in MPTP neurotoxicity and parkinson’s disease. Neurol Sci 2008;29:293–301. Guan, Z. Z. Corss-talk between oxidative stress and modification of cholinergic and glutaminergic receptors in the pathogenesis of Alzheimer’s disease. Acta Pharmacol Sci 2008;29:773–780. Manda, K., Ueno, M., Anzai, K. Space radiation-induced inhibition of neurogenesis in the hippocampal dentate gyrus and memory impairment in mice: Ameliorative potential of the melatonin metabolite, AFMK. J Pineal Res 2008;45:430–438. Thotala, D. K., Hallahan, D. E., Yazlovitskaya, E. M. Inhibition of glycogen synthase kinase 3β attenuates neurocognitive dysfunction resulting from cranial irradiation. Cancer Res 2008;68:5859–5868. Fuller, C. D., Schillerstrom, J. E., Jones, W. E., 3rd, Boersma, M., Royall, D. R., Fuss, M. Prospective evaluation of pretreatment executive cognitive impairment and depression in patients referred for radiotherapy. Int J Radiat Oncol Biol Phys 2008;72:529–533. Utsumi, H., Muto, E., Masuda, S., Hamada, A. In vivo ESR measurment of free radicals in whole mice. Biochem Biophys Res Commun 1990;172:1342–1348. Utsumi, H., Yamada, K. In vivo electron spin resonance-computed tomography/nitroxyl probe technique for noninvasive analysis of oxidative injuries. Arch Biochem Biophys 2003;416:1–8.

417

36. Phumala, N., Ide, T., Utsumi, H. Noninvasive evaluation of in vivo free radical reactions catalyzed by iron using in vivo ESR spectroscopy. Free Radic Biol Med 1999;26:1209–1217. 37. Kamatari, M., Yasui, H., Ogata, T., Sakurai, H. Local pharmacokinetic analysis of a stable spin probe in mice by in vivo L-band ESR with surface-coil-type resonator. Free Radic Res 2002;36:1115–1125. 38. Ilangovan, G., Li, H., Zweier, J. L., Kuppusamy, P. In vivo measurement of tumor redox environment using EPR spectroscopy. Mol Cell Biochem 2002;234/235:393–398. 39. Yasukawa, K., Kasazaki, K., Hyodo, F., Utsumi, H. Non-invasive analysis of reactive oxygen species generated in rats with water immersion restraint-induced gastric lesions using in vivo electron spin resonance spectroscopy. Free Radic Res 2004;38: 147–155. 40. Takechi, K., Tamura, H., Yamaoka, K., Sakurai, H. Pharmacokinetic analysis of free radicals by in vivo BCM (blood circulation monitoring)-ESR method. Free Radic Res 1996;26:483–496. 41. Matsumoto, K., Krishna, M. C., Mitchell, J. B. Novel pharmacokinetic measurement using X-band EPR and simulation of in vivo decay of various nitroxyl spin probes in mouse blood. J Pharmacol Exp Ther 2004;310:1076–1083. 42. Matsumoto, K., Okajo, A., Kobayashi, T., Mitchell, J. B., Krishna, M. C., Endo, K. Estimation of free radical formation by beta-ray irradiation in rat liver. J Biochem Biophys Methods 2005;63:79–90. 43. Okajo, A., Matsumoto, K., Mitchell, J. B., Krishna, M. C., Endo, K. Competition of nitroxyl contrast agents as an in vivo tissue redox probe: Comparison of pharmacokinetics by the bile flow monitoring (BFM) and blood circulating monitoring (BCM) method suing X-band EPR and simulation of decay profiles. Magn Reson Med 2006;56:422–431. 44. Kuppusamy, P., Afeworki, M., ShankarR, A., Coffin, D., Krishna, M. C., Hahn, S. M., Mitchell, J. B., Zweier, J. L. In vivo electron paramagnetic resonance imaging of tumor heterogeneity and oxygenation in murine model. Cancer Res 1998;58:1562–1568. 45. Kuppusamy, P., Li, H., Ilangovan, G., Cardounel, A. J., Zweier, J. L., Yamada, K., Krishna, M. C., Mitchell, J. B. Noninvasive imaging of redox status and its modification by tissue glutathione levels. Cancer Res 2002;62:307–312.

418

Matsumoto et al.

46. Ilangovan, G., Li, H., Zweier, J. L., Krishna, M. C., Mitchell, J. B., Kuppusamy, P. In vivo measurement of regional oxygenation and imaging of redox status in RIF-1 murine tumor: Effect of carbogen-breathing. Magn Reson Med 2002;48:723–730. 47. Yamada, K., Kuppusamy, P., English, S., Yoo, J., Irie, A., Subramanian, S., Mitchell, J. B., Krishna, M. C. Feasibility and assessment of non-invasive in vivo redox status using electron paramagnetic resonance imaging. Acta Radiol 2002;43:433–440. 48. Ohno, K., Watanabe, M. Electron paramagnetic resonance imaging using magneticfield-gradient spinning. J Magn Reson 2000;143:274–279. 49. Deng, Y., He, G., Petryakov, S., Kuppusamy, P., Zweier, J. L. Fast EPR imaging at 300 MHz using spinning magnetic field gradients. J Magn Reson 2004;168: 220–227. 50. Subramanian, S., Koscielniak, J. W., Devasahayam, N., Pursley, R. H., Pohida, T. J., Krishna, M. C. A new strategy for fast radiofrequency CW EPR imaging: Direct detection with rapid scan and rotating gradients. J Magn Reson 2007;186:212–219. 51. Di Giuseppe, S., Placidi, G., Sotgiu, A. New experimental apparatus for multimodal resonance imaging: Initial EPRI and NMRI experimental results. Phys Med Biol 2001;46:1003–1016. 52. He, G., Deng, Y., Li, H., Kuppusamy, P., Zweier, J. L. EPR/NMR co-imaging for anatomic registration of free radical images. Magn Reson Med 2002;47:571–578. 53. Hyodo, F., Yasukawa, K., Yamada, K., Utsumi, H. Spatially resolved time-course studies of free radical reactions with an EPRI/MRI fusion technique. Magn Reson Med 2006;56:938–943. 54. Lurie, D. J., Bussell, D. M., Bell, L. H., Mallard, J. R. Proton electron double resonance imaging of free radical solutions. J Magn Reson 1988;76:366–370. 55. Lurie, D. J., Nicholson, I., Foster, M. A., Mallard, J. R. Free radicals imaged in vivo in the rat by using proton-electron doubleresonance imaging. Phil Trans R Soc Lond 1990;A333:453–456. 56. Krishna, M. C., English, S., Yamada, K., Yoo, J., Murugesan, R., Devasahayam, N., Cook, J. A., Golman, K., Ardenkjaer-Larsen, J. H., Subramanian, S., Mitchell, J. B. Overhauser enhanced magnetic resonance imaging for tumor oximetry: Coregistration of tumor anatomy and tissue oxygen concentration. Proc Natl Acad Sci USA 2002;99:2216–2221.

57. Overhauser, A. W. Polarization of nuclei in metals. Phys Rev 1953;92:411–415. 58. Utsumi, H., Yamada, K., Ichikawa, K., Sakai, K., Kinoshita, Y., Matsumoto, S., Nagai, M. Simultaneous molecular imaging of redox reactions monitored by overhauserenhanced MRI with 14n- and 15n-labeled nitroxyl radicals. Proc Natl Acad Sci USA 2006;103:1463–1468. 59. Anzai, K., Saito, K., Takeshita, K., Takahashi, S., Miyazaki, H., Shoji, H., Lee, M. C., Masumizu, T., Ozawa, T. Assessment of ESR-CT imaging by comparison with autoradiography for the distribution of a bloodbrain-barrier permeable spin probe, MCPROXYL, to rodent brain. Magn Reson Imaging 2003;21:765–772. 60. Lee, M. C., Shoji, H., Miyazaki, H., Yoshino, F., Hori, N., Miyake, S., Ikeda, Y., Anzai, K., Ozawa, T. Measurement of oxidative stress in the rodent brain using computerized electron spin resonance tomgraphy. Magn Reson Med Sci 2003;2:79–84. 61. Miyake, N., Shen, J., Liu, S., Shi, H., Liu, W., Yuan, Z., Pritchard, A., Kao, J. P. Y., Kiu, K. J., Rosen, G. M. Acetoxymethoxycarbonyl nitroxides as electron paramagnetic resonance proimaging agents to measure O2 levels in mouse brain: A pharmacokinetic and pharmacodynamic study. J Pharmacol Exp Ther 2006;318:1187–1193. 62. Yamada, K., Inoue, D., Matsumoto, S., Utsumi, H. In vivo measurement of redox status in streptozotocin-induced diabetic rat using targeted nitroxyl probes. Antioxid Redox Signal 2004;6:605–611. 63. Tsutsumi, T., Ide, T., Yamato, M., Kudou, W., Andou, M., Hirooka, Y., Utsumi, H., Tsutsui, H., Sunagawa, K. Modulation of the myocardial redox state by vagal nerve stimulation after experimental myocardial infarction. Cardiovasc Res 2008;77:713–721. 64. Okajo, A., Ui, I., Manda, S., Nakanishi, I., Matsumoto, K., Anzai, K., Endo, K. Intracellular and extracellular redox environments surrounding redox-sensitive contrast agents under oxidative atmosphere. Biol Pharm Bull 2009;32:535–541. 65. Utsumi, H., Yasukawa, K., Soeda, T., Yamada, K., Shigemi, R., Yao, T., Tsuneyoshi, M. Noninvasive mapping of reactive oxygen species by in vivo electron spin resonance spectroscopy in indomethacin-induced gastric ulcers in rats. J Pharmacol Exp Ther 2006;317:228–235. 66. Matsumoto, K., Hyodo, F., Matsumoto, A., Koretsky, A. P., Sowers, A. L., Mitchell, J. B., Krishna, M. C. High resolution mapping of tumor redox status by magnetic res-

Brain Redox Imaging onance imaging using nitroxides as redoxsensitive contrast agents. Clin Cancer Res 2006;12:2455–2462. 67. Matsumoto, K. Utility decay rates of T1 -weighted MRI contrast based on redox-sensitive paramagnetic nitroxyl contrast agents. Biol Pharm Bull 2009;32: 711–716. 68. Hyodo, F., Matsumoto, K., Matsumoto, A., Mitchell, J. B., Krishna, M. C. Probing the intracellular redox status of tumors with magnetic resonance imaging and redox-sensitive contrast agents. Cancer Res 2006;66:9921–9928.

419

69. Chuang, K. H., Korestsky, A. Improved neuronal tract tracing using manganese enhanced magnetic resonance imaging with fast T1 mapping. Magn Reson Med 2006;55:604–611. 70. Zhelev, Z., Bakalova, R., Aoki, I., Matsumoto, K., Gadjeva, V., Anzai, K., Kanno, I. Nitroxyl radicals for labeling of coventional therapeutics and non-invasive magnetic resonance imaging of their permeability for blood-brain barrier: Relationship between structure, blood clearance, and MRI signal dynamic in the brain. Mol Pharm 2009;6:504–512.

wwwwwww

Chapter 21 Systems Biology Approach to Imaging of Neural Stem Cells ´ and Mirjana Maletic-Savati ´ Li Hua Ma, Yao Li, Petar M. Djuric, c´ Abstract Over the past decade, the advances in human brain magnetic resonance imaging (MRI) have significantly improved our ability to gain insightful information about the structure and function of the brain. One of the MRI imaging modalities that still awaits more comprehensive data mining is magnetic resonance spectroscopy (MRS). MRS provides information on the functional status of the brain tissue and can detect metabolic abnormalities that precede structural changes. The chemical specificity of proton MRS (1 H-MRS) allows detection of several biomarkers that are specific for neurons (N-acetyl aspartate, NAA) and astrocytes (myoinositol (mI) and choline (Cho)), the two most abundant cell types present in the brain tissue. However, apart from a dozen metabolites, current methodologies utilized for MRS analysis do not allow further biomarker discoveries. Herein, we introduce a bioinformatics approach to MRS data processing and discuss possible discoveries that such approach may provide. Specifically, we describe the methodology for neural stem/progenitor cell (NPC) detection in vitro and in vivo, utilizing metabolomic profiling and singular value decomposition analyses. Key words: NMR, MRS, metabolomics, neuroprogenitors, hippocampus, multivariate statistics, singular value decomposition.

1. Introduction The ability to identify human neural stem/progenitor cells (NPCs) by brain imaging may have profound implications for diagnostic, prognostic, and therapeutic purposes. The study of human NPC in vivo has been hindered by the absence of welldefined markers that can distinguish them from other neural cell types. Recently, this obstacle has been approached from a systems biology point of view; namely metabolomics (1).

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_21, © Springer Science+Business Media, LLC 2011

421

422

Ma et al.

Metabolomics investigates a global metabolic pool of cells, tissues, and fluids. While metabonomics aims to measure the global, dynamic metabolic response of living systems to biological stimuli or genetic manipulation, metabolomics represents an analytical description of complex biological samples, aiming to characterize and quantify all small molecules in such a sample (2). These small molecules represent a “metabolome,” a metabolic state as regulated by net interactions between the genes and the environment (2). Two major modes of data acquisition in metabolomics are nuclear magnetic resonance spectroscopy (NMR) and mass spectrometry (MS) (3). Despite the relative lower sensitivity and specificity, NMR has several advantages over MS, in particular that it could be translated to in vivo studies utilizing the NMR correlate, magnetic resonance spectroscopy (MRS). However, due to inherent limitations of both NMR and MRS instrumentation and spectral analysis, specific care needs to be taken to ensure that data acquisition is as optimal as it can be. Precise data acquisition and processing are most critical when differences between classes approach the limitations of instrument precision. Studies where the key metabolites are present in high concentrations may not require great sensitivity, whereas studies of less abundant metabolites will require optimizing sensitivity and signal-to-noise ratios (SNR). Thus, new NMR and MRS methodologies geared toward a specific application of interest have to provide standardized experimental and mathematical approaches to enable reliable and reproducible data generation across different users. In this chapter, we focus on NMR and MRS data acquisition and analysis aimed to identify a variety of metabolic biomarkers in neural cell types in vitro and brain tissue in vivo. Specifically, we describe our methodology for detection of metabolites enriched in NPC in both in vitro and in vivo conditions. We describe spectral preprocessing, statistical multivariate analysis for potential biomarker discovery, and analytical methods used for identification and quantification of individual metabolites. We also note the specific empirical lessons we have learned from our studies. We propose that our methodology and lessons we learned potentially provide the foundation of general approaches, amenable for further modifications based on specific experimental hypothesis, and thus not only for detection of NPC and metabolomic profiling of the human brain.

2. Materials 2.1. Reagents

1. Phosphate buffered saline (PBS). 2. PBS + 10% antibiotic (Invitrogen Cat. #15140-122).

Systems Biology Approach to Imaging of Neural Stem Cells

423

3. Epithelial growth factor (EGF). Prepare a stock solution of 20 μg/mL of rhEGF (Sigma Cat. #E9644, powder at 200 g/vial) by adding 0.1 mL of sterile 10 mM acetic acid (Mallinckrodt Cat. #V193-05) with 0.1% BSA (Sigma Cat. #A1470) to initially dissolve the rhEGF powder and then add 9.9 mL of NeuroCult basal medium with 10% proliferation supplements. Store stock solution as 0.5 mL aliquots at –20◦ C. Add 1 μL of 20 μg/mL rhEGF to every 1 mL of NeuroCult basal medium with 10% proliferation supplements to give a final concentration of 20 ng/mL of rhEGF. Mix well. 4. NeuroCult proliferation supplements (StemCell Technologies Cat. #05701) should be thawed for 1 h at 37◦ C before addition to the NeuroCult basal medium (StemCell Technologies Cat. #05700). A white precipitate may form during storage at –20◦ C which will disappear upon thawing at 37◦ C. Add the entire volume of NeuroCult proliferation supplements (50 mL) to one bottle (450 mL) of NeuroCult basal medium. Avoid repeated exposure of medium to room temperature and light. If the entire volume is not needed immediately, aliquot into appropriate volumes to be used within 1 week. 5. NeuroCult proliferation medium (StemCell Technologies Cat. #05700) with EGF (20 ng/mL). 6. NeuroCult chemical dissociation kit (StemCell Technologies Cat. #05707). 7. Trypan blue. 8. D2 O, >99.9% atom% D. 9. 4,4-Dimethyl-4-silapentane-1-sulfonic acid (DSS, Cambridge isotope laboratories) for calibration of NMR spectra. 2.2. Equipments

1. For metabolomic experiments of cultured cells: NMR spectrometer (500–900 MHz, Bruker or Varian). NMR fingerprinting can be done at any operating frequencies, but the higher, the better. Use a 5-mm diameter 1 H-detection NMR probe and NMR processing software (Topspin 2.1 (Bruker); HiRes (http://mrs.cpmc.columbia.edu/hires.html)). 2. For metabolomic experiments on animal brain in vivo: 9.4 T mMRI scanner (Biospec Avance 94/20 as, Bruker Biospin MRI, Inc.) with operating system Paravision 5.0 and newer versions. Higher magnetic strength may also be used; however, MRS acquisition might be technically more difficult. 3. For metabolomic experiments on human brain in vivo: 3 T MRI scanner (Phillips, Siemens, GE). 4. Software: Excel for data normalization; SIMCA-P, version 11 (Umetrics AB) for multivariate statistical analysis.

424

Ma et al.

3. Methods 3.1. NPC Preparation In Vitro

1. NPCs are derived from the mouse embryos at gestational day E12–13. It is very important to dissect the embryos at this stage and not later, as properties of NPC change with age. Thus, to derive NPC at E12-13 consistently, one needs to establish a timed-pregnancy colony. NPC can be derived from both wild type and transgenic animals in which NPCspecific promoter nestin drives the expression of fluorescent proteins, as the expression of these proteins does not affect the metabolomic profile of cultured NPCs. 2. Mother is euthanized in the CO2 chamber. 3. Abdomen is washed with 70% ethanol (EtOH), opened, and the embryos are removed. They are transferred to a 10-cm dish with cold PBS, pH 7.25 and 10% antibiotics, placed on ice. This part can be done on the lab bench, particularly if using heterozygote colony of transgenic mice. Thus, embryos that express fluorescent proteins may be selected under the microscope. 4. The selected embryos are transferred to the tissue culture room. The embryos are taken out of the sac and rinsed twice with cold PBS, pH 7.25 with 10% antibiotics. 5. The remaining of the procedure is done in a laminar flow hood under sterile conditions. Embryonic brains are dissected out and transferred to a 35-mm plate containing cold PBS. Using tweezers, the tissue is triturated to obtain a fine suspension (make special care not to introduce air bubbles into the tissue suspension). The suspension is filtered through a 40-μm strainer (BD Cat. # 352340) directly into a 50-mL tube. 6. The cells are centrifuged at 1,000 rpm (or 200×g) for 5 min. Supernatant is removed and the pellet resuspended smoothly (no bubbles) in NeuroCult proliferation medium with EGF (20 ng/mL). Count cell numbers (a 1/5 or 1/10 dilution could be necessary) with Trypan Blue and hemacytometer. Plate 5 × 106 cells per 75 cm2 flask and maintain the cells in the tissue culture incubator at 37◦ C and 5% CO2 . 7. After 2 days in culture, the cell aggregates are visible. Collect the medium containing floating neurospheres and centrifuge them at 1,000 rpm for 5 min. Discard the medium and cell debris and proceed with the NeuroCult chemical dissociation kit (StemCell Technologies Cat. #05707) to get single cell suspension. If necessary, use a cell strainer to discard non-disaggregated neurospheres.

Systems Biology Approach to Imaging of Neural Stem Cells

425

8. Count cell numbers (a 1/5 or 1/10 dilution could be necessary) with Trypan Blue and hemacytometer. Again, plate 5 × 106 cells per 75-cm2 flask and maintain them at 37◦ C/5% CO2 incubator. 9. Keep neurospheres at 100–200 μm in diameter using the NeuroCult chemical dissociation kit twice per week. As the neurosphere size increases, the possibility of hypoxia in the neurosphere core also increases. This may jeopardize the “healthiness” of the neurosphere as cells in the core may undergo cell death. Thus, regular dissociation and replating are necessary. 10. After two passages/dissociations (approx. 7 days of culture), proceed with experiments. Cultured neurospheres obtained and maintained the way described here may be used for up to 10 passages and optimally by 6–7 passages. Thus, we routinely prepare fresh batch of neurosphere cultures once a month. 3.2. NMR Data Acquisition

1. Dissociate the neurospheres to single cell level. The cell quantity required to obtain a reliable and reproducible metabolomic profile using the NMR depends on the magnet strength. For our studies, we used 10 × 105 to 106 NPC to obtain the NPC-associated signal utilizing 700 MHz NMR (4). Furthermore, all experiments are done in triplicate within the same culture (i.e., on the same day) to ensure reproducibility, with N ≥ 3 different cultures to ensure specificity (see Note 1). Control solutions are run with each experiment (see Note 2). 2. Centrifuge the sample at 1,000g for 1–2 min at 4◦ C to remove the culture media. Use PBS to resuspend the pellet. Add 0.45 ml PBS and 0.05 ml D2 O with 0.3 mM DSS as concentration standard and chemical shift reference (calibrated to 0.00 ppm). The sample may be recovered after the spectra collection and stored at –80◦ C for several weeks to allow further measurements by NMR or other analytical techniques. 3. Set acquisition temperature for NMR between 26 and 36◦ C depending on the type of analysis. Load sample into probe and leave sufficient time to equilibrate. Pay special attention to use powder-free gloves when handling NMR tubes to avoid fatty and other deposits from bare hands onto the glass (see Note 3). 4. Tune and match the probe. Set the RF carrier frequency offset value to the H2 O resonance and determine the water saturation power. Transfer these settings to the experiments to be submitted. Select 1D NOESY-presat pulse sequence

426

Ma et al.

(on a Bruker instrument, this is called noesypr1d). Record an NMR spectrum. The time required to obtain the consistent spectra depends on the magnet strength and can vary from 3 to 30 min. Figure 21.1 shows NMR of the NPC acquired at 800 MHz NMR (Bruker) with specific emphasis on the 1.28 ppm peak enriched in NPC (4) (see Note 4).

Fig. 21.1. Spectral profile of live cultured neural progenitor cells (NPC; 100,000 cells) obtained using 800 MHz NMR and 10 min acquisition. Arrow indicates the spectral peak of a metabolite enriched in NPC, resonating at 1.28 ppm. Inset shows enlarged spectral area between 0.8 and 1.4 ppm. Lactate doublet (1.33 ppm) is visible, as well as a CH3 associated peak (0.88 ppm).

3.3. microMRI Data Acquisition

1. Sprague-Dawley adult rats are anesthetized with isoflurane and allowed to breathe spontaneously. Anesthesia is maintained with 1–2% isoflurane in a 1:1 O2 /air mixture delivered via a T-piece. 2. To reduce salivation and maintain optimal conditions for spontaneous ventilation, glycopyrulate (0.1 mg/kg) is given prior to positioning within the mMRI scanner. Rodent head is immobilized by a custom-build head holder and positioned on a 1 cm RF surface coil. Initial hydration is given by an intraperitoneal injection of lactated ringer (4 cc/kg/h). Fluids are continuously administered through the intraperitoneal catheter. Heart rate, body temperature, and respiratory rate are continuously monitored. The rodent ears are covered with cotton/gauze custom-build ear fittings to protect against noise from the MRM gradients during scanning.

Systems Biology Approach to Imaging of Neural Stem Cells

427

3. The MR acrylic holder with the anesthetized and monitored rodent and RF coil is positioned in the center of the 9.4 T magnet. Physiological monitoring is continued from afar via a computerized system (PC-SAM, SA Inc.) which also allows 1 H-MRS data acquisition using cardiac and respiratory scan synchronizations. 4. T2-weighted fast spin echo MR images (TR/TE 2,000/20 ms) are collected with in-plane spatial resolution of 100–200 μm and consecutive 500 μm slices. Single 1 H-MRS voxel-of-interest (VOI) is centered over the region of interest. For investigations of NPC, the VOI is placed within the hippocampus. Voxel size is approximately 2.5 × 2.5 × 2.5 mm3 , tailored to the size of animal’s brain. 5. Data are acquired using a double spin echo sequence (Point RESolved Spectroscopy, PRESS, 8,192 acquisitions, 4 kHz bandwidth, TR/TE=2,000/20 ms, NEX=1,024 water suppressed spectra, total acquisition time = 60 min). The T2 decay of the unsuppressed water signal from the PRESS experiment is measured at 10 different echo times to determine the metabolite concentrations corrected for the partial volume of CSF (5). This approach allows minimization of inter- and intra-subject variability of the 1 H-MRS data sets as it allows measuring of the water content in vivo with a precision of 1.5% (5) (see Note 5). 3.4. MRI Data Acquisition

1. The MRI imaging of human subjects is done at 3 T MRI scanner (Phillips Achieva). The hippocampal voxel is planned using a 3D T1-weighted, inversion-prepared, spoiled gradient echo image, acquired axially and reformatted in the sagittal and oblique coronal planes to visualize the hippocampus (sequence parameters: TE/TR/TI = 5/8/400, matrix = 256 × 256, resolution 1 mm isotropic, SENSE factor 1.6, 160 slices). 2. Hippocampal voxel is placed along the full length of hippocampus, as determined on axial, sagittal, and coronal sections. The VOI size is tailored to the individual brain and is approximately 27.5 × 10 × 10 mm3 for the adult brain. 3. Pre-scan calibrations include an in-house developed, secondorder iterative shimming procedure to obtain minimal linewidth and maximal separation of metabolic components and an adiabatic, hyperbolic secant CHESS water suppression pulse utilizing an in-house developed optimized search algorithm for obtaining complete water suppression and minimal baseline deviation. Minimal line-width should be between 10 and 15 Hz.

428

Ma et al.

4. The 1 H-MRS is preformed using a PRESS sequence with the following acquisition parameters: TE/TR = 30/2,000 ms, spectral width 2 kHz, 1,024 points, NEX 128, and a total scan time of approximately 5 min (see Note 5). 3.5. Spectral Data Processing for Metabolomics Analysis

1. Preprocessing converts the raw spectra into an appropriate form for the statistical analysis. In general, preprocessing includes data import and organization by highresolution MRS (HiRes). In HiRes, NMR spectra acquired from Bruker spectrometers are imported automatically or by manually entering parameters. Series of spectra are imported in a single step. All the spectra and associated information are organized in a project and listed in a table that can be easily sorted by each column. 2. Spectral data processing procedures can be applied either to all the spectra or to the selected individual spectra. Free induction decays (FIDs) are zero-filled or truncated to desired, specified length. Binning is assigned, and spectra are divided into small regions or bins. The averaged intensity of each bin is used to generate a list of variables that represent the original spectrum. Spectra are compared based on the same bins or variables. Constant phase and linear phase corrections are performed. The disadvantages of binning are loss of spectral resolution, generation of possible artifacts caused by signals on the border of bin boundaries, and compensation of intensities within individual bins. Temporal filtering (Gaussian 5) is performed and then the baseline is corrected to remove the baseline distortion and obtain the correct signal intensities. The next step in data processing is water suppression, resonance adjustment (ppm), and calibration or peak alignment (0.00 ppm of DSS). 3. The spectra are then exported from HiRes in the text format into Excel, where they are normalized. Normalization is essential to reduce the systemic variations in different samples and bring them to a more comparable scale. Spectra can be normalized to the total spectral integral or to the specific reference metabolite, such as creatine – a single peak around 2.98–3.08 ppm. We prefer to use the former method, as it is more general, particularly when analyzing in vitro data and spectra from different species where exact metabolite identities are not known for all the peaks. Furthermore, creatine amplitude may differ in disease conditions and thus may skew the data interpretation. The normalized file is saved as delimited unicode text file for multivariate statistical analysis. 4. Multivariate statistical analysis of metabolite fingerprints is performed by SIMCA 11 P. The statistical methods used to analyze sets of metabolite NMR/MRS fingerprints fall into

Systems Biology Approach to Imaging of Neural Stem Cells

429

two main categories: unsupervised methods such as principal component analysis (PCA) and supervised methods such as partial least squares-discriminant analysis (PLS-DA). PCA is an unsupervised analysis which highlights the differences in a data set using a smaller number of factors, called principal components (PC), which contain the most variance through the data set (6). The collection of all NMR or 1 H-MRS spectra can be thought of as a matrix with a size of N × K dimensions, where N is the number of observations (individual patient spectra) and K is the number of variables, often integrals over small segments of the spectra, which can be narrowed down to the individual digital points of the spectra (6). The first principal component (PC) is the vector that accounts for the most variation in the data, and each observation is projected onto this line. The second PC is represented by another vector in K dimensional space, orthogonal to the first PC, and represents the second largest variation in the data set. Many PCs may be obtained from a single data set; however, the variation quickly tapers off within a few consequent PCs. The PCA data are presented as a score plot, which reveals clustering in the observations based on similarities in the spectra. Additionally, a loadings plot is obtained from the PCA, which assists in the interpretation of the score plot by highlighting the spectral regions responsible for clustering (7). In 1 H-MRS data, it is expected that actual peaks, not noise, are the highest contributing factors. 5. Forms of scaling. The results of both PC and PLS modeling are scale dependent. Selecting the scaling of the variables is therefore an important step in fitting projection models such as PC or PLS. The data are scaled to reduce the dominance of large spectral features. If one has no prior information about the importance of the variables, autoscaling all variables to unit variance (UV) is recommended. This is equivalent to making all of the variable axes have the same length, which is, giving all of the variables equal importance. In UV, the variable j is centered and scaled to “unit variance,” i.e., the base weight is computed as 1/sdj, where sdj is the standard deviation of variable j computed around the mean. If one has prior information about the importance of the variables, it may be desirable to scale some variables up or down by modifying their unit variance weights. A modifier can be selected (default = 1.0) that changes the scaling of a variable relative to its base weight. Block scaling can also be specified. The SIMCA default for all selected variables is a UV base weight and a block scaling factor of 1 (no block scaling) and thus needs to be changed if one desires to apply modified scaling. Another form of scaling is Pareto scaling. There, the variable j is centered and scaled to Pareto variance, i.e.,

430

Ma et al.

the base weight is computed as 1/sqrt (sdj), where sdj is the standard deviation of variable j computed around the mean. Pareto scaling is in between no scaling and UV scaling and gives the variable a variance equal to its standard deviation instead of unit variance. The scaling weight of the variable may also be frozen (will not be recomputed when observations in the workset change or the variable metric is modified after the freezing) (see Note 6). 6. The default workset, at the project start, is the whole data set with variables defined as Xs and Ys as specified at import and scaled to UV. The associated model (unfitted) is listed in the active area. To fit a model with excluded variables, transformations, or different scaling, it is necessary to first modify the workset. An unfitted model is generated by SIMCA-P when the workset is specified (select a starting workset new or as model). It uses the whole original primary data set with Xs and Ys as defined at import. The next step is to include or exclude observations or group them into classes for classification. X/Y variables are defined, transformation is performed, and scaling is selected. A modifier can be selected (default = 1.0) that changes the scaling of a variable relative to its base weight. Block scaling can also be specified. Trimming/Winsorizing the workset does not affect the data set as a whole but just that particular workset. The model level options are specified and then the model is fitted. 7. The scores plot is generated by PCA or PLS-DA. PC of the given table (or parts of it) gives vectors of scores, with values tia, which summarize all the variables entering the analysis. It is customary to calculate two or three score vectors, up to 10, and then plot them against each other (tt-plots). This provides the best summary of the process behavior. In scores plot we can observe trends, unusual behaviors (such as outliers), and other information of interest. 8. In addition, the PC loadings plot is generated, demonstrating the variables responsible for data clustering in the scores plot. The PCA also provides values of the so-called loading vectors, showing how the variables are combined to form the scores. A plot of the loading vectors indicates which of the variables are important and correspond to the directions in the scores plot (see Note 7) (8). 3.6. Spectral Data Processing Using Singular Value DecompositionBased Algorithm

The SVD-based program may be utilized for detection and quantification of molecules that are present in low quantities in the obtained spectra. The main assumption of our SVD-based algorithm is that the individual metabolites resonating in the magnetic field have Lorentzian shape. This assumption can be challenged if the magnetic field is not homogeneous during the acquisition.

Systems Biology Approach to Imaging of Neural Stem Cells

431

Thus, the robustness of the SVD-based algorithms has to be tested if such inhomogeneity is highly suspected. Our SVD-based software has been developed in MATLAB and primarily used for detection and quantification of 1.28 ppm peak, associated with NPC, in the spectra acquired in vivo (4). The input to the program is a time domain free induction decay (FID) signal represented by complex data. 1. Remove the water in the raw data using an SVD-based method. Apply a fixed model of order 10 and remove the peaks located in the water region. 2. From the obtained data, estimate the noise variance σ 2 . Perform line broadening by multiplying the signal with an exponentially decaying sequence. 3. Apply fast Fourier transform (FFT) to the data and calibrate the spectrum using the DSS peak (0.00 ppm) for in vitro spectra or N-acetyl aspartate (NAA, 2.02 ppm) peak for in vivo spectra. 4. If necessary, using SVD, remove the nuisance metabolic peaks in the vicinity of the peak of interest. 5. Select the final quantification region based on the location of the peak of interest. Filter the data with the predefined bandwidth which is around 0.5 ppm. Process the filtered data with models of orders K = 2, 3,. . ., Kmax . A signal is considered to be detected with a given model if the SVD frequency estimates of the signal are in a small bandwidth centered at the frequency of the biomarker. Soft constraints are imposed so that the estimated damping factor of the detected signal must be between 0 and 15 Hz. The estimated signal power is the median of the estimated powers from the different models. 6. The SNR of the detected signal is calculated as a function of the estimated signal amplitude a, estimated signal damping factor d, signal length N, and the estimated data noise variance σ 2 and according to SNR = 10 log

1 − e−2dN |a|2 + 10 log , σ2 N (1 − e−2d )

where the SNR is expressed in dB. Based on the SNR value, the reliability of the quantification is determined. For example, for human brain 1 H-MRS data, 80% reliability in signal quantification (such as 1.28 ppm peak enriched in NPC) is obtained when the minimal SNR required is –20 dB. An example of the human MRS of the hippocampal voxel and the SVD-based processing of the spectral range of interest is shown in Fig. 21.2 (see Note 8).

432

Ma et al.

Fig. 21.2. 1 H-MRS of the human brain hippocampus in vivo, obtained using a 3 T MRI scanner (Phillips, Achieva) and 5 min acquisition. a Anatomical scans are used to localize the hippocampal voxel of interest. FFT spectra are shown. b Singular value decomposition (SVD)-based spectral processing of the spectra obtained in (a). Modeling of the individual peaks within the region of interest (1.0–1.5 ppm) outlines the 1.28 ppm metabolite (peak of interest).

4. Notes 1. The same NMR acquisition methodology can be applied for metabolome analysis of other cell types, such as primary cultures of neurons, astrocytes, oligodendrocytes, other stem cells, or any other isolated cell population of interest (4). 2. To ensure the specificity of the findings, all control solutions should be analyzed in parallel. These include media, conditioned media (also useful for footprinting of excreted metabolites), and any compounds or kits used to dissociate and wash the cells. 3. The described method utilizes live cells and not cell extracts. Thus, special care needs to be taken to ensure the homogeneous suspension of cells in the NMR probe. 4. Since the cells are resuspended in PBS and not media, they are deprived from all the nutrients and most likely switch to an anaerobic metabolism, leading to the presence of the lactate doublet (1.33 ppm) observed in most of the in vitro spectra. 5. Metabolomic analysis of the MRS data is challenging and high-quality data are critical for the success of this work. Some of the issues which may be encountered in MRS are (1) poor shimming, in particular of the hippocampal spectra (the irregular structure and adjacent CSF provide a challenge for shimming), (2) suboptimal water suppression, which introduces baseline deviations and affects accurate metabolomic peak area estimates, (3) measurement of short T2 metabolites, which may be important for improving the metabolomic profile, (4) insufficient signal-to-noise ratio (SNR) for detection of low concentration metabolites.

Systems Biology Approach to Imaging of Neural Stem Cells

433

SNR can always be improved with increased scan time or by testing alternate coil options, including highly parallel coils (>64 channels) which are beginning to show marked SNR improvements for brain scanning. 6. All scaling methods should be tested on the data before selecting the best method. We recommend using UV scaling for in vitro data and Pareto scaling for in vivo data. 7. Multivariate statistical analysis indicates metabolomic differences in main neural cell types and points to certain metabolites enriched in NPC which distinguishes them from the rest of the cells found in the healthy brain (1). However, this methodology may be applied to any spectral analysis regardless of the mode of acquisition and thus for both in vitro and in vivo data acquisition. 8. The presented methodologies for metabolomics analysis and quantification of the selected peaks using SVD-based algorithms are always subjected to problems associated with biological variation and closely tied analytical validation. Thus, optimization of the acquisition and processing methods is essential to provide informative and reproducible data.

Acknowledgments We would like to thank Lisa Vingara and Dr. Istvan Pelczer for helpful discussions on metabolomics, Dr. Helene Benveniste for invaluable advice and continued support of our animal brain imaging studies, and Dr. Juan Jose Pena Deudero for critically reading the chapter. This work was supported by the National Institute of Neurological Disorders and Stroke (NINDS: R21NS05875-1), Phillip R Dodge Young Investigator Award (Child Neurology Society); U.S. Army Medical Research Grant (DAMD170110754) and NIH Intellectual and Developmental Disabilities Research Grant (P30HD024064 ) (M.M.-S.). References 1. Maleti´c-Savati´c, M., Vingara, L. K., Manganas, L. N., Li, Y., Zhang, S., Sierra, A., Hazel, R., Smith, D., Wagshul, M. E., Henn, F., Krupp, L., Enikolopov, G., Benveniste, H., Djuri´c, P. M., Pelczer, I. Metabolomics of neural progenitor cells:

A novel approach to biomarker discovery. Cold Spring Harb Symp Quant Biol 2008;73: 389–401. 2. Holmes, E., Wilson, I. D., Nicholson, J. K. Metabolic phenotyping in health and disease. Cell 2008;134:714–717.

434

Ma et al.

3. Weckwerth, W., Morgenthal, K. Metabolomics: From pattern recognition to biological interpretation. Drug Discov Today 2005;10:1551–1558. 4. Manganas, L. N., Zhang, X., Li, Y., Hazel, R. D., Smith, S. D., Wagshul, M. E., Henn, F., Benveniste, H., Djuri´c, P. M., Enikolopov, G., Maleti´c-Savati´c, M. Magnetic resonance spectroscopy identifies neural progenitor cells in the live human brain. Science 2007;318:980–985. 5. Benveniste, H., Blackband, S. MR microscopy and high resolution small

animal MRI: Applications in neuroscience research. Prog Neurobiol 2002;67: 393–420. 6. Jolliffe, I. T. Principal Component Analysis. New York, NY: Springer; 1986. 7. Lindon, J. C., Holmes, E., Nicholson, J. Metabonomics pattern recognition methods and applications in biomedical magnetic resonance. Prog NMR Spec 2001;39:1–40. 8. Ebbels, T. M. D., Cavill, R. Bioinformatic methods in NMR-based metabolic profiling. Prog NMR Spec 2009;55: 361–374.

Chapter 22 MRI of Transplanted Neural Stem Cells Stacey M. Cromer Berman, Piotr Walczak, and Jeff W.M. Bulte Abstract Stem cell-based therapy has the potential to improve the prospect of patients suffering from many untreatable diseases. Applications of stem cells for therapy of neurological conditions, such as Parkinson’s disease, Alzheimer’s disease, multiple sclerosis, or stroke, are particularly appealing as damage of the central nervous system is irreversible and the efficacy of conventional therapy is limited. Despite a broad interest among researchers and clinicians, progress in this field has been slow due to the remarkable complexity of the brain, which makes the task of repairing damaged tissue with stem cells extremely challenging. Making advances can be expedited by novel technologies that can monitor transplanted cells non-invasively. In vivo cellular imaging allows for the repetitive, real-time observation of targeted cells over the course of treatment without the need for invasive biopsies. Magnetic resonance imaging (MRI) is an excellent non-invasive cellular imaging modality as it has superior resolution, is widely used clinically, and has no radiation. This technique currently requires magnetic labeling of cells using superparamagnetic iron oxide particles (SPIOs) and transfection agents. In this chapter, methods for cellular labeling with SPIOs, transplantation of stem cells into the mouse brain, and MR imaging of the cells both in vivo and postmortem will be described. Additional histological and immunohistochemical procedures for analysis of the transplanted cells and the diseased brain environment are also provided. Key words: Magnetic resonance imaging, MR contrast agent, stem cell, superparamagnetic iron oxide, cell transplantation, cell tracking.

1. Introduction The central nervous system (CNS) unarguably has the most complex anatomical and functional organization of all body systems. The endogenous regenerative ability of the CNS is extremely limited, and neurons that are damaged or lost due to aging or disease are typically replaced by glial scarring. Therefore, the prospect of CNS regeneration through transplantation of exogenous cells is M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_22, © Springer Science+Business Media, LLC 2011

435

436

Cromer Berman, Walczak, and Bulte

particularly appealing. Neurological and neurodegenerative diseases afflict many patients every year, reaffirming the need for an effective treatment. Several cell types have been identified as having potential for neural repair including neural progenitor cells (1), embryonic stem cells (2), or induced pluripotent stem cells (3). Further progress in stem cell research may be expedited by the use of real-time, non-invasive imaging of transplanted cells (4). Imaging helps to determine whether the transplanted cells survive, differentiate, or migrate to the targeted lesions to monitor the effectiveness of the stem cell treatment (5). In addition, as (stem) cell-based therapy transitions to clinical trials (4) there is a need for monitoring the status of transplanted cells in order to assure the safety and efficacy of these therapies. Several commercially available technologies have been utilized for in vivo imaging of experimental animal disease models, including magnetic resonance imaging (MRI) (6), X-ray computed tomography (CT) (7), positron emission tomography (PET) (8), single photon emission computed tomography (SPECT) (9), and optical imaging (10). The ideal imaging modality would combine three-dimensional tomographic capabilities, sensitivity, and resolution allowing single cell imaging, low toxicity to stem cells and patients, and utilize a contrast mechanism that produces a non-diluting, cell-specific signal reporting on the status of the stem cells (5). PET and SPECT both have high sensitivity, but unfortunately they subject patients to ionizing radiation, whereas MRI has high resolution and sensitivity without exposing patients to radiation (11). Due to its superior resolution and anatomical information, MR imaging of magnetically labeled cells has therefore been the preferred imaging modality for image-guided transplantation and subsequent tracking of stem cells in the CNS (12). A limitation of this cell labeling approach is the dilution of the contrast agent, such as iron oxide nanoparticles (13) in proliferating stem cells, which can be avoided in the future with the use of MR reporter genes (13, 14). In order to detect the transplanted stem cells and monitor the delivery and migration with MRI, the current technique requires the use of an intracellular contrast agent (15). The most commonly used technique for cell labeling is based on combining SPIOs with a transfection agent facilitating uptake of the iron particles (16). SPIO nanoparticles lead to strong T2 relaxation enhancement, shortening the T2 , and create hypointensities on MR images. The dextran-coated iron oxide nanoparticle agent R (Berlex Laboratories, Wayne, NJ) is FDA approved for Feridex clinical use as a liver contrast agent. The contrast agent is biocompatible and biodegradable through liver biodegradation and processing.

MRI of Transplanted Neural Stem Cells

437

In this chapter, methodologies for neural stem cell tracking by MRI are described. In typical stem cell tracking experiments, the cells of interest are cultured, magnetically labeled in vitro using Feridex (17, 18) and stereotactically transplanted into the brain or infused systemically in a mouse or rat model of neurological disease (6). Animals with transplanted cells are then imaged in vivo using two-dimensional or three-dimensional T2 and T2 ∗ -weighted MRI (19). High-resolution MRI can also be performed postmortem for improved sensitivity and spatial resolution (20). Validation techniques for cellular MR imaging, such as histological and immunohistochemical identification of iron oxide nanoparticles and transplanted cells, are also outlined in this chapter.

2. Materials 2.1. Culture of Stem Cells

1. Primary neural stem cells or cell line (e.g., C17.2 cells). 2. Culture medium composition required for C17.2 cells: ◦ Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Bethesda, MD), ◦ 10% fetal bovine serum (Gibco), ◦ 5% horse serum (Gibco), ◦ 2 mM L-glutamine (Gibco), ◦ 1% penicillin/streptomycin (Sigma-Aldrich, St. Louis, MO, USA), ◦ 1% amphotericin B (Sigma-Aldrich) (optional). 3. Polystyrene tissue culture dishes (Corning, NY, USA).

2.2. Preparation of Iron Oxide-Labeled Cells

R 1. Feridex (11.2 mg Fe/mL, Berlex Laboratories, NJ, USA).

2. Transfection agent, poly-L-lysine (PLL, MW = 350–400 kDa, Sigma-Aldrich). 3. Rotating shaker.

2.3. Histochemical Staining of Magnetically Labeled Cells

1. 4% paraformaldehyde (freshly prepared, Sigma-Aldrich). 2. Nuclear fast red (NFR, Fluka Chemika, Switzerland). 3. Aluminum sulfate (Sigma-Aldrich). 4. Thymol (Sigma-Aldrich). 5. Potassium hexacyano-ferrate III (Sigma-Aldrich). 6. Hydrochloric acid (J.T. Baker, Phillipsburg, NJ, USA). 7. Ethanol (solutions – 70, 80, and 100% ethanol). 8. Histoclear (National Diagnostics, Georgia, USA).

438

Cromer Berman, Walczak, and Bulte

9. TBS SHUR mount (Electron Microscopy Sciences, Hatfield, PA, USA). 10. 3 ,3 -Di-aminobenzidine (Sigma-Aldrich). 2.4. Transplantation of Labeled Cells

1. Hemocytometer. 2. Trypan blue. 3. Mice or rats. 4. Isoflurane (Baxter Healthcare Corporation, Deerfield, IL, USA). 5. Isoflurane chamber induction system and anesthesia machine (Surgivet, Wisconsin, USA). 6. Electric clippers. 7. Stereotactic brain apparatus (Stoelting, Wood Dale, IL, USA). 8. Betadine and hydrogen peroxide disinfectants. 9. Surgical instruments – scalpel blade and handle, forceps, scissors, needle holder, 3.0-gauge silk suture. 10. Electric drill. 11. 10 μL Hamilton glass syringe with 31 G microinjection needle (Hamilton, Reno, NV, USA). 12. Ketofen (2–5 mg/kg). 13. Warming pad.

2.5. MR Imaging of Labeled Cells After Transplantation into the Brain

1. Respiration monitor system for small animals (SA instruments). 2. Eye ointment. 3. MR scanner (such as Bruker Biospec 9.4 T horizontal bore magnet and Bruker 400 MHz vertical magnet). 4. MR coils. 5. Fomblin LC08 (Ausimont, USA) (only for postmortem imaging).

2.6. MRI Data Analysis

1. ImageJ (National Institutes of Health, Bethesda, MD, USA). 2. Amira (Mercury Computer Systems, San Diego, CA, USA) or similar image processing software.

2.7. Histological and Immunohistochemical Analysis

1. 4% PBS buffered paraformaldehyde (freshly prepared, Sigma-Aldrich). 2. Microscope tissue slides (Fisherbrand). 3. Primary and secondary antibodies for immunohistochemistry.

MRI of Transplanted Neural Stem Cells

439

4. Hoechst dye (Invitrogen, Carlsbad, CA, USA). 5. Vectashield mounting medium (Vector, Burlingame, CA, USA). 6. Glass cover slips (Fisherbrand).

3. Methods 3.1. Culture of Stem Cells

Cell culture conditions will depend on the type of stem cell used. The following methods are used for C17.2 neural stem cells, derived from neonatal mouse cerebellum (21). 1. A cryopreserved vial of cells should be thawed quickly in a 37◦ C water bath. The cells should be added to warm, fresh culture medium and centrifuged at 150g for 5 min to remove the DMSO. After the cells are washed, plate the cells in polystyrene tissue culture flasks. 2. Maintain the cells in Dulbecco’s modified Eagle’s medium, supplemented with 10% fetal bovine serum, 5% horse serum, 2 mM L-glutamine, 1% penicillin/streptomycin, and 1% amphotericin B. 3. Change the medium every 3 days. Subculture the cells when the culture reaches 80–90% confluence. 4. Grow cells at 37◦ C and 5% CO2 on 75-cm polystyrene tissue culture dishes until there are enough viable cells for magnetic labeling and transplantation.

3.2. Preparation of Iron Oxide-Labeled Cells

The following protocol uses superparamagnetic iron oxide nanoparticles (Feridex) and poly-L-lysine (PLL) as the transfection agent. Other nanoparticles (22), transfection agents (23), or labeling methods (18) may also be used depending on the cell type. 1. Prepare a 1.5 mg/mL PLL stock solution in sterile water. Store the stock at –20◦ C. R stock 2. Under sterile conditions, add 2.2 μL Feridex (11.2 mg Fe/mL) to fresh culture medium to achieve a final concentration of 25 μg Fe/mL. Mix well.

3. Add 0.25 μL PLL transfection agent stock solution per milliliter of Feridex/medium to a final concentration of 375 ng/mL. Mix well (see Notes 1 and 2). 4. Incubate the Feridex–PLL medium for 1 h at room temperature on a rotating shaker in order to allow the Feridex–PLL complexes to form by electrostatic interactions (see Note 3).

440

Cromer Berman, Walczak, and Bulte

5. When working with adherent cells, culture the cells until they reach about 70–80% confluence. Remove and discard the old medium and wash the culture flask with PBS. For floating cells, spin the cells down at 150g for 5 min and discard the supernatant. 6. Apply the Feridex–PLL containing medium to the cells or resuspend the cell pellet in Feridex–PLL containing medium. 7. Incubate the cells for 24 h at 37◦ C and 5% CO2 to allow uptake of the iron by the cells (see Note 4). 3.3. Histochemical Analysis of Magnetically Labeled Cells

In order to ensure that the cells are properly labeled with iron oxide, it is important to perform histochemical staining of the cells prior to transplantation. Prussian blue staining with Perls’ reagent is the preferred method of staining for iron oxide. 1. Remove the medium from labeled cells and gently wash the cells three times with PBS, avoiding disruption of the monolayer. 2. Fix the cells with 4% paraformaldehyde for 10 min (see Notes 5 and 6). 3. Remove the fixative and wash the cells three times with PBS. Fixed cells can be stored at this stage in 4◦ C for up to 2 weeks. 4. Prepare nuclear fast red solution (NFR) for counter staining by dissolving 0.1 g of NFR in 1,000 mL of deionized water. Add 5 g of aluminum sulfate and heat to a boil slowly with stirring. Cool to room temperature, then filter the NFR solution with a 0.2 μm filter. Add a few grains of thymol as a preservative so the solution can be used for several weeks. 5. Prepare fresh Perls’ reagent by dissolving 1 g of potassium ferrocyanide (potassium hexacyano-ferrate III) in 42 mL of deionized water. Add 8 mL of 37.5% HCl in the fume hood because the reaction releases cyanide. Perls’ reagent contains 2% potassium ferrocyanide and 6% HCl (see Note 7). 6. In the dark (e.g., covered with aluminum foil), incubate the fixed cells with Perls’ reagent for 30 min. 7. Wash the cells with distilled water. 8. (OPTIONAL) Diaminobenzidine (DAB) enhanced Prussian blue staining can also be performed to further enhance and improve the sensitivity of iron staining (see Note 8). 9. Incubate the cells with NFR solution for 10–15 min for desired intensity of nuclear counterstaining. 10. Wash the cells three times with PBS.

MRI of Transplanted Neural Stem Cells

441

11. Cells stained in polystyrene culture vessels need to be viewed under the microscope immediately. Microscope glass slides can be dehydrated for permanent preservation of the staining. Dehydrate the slides for 1 min each with 70% ethanol, 80% ethanol, and 100% ethanol. Put the slides in Histoclear solution for 5 min and then cover with TBS SHUR mount and a glass cover slip. 12. After the slides have dried, image with a bright field microscope. 13. Quantification of the amount of iron per cell can be performed using a ferrozine assay (16). An example of neural stem cells labeled with Feridex–PLL and stained for Prussian blue is shown in Fig. 22.1.

R Fig. 22.1. Prussian blue staining of C17.2 neural stem cells labeled with Feridex and PLL without (a) and with (b) DAB-enhanced Prussian blue staining demonstrating intracellular labeling with iron oxide particles.

3.4. Transplantation of Labeled Cells into the Brain Parenchyma

The method of choice for targeted intracerebral cell injection is using a stereotaxic device and a Hamilton syringe/needle. There are many animal models of neurological disease including stroke, experimental autoimmune encephalitis, and Alzheimer’s disease. Most of these models are fully immunocompetent and transplanted stem cells, particularly human cells, are subject to immunorejection. Except for the unique and rare situation when the donor and the recipient are immunologically matched (syngenic), immunosuppression should be considered to prevent the rejection of the transplanted stem cells. 1. Prepare the labeled stem cells for transplantation by removing the Feridex–PLL labeling medium and washing with PBS. Harvest the cells from the flask with mild trypsinization (0.05% trypsin/EDTA, 5 min). Centrifuge the cells for 5 min at 150g. Count the cells using a hemocytometer

442

Cromer Berman, Walczak, and Bulte

using trypan blue staining to determine the number and viability of Feridex-labeled cells. Suspend the cells at the desired concentration in sterile PBS or Hanks Balanced Salt Solution and keep the cells on ice until transplantation. 2. Anesthetize the mouse with 1–2% isoflurane inhalation. Check the toe-pinch reflex to determine if the mouse is properly anesthetized. 3. Shave the head using electrical clippers. 4. Place the mouse into a stereotaxic device. Stabilize the head using the ear and nose/tooth bars (see Note 9). 5. Prior to beginning the surgery, apply betadine disinfectant solution using sterile cotton swabs to sanitize the surgical area. 6. Using a scalpel, incise the skin in the midline to expose the skull. 7. Clean the area with diluted hydrogen peroxide and locate bregma (the crossing of sagittal and coronal sutures). Confirm the correct coordinates according to a brain atlas (e.g., (24)) for injecting the cells into the desired location (e.g., right corpus callosum AP = 0.0 mm; ML = +2.0 mm; DV = 1.5 mm). 8. Drill a small burr hole in the scull above the injection site using a micro drill and 0.7 mm burr. 9. Load the cells into a 10 μL Hamilton microinjection syringe with an attached 31 G needle. Touch the needle to the surface of the brain, measure 0.0 reference point for DV coordinate, and lower the needle into the brain (–1.5 mm for corpus callosum targeting) (see Note 10). 10. Inject cells manually or using a motorized injector at a rate of 1 μL/min (see Note 11). 11. After transplantation, let the needle remain for 1 min to prevent backflow and then slowly retract the needle (see Note 12). 12. Close the incision tightly using 3.0 vicryl suture. 13. Follow appropriate animal care and use guideline providing appropriate postsurgical care and analgesia. 14. Place the mouse on a warming pad and when fully recovered from anesthesia, return it to its cage. 3.5. MR Imaging of Labeled Cells After Transplantation into the Brain

After transplantation, the localization and migration of magnetically labeled cells can be visualized and tracked using magnetic resonance imaging. High-field, dedicated animal MR scanners, such as a Bruker Biospec 9.4 T horizontal bore spectrometer,

MRI of Transplanted Neural Stem Cells

443

are best suited for cellular imaging and providing high sensitivity. The following procedure describes the method for obtaining MR images of SPIO-labeled cells in the brain. 1. Anesthetize the mouse with 1–2% isoflurane inhalation and 0.5 L/min O2 flow in an induction chamber (see Note 13). 2. Immobilize and secure the mouse in a manufacturer provided or custom-made cradle, equipped with a surface or volume coil that is connected to a continuous flow of 1–2% isoflurane. Make sure that the mouse is positioned properly and motion of the head is minimized. 3. Apply eye ointment to prevent the eyes from drying. 4. Connect a respiration monitor to detect the breathing rate. Throughout the imaging session, the breathing rate should be maintained between 50 and 70 breaths per minute, which can be maintained by adjusting the level of isoflurane anesthesia. 5. Position the mouse in the coil so that the mouse head is located at the center of the coil. Make sure that the mouse and the coil are straight and place them in the magnet. 6. In the MR software, create a new patient and select that the mouse is positioned head first in the prone position. 7. After the probe is tuned and matched, use the position scan to look at the position of the mouse and ensure that the head is positioned in the coil. 8. For proper orientation of MRI scans, acquire a scout scan using a tripilot sequence with the following parameters: rapid acquisition with relaxation enhancement (RARE) pulse sequence, echo time (TE) = 12.5 ms, repetition time (TR) = 2,000 ms (number of averages) NAV = 1, field of view (FOV) = 4 × 4 cm, matrix = 128 × 128, RARE factor = 8, total scan time = 32 s. 9. If needed, adjust the geometry on the RARE tripilot sequence so that the three main axes of the brain (sagittal, coronal, and axial) are clearly visible on the scout scans. Geometry coordinates for future scans are based on this tripilot scan (see Notes 14 and 15). 10. For T2 ∗ -weighted gradient echo scans, select a sequence (e.g., multi-gradient echo, MGE), adjust the geometry, and scan with the following parameters: TE = 5 ms, TR = 500 ms, NAV = 4, FOV = 1.70 × 1.70 cm (modify if needed to cover the entire sample and avoid folding artifacts), matrix = 256 × 256, if available use respiratory

444

Cromer Berman, Walczak, and Bulte

gating for reduced motion artifacts, total scan time = 4 min 16 s (scan time may be longer if gating is used). 11. For T2 -weighted scans, select a spin echo sequence, adjust the geometry, and scan with the following parameters: TE = 12 ms, TR = 2,000 ms, NAV = 4, FOV = 2.0 × 2.0 cm, matrix = 256 × 256, RARE factor = 2, if available use respiratory gating for reduced motion artifacts, total scan time = 8 min 32 s. 12. After completion of the MR imaging, remove the mouse from the chamber, place the mouse on a warming pad until it has recovered from anesthesia and then return it to its cage. 13. Additionally, postmortem MRI can be performed for improved resolution and sensitivity. Following transcardial perfusion with PBS and 4% paraformaldehyde (PFA) fixation. Decapitate the mouse and submerge the head in PBS or Fomblin in a polystyrene or glass tube. Place the tube in an MR coil. Acquire a 3D gradient-echo image with the following parameters: TE = 5 ms, TR = 120 ms, NAV = 4, matrix = 368 × 256 × 256, flip angle = 30◦ , total scan time = 2 h 11 min 320 s. An example of MR images and the hypointensity from SPIOlabeled cells is shown in Fig. 22.2. 3.6. MRI Data Analysis

1. Refer to a mouse brain atlas (e.g., (24)) to identify anatomical brain structures visible on MR images. 2. With NIH ImageJ software or a similar program, analyze the MR images for cellular imaging, e.g., calculate hypointense pixels, the surface area, or the volume occupied by hypointense labeled cells. Example 1: Labeled stem cells are transplanted into the corpus callosum of mice. MRI is performed weekly for several months after transplantation to monitor SPIO hypointensities over time. The number of black pixels is quantified for the SPIO hypointensities and compared over time. A Z-projection of minimum intensity is made for each MR image using Image J. A region of interest is drawn around the control hemisphere, contralateral to the injection site, and a baseline pixel intensity histogram is created to establish the minimum signal intensity in the absence of iron. A second region on interest is drawn around the ipsilateral hemisphere with the hypointensity, and a pixel intensity histogram is created for this region of interest. The number of pixels in the ipsilateral hemisphere below the minimum signal intensity is calculated. This represents the black pixel count for the SPIO hypointensity.

MRI of Transplanted Neural Stem Cells

445

R Fig. 22.2. Feridex -labeled cells proliferate normally in vivo and can be detected by MR imaging 4 days after intrastriatal transplantation. a T2 ∗ -weighted gradient echo image shows a prominent hypointense signal around the area of labeled injection in the left hemisphere. Unlabeled cells injected in the contralateral striatum (right side) do not exhibit contrast. b Corresponding histology immunostained for beta-galactosidase (red) and myelin basic protein (green). Transplanted cells (red) are visible in both striata with a good anatomical correlation of labeled cells and MR hypointense signal. Magnification of the brain region with grafted MEP-labeled (c) and unlabeled (d) cells shows that both the migration distance and the cell morphology are indistinguishable between labeled and unlabeled cells. Reproduced, with permission, from Ref. (18).

Example 2: Labeled stem cells are transplanted into the ventricular system, which allows for migration of stem cells throughout the entire neuroaxis. MR images are taken over time to monitor the cellular migration patterns. The distance that the stem cells migrate is calculated from sequential MR images. The maximum distance of the migration of transplanted cells is determined from the MR slices as the distance from the proximal site of the hypointensity at the edge of the ventricle to the leading edge of migrating cells in the corpus callosum. The downward migration is determined from coronal images and is calculated by the number of axial or sagittal slices spanning the hypointense signal. An example is shown in Fig. 22.3. 3. Reconstruct and co-register postmortem 3D MR images using Amira software and the label voxel module.

446

Cromer Berman, Walczak, and Bulte

Fig. 22.3. Examples of in vivo and ex vivo MR images of ferumoxide-labeled neural progenitor cells. Labeled NPCs were transplanted to the right ventricle of EAE mice (black arrow). At day 1 after ICV transplantation (a), cells indicated by hypointense (black) MRI signal are found exclusively within the cerebral ventricles and are absent within the corpus callosum (white arrow). At 4 (b) and 7 (c) days after ICV transplantation, some cells had migrated into the corpus callosum (white arrow). Ex vivo MRI at day 22 posttransplantation confirmed this pattern of migration (d). Reproduced, with permission, from Ref. (19).

3.7. Histological and Immunohistochemistry Analysis

1. After completion of MR imaging experiment, sacrifice and transcardially perfuse mice with 4% PFA. Postfix the brain tissue in 4% PFA overnight, cryopreserve the brains in 30% sucrose, freeze on dry ice, and cryosection the brains at 20–30 μm slices, mounting the slices on charged microscope slides for histology and immunohistochemistry. 2. Stain brain slices for Prussian blue following the procedure outlined in Section 3.3 to detect intracellular iron oxide deposits. Image using bright field microscopy (see Note 16). 3. For immunohistochemistry, stain brain slices for primary antibodies of interest to detect the transplanted stem cells or determine their phenotypic features. Block tissue samples for 1 h at room temperature with PBS containing 10% normal goat serum and 0.3% Triton. Incubate the slides overnight at 4◦ C with primary antibody diluted with PBS, 2% normal goat serum, and 0.3% Triton. The next day, incubate the slides with the secondary antibody for 2 h at room temperature with PBS and 0.3% Triton. Stain the nuclei with 1 μg/mL Hoechst dye and embed with Vectashield mounting medium to preserve the fluorescence signal. Image on a fluorescent microscope.

MRI of Transplanted Neural Stem Cells

447

4. Notes 1. PLL is just one example of the transfection agents that can be used for labeling with SPIOs. Other transfection agents, such as protamine sulfate (23) (American Pharmaceuticals Partner), Lipofect (Gibco), Superfect (Qiagen), and FuGENE (Roche), can also be applied. 2. The amount of transfection agent used should be optimized for each cell type. The concentration of 375 ng/mL PLL has been found to label neural stem cells with SPIO efficiently without affecting cell proliferation and differentiation, but this should be determined for each cell type. R with the medium well by pipetting before 3. Mix Feridex adding PLL. If not mixed properly, precipitation can occur R –PLL complexes. due to aggregation of Feridex

4. Cells are usually magnetically labeled with SPIOs for 24 h in normal culture conditions, but some cell lines can benefit from up to 48 h of incubation. An alternative cell labeling method is based on electroporation as described by Walczak et al. (18). 5. Prepare fresh 4% PFA for fixing cells or tissue and perfusing animals. Storing PFA solution may lead to crystallization, which results in improper fixation and suboptimal histological stains. 6. Prepare 4% PFA in a chemical fume hood as PFA is toxic. 7. Use proper caution when preparing Perl’s reagent as the reaction releases cyanide and should be performed in a fume hood. 8. For DAB-enhanced Prussian blue staining, prepare an inactivated solution of 7 mg of DAB with 50 mL PBS (140 μg/mL DAB) and add to the cells for 15 min, in the dark and with shaking. Without washing, add 50 μL of H2 O2 (0.025% H2 O2 ) to make an activated DAB solution for 15 min, in the dark with shaking. Then wash the cells with distilled water and proceed with NFR staining (step 9). 9. Proper placement and immobilization of the mice in the stereotaxic device is crucial to the success of the surgery and implantation of the cells precisely into the desired location. 10. Prior to each loading of cells into the Hamilton syringe, mix cells well with a pipet to avoid clumping. 11. Inject cells into the brain slowly to avoid any injury to the tissue.

448

Cromer Berman, Walczak, and Bulte

12. When the Hamilton needle is removed from the brain, carefully watch for any backflow of injection fluid that would result in a loss of cells. 13. The amount of isoflurane for induction and throughout the MR imaging may need to be adjusted based on the strain of mice as some mice strains react differently to the anesthesia. 14. The entire brain should be displayed within the MR image without wrapping artifacts. If the entire brain is not visualized, check the position of the animal and the FOV settings. 15. If the MR images are blurry, check that the animal is hold secure and that the anesthesia is still working properly with the breathing rate in the desired range. 16. When dehydrating thicker samples, e.g., cryosectioned tissue slides, dehydrate in each ethanol solution for 5 min instead of 1 min. References 1. Wang, L., Martin, D. R., Baker, H. J. et al. Neural progenitor cell transplantation and imaging in a large animal model. Neurosci Res 2007;59:327–340. 2. Ben-Hur, T., Idelson, M., Khaner, H. et al. Transplantation of human embryonic stem cell-derived neural progenitors improves behavioral deficit in parkinsonian rats. Stem Cells 2004;22:1246–1255. 3. Yu, J., Vodyanik, M. A., Smuga-Otto, K. et al. Induced pluripotent stem cell lines derived from human somatic cells. Science 2007;318:1917–1920. 4. Bulte, J. W. In vivo MRI cell tracking: Clinical studies. AJR Am J Roentgenol 2009;193:314–325. 5. Walczak, P., Bulte, J. W. The role of noninvasive cellular imaging in developing cellbased therapies for neurodegenerative disorders. Neurodegener Dis 2007;4:306–313. 6. Ben-Hur, T., van Heeswijk, R. B., Einstein, O. et al. Serial in vivo MR tracking of magnetically labeled neural spheres transplanted in chronic EAE mice. Magn Reson Med 2007;57:164–171. 7. Connor, D. M., Benveniste, H., Dilmanian, F. A., Kritzer, M. F., Miller, L. M., Zhong, Z. Computed tomography of amyloid plaques in a mouse model of Alzheimer’s disease using diffraction enhanced imaging. Neuroimage 2009;46:908–914. 8. Adonai, N., Nguyen, K. N., Walsh, J. et al. Ex vivo cell labeling with 64cu-pyruvaldehydebis(N4-methylthiosemicarbazone) for imag-

9.

10.

11.

12. 13.

14.

15. 16.

ing cell trafficking in mice with positronemission tomography. Proc Natl Acad Sci USA 2002;99:3030–3035. Acton, P. D., Zhou, R. Imaging reporter genes for cell tracking with PET and SPECT. Q J Nucl Med Mol Imaging 2005;49: 349–360. Chaudhari, A. J., Darvas, F., Bading, J. R. et al. Hyperspectral and multispectral bioluminescence optical tomography for small animal imaging. Phys Med Biol 2005;50: 5421–5441. Judenhofer, M. S., Wehrl, H. F., Newport, D. F. et al. Simultaneous PET-MRI: A new approach for functional and morphological imaging. Nat Med 2008;14:459–465. Modo, M., Hoehn, M., Bulte, J. W. Cellular MR imaging. Mol Imaging 2005;4:143–164. Walczak, P., Kedziorek, D. A., Gilad, A. A., Barnett, B. P., Bulte, J. W. Applicability and limitations of MR tracking of neural stem cells with asymmetric cell division and rapid turnover: The case of the shiverer dysmyelinated mouse brain. Magn Reson Med 2007;58:261–269. Gilad, A. A., McMahon, M. T., Walczak, P. et al. Artificial reporter gene providing MRI contrast based on proton exchange. Nat Biotechnol 2007;25:217–219. Bulte, J. W., Kraitchman, D. L. Iron oxide MR contrast agents for molecular and cellular imaging. Nmr Biomed 2004;17:484–499. Frank, J. A., Miller, B. R., Arbab, A. S. et al. Clinically applicable labeling of mammalian

MRI of Transplanted Neural Stem Cells

17.

18.

19.

20.

and stem cells by combining superparamagnetic iron oxides and transfection agents. Radiology 2003;228:480–487. Bulte, J. W., Arbab, A. S., Douglas, T., Frank, J. A. Preparation of magnetically labeled cells for cell tracking by magnetic resonance imaging. Methods Enzymol 2004;386: 275–299. Walczak, P., Kedziorek, D. A., Gilad, A. A., Lin, S., Bulte, J. W. Instant MR labeling of stem cells using magnetoelectroporation. Magn Reson Med 2005;54:769–774. Cohen, M. E., Muja, N., Fainstein, N., Bulte, J. W., Ben-Hur, T. Conserved fate and function of ferumoxides-labeled neural precursor cells in vitro and in vivo. J Neurosci Res 2010;88:936–944. Magnitsky, S., Watson, D. J., Walton, R. M. et al. In vivo and ex vivo MRI detection of localized and disseminated neural stem

21.

22.

23.

24.

449

cell grafts in the mouse brain. Neuroimage 2005;26:744–754. Snyder, E. Y., Deitcher, D. L., Walsh, C., Arnold-Aldea, S., Hartwieg, E. A., Cepko, C. L. Multipotent neural cell lines can engraft and participate in development of mouse cerebellum. Cell 1992;68:33–51. Alric, C., Taleb, J., Le Duc, G. et al. Gadolinium chelate coated gold nanoparticles as contrast agents for both X-ray computed tomography and magnetic resonance imaging. J Am Chem Soc 2008;130:5908–5915. Arbab, A. S., Yocum, G. T., Kalish, H. et al. Efficient magnetic cell labeling with protamine sulfate complexed to ferumoxides for cellular MRI. Blood 2004;104: 1217–1223. Paxinos, G., Franklin, K. B. J. The Mouse Brain – In Stereotactic Coordinates. San Diego, CA: Academic Press; 2001.

wwwwwww

Chapter 23 MRI of Experimental Gliomas Frits Thorsen Abstract Malignant gliomas are the most frequent primary brain tumours and they are associated with a grim prognosis. In order to elucidate the biological properties of these tumours and to assess treatment responses, valid animal models are needed. We have developed a model where human glioma specimens are operated into the brains of immunodeficient animals. Tumour development is followed by MR imaging and proton spectroscopy. In this chapter, operating procedures and the MR techniques are presented. Key words: Glioblastoma multiforme, animal models, tumour invasion, vascularisation, tumour metabolism, patient biopsies, implantation, proton spectroscopy.

1. Introduction 1.1. Common Features of Glioblastoma Multiforme (GBM)

Numerically, malignant gliomas (the tumours derived from glial cells) are the most frequent primary brain tumours. The most malignant glioma, glioblastoma multiforme (GBM) (grade IV according to the WHO classification), accounts for approximately 50–60% of all astrocytic tumours and 12–15% of all intracranial neoplasms. The incidence is 3–6 new cases per 100,000 individuals per year (1–3). Expected postoperative patient survival is less than 1 year (4). Histologically, the GBM is composed of poorly differentiated neoplastic astrocytes with areas showing endothelial hyperplasia and necrosis. The tumours display cellular and nuclear polymorphism and a high degree of anaplasia. Diffuse and infiltrative tumour cell invasion into the surrounding brain tissue is a hallmark of GBMs. The tumour cells rapidly invade normal brain

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_23, © Springer Science+Business Media, LLC 2011

451

452

Thorsen

structures and migrate along fibre tracts, often through the corpus callosum and into the contralateral hemisphere. The central part of the tumour is often necrotic, with viable tumour cells restricted to a narrow, peripheral zone. Vascular endothelial cell proliferation is present, often with mitotic activity. As a consequence, intratumoural blood vessels are increased both in number and in size (1). 1.2. Tumour Models Reflecting Glioblastoma Progression

In order to better understand brain tumour biology and to develop new treatment strategies, several experimental animal models of glial tumours have been developed. Recently our group has developed human brain tumour models in immunodeficient rats to study separately two distinct biological phenotypes that characterise GBM growth and progression (5). One tumour type is exhibiting a highly infiltrative non-angiogenic phenotype that displays numerous stem cell markers (low-generation tumour, LG). The other tumour phenotype, which eventually evolves from LG tumours, shows a typical GBM appearance characterised by rapid cell proliferation, angiogenesis, and necrosis (high-generation tumour, HG). LG tumours are generated directly from GBM biopsy spheroids and are essentially slow growing, highly invasive tumours which show very little neovascularisation, as indicated by the absence of contrast enhancement on T1-weighted (T1w) MR images. In contrast, HG tumours, which are initiated by serial passaging in animals, show rapid growth, increased vascularity, and contrast enhancement on T1w MR images, indicating extensive vascular leakage and angiogenesis. Thus, these two distinct GBM phenotypes that are derived from the same tumour sample occur by a selection process in vitro and in vivo, where most cells in the biopsy specimens used to generate LG tumours succumb after implantation in the rat brain. The cells that survive most likely have stem-like properties and are able to adapt to the new microenvironment, where they divide and produce new tumour clones which show rapid growth and angiogenesis. Therefore, these two tumour models allow biological studies on important biological entities characterising GBMs, infiltrative growth, extensive cell proliferation, angiogenesis, and necrosis.

1.3. Magnetic Resonance Imaging (MRI) of Intracranial Neoplasms

Since the introduction of the first magnetic resonance imaging (MRI) system into the clinic, it has become the most accepted imaging modality for the visualisation of brain tumours, due to its excellent soft tissue contrast (6). MRI now plays a crucial role in diagnosis, therapy planning, evaluating therapeutic effects and detecting early recurrence of the intracranial neoplasms. High signal on T2-weighted (T2w) MR images reflects fluid regions within the brain parenchyma and is commonly

MRI of Experimental Gliomas

453

attributed to vasogenic oedema associated with tumour infiltration. Bright lesions on T1w MR images appear after intravenous administration of a contrast agent, such as gadolinium-DTPA, as the blood–brain barrier has been compromised in this region and the contrast agent has leaked into the interstitium. The appearance of GBMs clinically is typically characterised as an irregular enhancing lesion with central non-enhancing areas and a marked mass effect secondary to vasogenic oedema. In preclinical animal models, longitudinal MR studies are commonly performed to assess tumour development and treatment efficacy (7–9). Although scanning protocols on clinical MR machines can be adapted to animal research (10) dedicated MR scanners are more and more used, allowing for higher magnetic field strengths, thus improving signal to noise ratio (SNR) and spatial resolution. To assess tumour progression in our animal models, we have been using a 7 T MR scanner for animal monitoring in several preclinical studies (11–13). 1.4. Proton Magnetic Resonance Spectroscopy

Conventional MRI produces high-resolution images of the brain, allowing direct visualisation of tumour location and size. Usually, such tumours are investigated by T1w MRI (with and without contrast injections) and T2w MR imaging. However, even with the precise morphological information obtained by T1w and T2w MR imaging, conventional MRI alone cannot distinguish between tumours with different prognosis or to classify brain tumours. Proton magnetic resonance spectroscopy (1 H-MRS) is a noninvasive imaging technique for measuring the biochemical content of living tissue. 1 H-MRS is becoming a complement to MRI for the initial diagnosis of brain pathologies, since it provides useful chemical information about determining metabolites which frequently are changed in tumour tissue. 1 H-MRS allows a noninvasive study of several metabolic substances in tumours and in healthy tissue (14, 15) by measuring absolute metabolic concentrations (16). Proton MR spectra of the brain are characterised by the presence of four major metabolites, N-acetylaspartate (NAA), creatine and phosphocreatine (Cr), choline-containing compounds (Cho) and myo-inositol (mI). Other metabolites, such as glutamine, glutamate, glucose, lactate, taurine and others, can also be observed and often quantified, although their absolute concentrations in general are lower (17). Previous work done in our group showed that GBM tumour phenotypes have their own distinct metabolic profiles (12). The HG tumours had elevated concentrations of choline and myo-inositol and decreased concentrations of glutamate and N-acetylaspartate. In the LG tumours, similar changes in metabolic concentrations were observed, although the alterations

454

Thorsen

were more pronounced. The LG tumours also had higher concentrations of choline, taurine and lactate. The results show that metabolic profiles produced by 1 H-MRS can be used to distinguish between two distinct glioma phenotypes. The elevated lactate levels observed in the LG tumours indicate a more pronounced anaerobic metabolism, suggesting that the infiltrative cancer stem-like phenotype utilises glycolysis during growth.

2. Materials This section describes in detail the equipment necessary to culture human glioblastoma specimens, to perform intracranial implantations of the tumour material into the rat brain, and necessary equipment for MR scanning of these tumours. 2.1. Preparation of Tumour Material

1. Dulbecco’s modified Eagles medium (DMEM) (Gibco BRL, Paisley, UK), supplemented with penicillin (100 IU/ml), streptomycin (100 μg/ml), 2% L-glutamine, 10% heat-inactivated newborn-calf serum and four times the prescribed concentration of non-essential amino acids (hereafter called complete medium) (all biochemicals from Bio-Whittaker, Verviers, Belgium). 2. Trypsin EDTA (Lonza, Verviers, Belgium). 3. 0.75% agar (Difco, Detroit, MI, USA). 4. Petri dishes (10 cm diameter; Nunc, Roskilde, Denmark). 5. Surgical blades (Swann Morton Ltd, Sheffield, England). 6. Medium sized (80 cm (2)) culture flasks (Nunc). 7. Fungicin 10 mg/ml (InvitroGen, San Diego, CA, USA).

2.2. Rats and Intracranial Implantations

1. Immunodeficient rats (Han:rnu/rnu Rowett; Harlan Laboratories Inc., Indianapolis, IN, USA). 2. Stereotactic frame (model 900; David Kopf Instruments, Tujunga, CA, USA). 3. Operation microscope (Zeiss, Germany). 4. Dental drill (Microspeed 317 IN; Silfradent, Forli, Italy). 5. Hamilton syringe model no. 7125 or similar (Hamilton Company, Reno, NE). 6. Two bulldog clamps (Agnthos AB, Lidingö, Sweden). 7. One scissors (Agnthos AB). 8. Two forceps (Agnthos AB).

MRI of Experimental Gliomas

455

9. One haemostatic forceps (Agnthos AB). 10. One scalpel (Agnthos AB) with surgical blade (Swann Morton Ltd.). 11. One microknife (Agnthos AB). 12. Steri-drape adhesive towel drape (3 M HealthCare, Neuss, Germany). 13. DPBS (Sigma-Aldrich, St. Louis, MO) with 2% (Merck, Darmstadt, Germany).

D -Glucose

14. Isofluorane (Schering-Plough, Lysaker, Norway). 15. Dormicum (Roche, Basel, Switzerland). 16. Fentanyl (fentanylcitrate 0.0785 mg; Hameln Pharmaceuticals, Hameln, Germany). 17. Domitor Vet 1 mg/ml (Orion Pharma, Oslo, Norway). 18. Marcain 2.5 mg/ml (AstraZeneca, Oslo, Norway). 19. Ethicon prolene 4–0 suture (Johnson & Johnson, Langhorne, PA). 20. Surgical sponges (Neurotechniques Ltd, Oxon, United Kingdom). 2.3. MR Studies

1. 0.9% NaCl solution, 100 mL bottles (Fresenius Kabi, Halden, Norway). 2. Heparin 5,000 IE/a.e./ml (Leo Pharma, Denmark). 3. Sterile water (B. Braun Medical, Vestskogen, Norway). 4. Isoba (Schering Plough). 5. Omniscan T1 contrast agent (0.5 mmol/mL; Nycomed Amersham, Oslo, Norway). 6. Disposable needles (25 G; B. Braun Medical). 7. Two gas anaesthesia units (MatrX, Midmark, Orchard Park, NY, USA). 8. One water heating unit, including a heating pad (Bruker Biospin, Ettlingen, Germany). 9. Monitoring unit for temperature and breathing cycle, including thermometer and breathing patch (Respiration module; SA Instruments, Stony Brook, NY, USA). 10. Heating pad 40×30 cm, to be used during intravenous placement of catheter (Sønnico, Oslo, Norway). 11. Intravenous catheters with injection port (size 24 G; Terumo Europe N.V., Leuven, Belgium). 12. Polyethylene tubing, 0.58 mm inner diameter (Becton Dickinson and Company, Sparks, MD, USA).

456

Thorsen

13. Home-made injection device for intravenous injections of contrast (see Section 3.1.1). 14. Surgical tape, width 12.5 and 25 mm (3 M, Nadarzyn, Poland).

3. Methods This section describes in detail the procedures for collecting and culturing human tumour material, the implantation procedure, the MR procedures for anatomical imaging of the tumours and proton spectroscopy of tumour metabolites. 3.1. Preparation of Tumour Material

1. Obtain glioblastoma material either directly from patients during surgery or from the animal brains, after the primary tumour material has been serially passaged in the rat brain. 2. Transfer the tumour pieces as quickly as possible into a sterile 50 mL tube containing 15 mL complete medium. 3. Bring the tumour samples to the cell culture lab. Work in sterile conditions from now on. 4. Put the tumour pieces into a sterile petri dish. Remove as much complete medium from the petri dish as possible. 5. Cut the tumour into small fragments using two surgical blades. For a tumour piece of 5–10 mm in diameter, this takes approximately 15 min. 6. Transfer the tumour fragments into a medium sized agarcoated tissue flask and add 25 mL complete medium. 7. Add 50 μg/ml fungicin directly into the medium to avoid contamination by fungus. 8. Keep the tumour material in a tissue culture incubator with 5% CO2 in air and relative humidity 100% at 37◦ C. 9. Change the growth medium at least once a week. You do not need to add any fungicin in the complete medium.

3.2. Rats and Intracranial Implantations 3.2.1. Before the Implantation

1. All animal procedures must be carried out according to approval from National Animal Research Authorities. 2. Go to the cell culture lab and pick up 15 tumour spheroids with a pipette. Put them in a 10 mL sterile tube containing 5 mL PBS with 2% D-glucose (Fig. 23.1c, d). Keep the spheroids on ice until use.

MRI of Experimental Gliomas

457

Fig. 23.1. Intracranial implantation of tumour biopsies. a Stereotactic frame and instruments for implantation. b Injection of anaesthesia. c Positioning of the animal in the stereotactic frame. d Subcutaneous injection of local anaesthesia. e Incision of the skin over the scalp. f Cleaning of the scalp. g Picking the tumour spheroids. h Positioning of the Hamilton syringe over the bregma point. i Drilling the implantation hole through the bone. j Cutting the dura with the microknife. k Lowering the Hamilton syringe down into the brain tissue. l Suturing the skin after implantation.

3. Go to the implantation lab and prepare your operating table (Fig. 23.1a). 4. Cover the operating table with sterile towel drapes. 5. Place the stereotactic frame, the operating equipment and your tubes with tumour material on the table (Fig. 23.1a). 6. Prepare the anaesthesia: mix 10 mL fentanyl with 1 mL domitor. Fill the gas anaesthesia equipment with isofluorane and put the bottle with marcain (local anaesthesia) on the operating table. 7. Localise your animals and bring them to the operating room. 3.2.2. The Implantation Procedure

1. Weigh the animals. Calculate how much fentanyl/domitor mixture you need to inject per animal. 0.5 mL fentanyl/domitor per 100 g rat is a commonly recommended dose, but this has to be tested for each individual rat strain.

458

Thorsen

Start with 0.3 mL/100 g, and if the animal does not sleep deeply after 10 min, inject an additional amount of 0.1 mL. Repeat until the rat is fully asleep. 2. Withdraw the correct amount of fentanyl/domitor in a 1.0 mL syringe. 3. Anaesthetise the animal. We usually anaesthetise the animal first using inhalation anaesthesia by putting the rat in a chamber connected to an anaesthesia unit delivering 4.0% isofluorane mixed with 50% O2 and 50% N2 O. 4. Inject the correct amount of fentanyl/domitor (according to weight) and wait 10 min until the rat is at sleep (Fig. 23.1b). 5. Sterilise your surgical instruments while you wait for the rat to be fully anaesthetised. 6. Position the rat in the stereotactic frame (Fig. 23.1c). 7. Inject 0.2 mL local anaesthesia (Marcain) subcutaneously in the middle of the head (Fig. 23.1d). 8. Make a 1.5–2.0 cm incision with the scalpel (Fig. 23.1e). 9. Attach the two bulldog clamps to the skin, one on each side of the incision wound. Clean the scalp with two neurosurgical sponges (Fig. 23.1f). 10. Pick up the tumour spheroids with the Hamilton syringe (Fig. 23.1g). 11. Attach the Hamilton syringe into its holder in the stereotactic frame. Localise the bregma point on top of the scalp, using the operating microscope. The bregma point is the interconnection between the coronal sutures and the sagittal suture, seen on top of the scalp on the animal. Position the tip of the syringe just over this point (Fig. 23.1h). Then move the syringe tip 1 mm posterior to the coronal suture and 3 mm right of the sagittal suture. 12. Lift the Hamilton syringe 5 mm and make a hole in the bone, using the dental drill with a 3 mm diameter (Fig. 23.1i). It is of importance to use the microscope when making the burr hole and take care not to penetrate the brain tissue when the burr hole is finished. Clean the burr hole with a neurosurgical sponge and remove any loose bone pieces. Sometimes there will be a light bleeding from the brain tissue, then a light pressure should be applied with one of the sponges until the bleeding stops (usually after 30–60 s). 13. Make two cross cuts in the dura, using the microknife (Fig. 23.1j).

MRI of Experimental Gliomas

459

14. Position the syringe into the middle of the burr hole and lower the syringe so that the tip touches the surface of the brain (Fig. 23.1k). 15. Move the Hamilton syringe 3 mm down into the brain tissue and then 0.5 mm up again. This creates a small pocket inside the brain tissue. Inject the tumour spheroids into this pocket over a time period of at least 1 min. 16. Slowly retract the syringe during 3–5 min. You should look into the operating microscope all the time, and ensure that there is no backflow of fluid containing spheroids. 17. Suture the skin with 4–5 stitches, 2–3 mm between each stitch (Fig. 23.1l), using the haemostatic forceps and a forceps. 18. Remove the animal from the stereotactic frame. 3.2.3. After the Implantation

1. Put the animals into separate cages, otherwise the rats will start biting off the stitches on each other when they wake up. 2. Bring the rats back to the animal room. 3. Clean up the implantation room. 4. Monitor the animals until they start waking up, then they should be looked after at least two times a day during the first 2 days after implantation.

3.3. MR Scanning of Experimental Gliomas

These instructions assume the use of a Bruker Pharmascan 7 T small animal MR machine, with ParaVision 5.0 software version. It should, however, be fairly easy to adapt the scanning protocols to other Paravision software versions, as well as to other Bruker machines, and MR scanners from other vendors. Furthermore, these instructions assume that there are two gas anaesthesia units available: one next to the MR scanner and the other one in the animal preparation area. It is also assumed that a software-controlled breathing cycle and temperature monitoring unit is present.

3.3.1. Preparing the Homemade Injection Device

For injecting the contrast agent, a homemade injection system (hereafter called injection device) is prepared at least 1 day prior to the MR experiments. 1. Cut two 35 cm pieces of polyethylene tubing (inner diameter 0.58 mm). 2. Take one 1.0 mL syringe and cut the tip off with a scalpel. 3. Put the two polyethylene tubes into the tip and fill the tip with glue. 4. When the glue has hardened, fill two 1.0 mL syringes with NaCl solution and connect them to the tubes with a 25

460

Thorsen

G needle into each tube. Both needles should be blunt, in order not to penetrate the tubes. 5. Check that both tubes are open and that there are no leakages, by squeezing the NaCl solution through the injection device (Fig. 23.2b).

Fig. 23.2. Preparing the rats for MR scanning. a Choosing the correct transmit/receive coil. b The homemade injection device. c Anaesthesia equipment. d Anaesthetising the rat. e Moving the rat onto the heating pad. f Checking that the catheter is correctly positioned into the tail vein. g Securing the catheter with surgical tape. h Immobilising the rat inside the rat bed. i Positioning the rat inside the MR scanner.

3.3.2. Preparing the T1 Contrast Agent

The Omniscan T1 contrast agent has a concentration of 0.5 mmol/mL. For i.v. injections on rats, a dose of 0.2 mmol/kg is commonly used. 1. Put 0.5 mL contrast agent into a 10 mL tube. 2. Dilute the contrast agent by putting 2.0 mL NaCl solution into the same tube. Mix well. The contrast agent is thereby diluted 1:5, giving a more suitable working concentration of 0.1 mmol/mL. 3. Withdraw the correct amount of contrast agent into a 1.0 mL syringe. As an example, a rat weighing 150 g would need the following amount of contrast agent: Amount of CA = 0.15 kg × 0.1 mmol/kg/0.1 mmol/ml = 0.15 mL of solution 4. Mix 6 drops of heparin into a 100 mL bottle of NaCl solution.

MRI of Experimental Gliomas

461

5. Fill one tube of the injection device completely with 0.9% NaCl/heparin solution and connect it to a syringe with 1.0 mL of NaCl/heparin. 6. Fill the other tube of the injection device with 0.1 mmol/mL of contrast agent and connect the syringe with the correct amount of contrast agent, according to the calculations made in point 3. To avoid confusion during injections, mark the syringe containing contrast agent with a permanent ink pencil. 7. Bring the injection device to the table in front of the MR scanner. 3.3.3. Preparing the MR with Accessories for Scanning

1. Helium and nitrogen levels must be checked and logged before start. The MR must not be used if the levels are too low. 2. Turn on the gradient electronics on the MR machine. 3. Turn on the scanner PC. Log in with your username and password. 4. Start the scanner software (Paravision 5.0 or other). 5. Select the correct animal transmit/receive coil (on our machine, rat head coil) and attach it inside the MR scanner (Fig. 23.2a). 6. Select the right animal bed and attach it to the front of the MR scanner. 7. The animals must be heated during scanning, otherwise they will lose the body temperature very quickly. Use heated air or a water heating unit with circulating hot water. The following instructions assume the use of a water heating unit. 8. Place the warming pad inside the animal bed and check that there is no leakage of water. 9. Connect the breathing patch to the monitoring device. 10. Fill the gas anaesthesia unit with isofluorane. Turn the unit on and adjust the unit so that it delivers around 2% isofluorane. Turn it off again and connect the anaesthesia tube to the animal bed.

3.3.4. Preparing the Animals for Scanning

1. Fill the gas anaesthesia unit in the preparation area (Fig. 23.2c). 2. Bring the animals from the animal room into the area for animal preparation. 3. The animals must then be weighed, in order to calculate the correct amount of contrast agent (proceed with Section 3.3.2 before continuing with this procedure).

462

Thorsen

4. Put the rat inside the anaesthesia chamber and anaesthetise the rat with 4% isofluorane, 50% O2 and 50% N2 O. Monitor the animal while it falls asleep (Fig. 23.2d). 5. Move the anaesthetised rat onto the large heating pad and position the anaesthesia mask over the nose of the animal (Fig. 23.2e). Change the stream of isofluorane gas from the chamber to the mask and adjust the isofluorane level down to 2%. 6. Place a 24 G intravenous catheter into the tail vein. Check that the catheter is correctly placed inside the tail vein by injecting 0.1 mL NaCl solution containing heparin (Fig. 23.2f). 7. Secure the catheter with surgical tape (Fig. 23.2g) (see Note 1). 8. Turn on the anaesthesia unit next to the MR scanner and check that the isofluorane level is 2.0, 50% O2 and 50% N2 O. 9. Move the anaesthetised rat quickly into the animal bed in front of the MR. 10. Place the breathing patch under the rat in the chest area. Secure the abdominal area of the animal with two pieces of surgical tape (width 25 mm), and the head area with one piece of tape (width 12.5 mm). Ensure that the breathing signal is visible on the monitor. 11. Adjust the amount of anaesthesia, so the respiration frequency is between 50 and 70 bpm (see Note 2). 12. Place the temperature probe into the rectum of the animal and secure it with surgical tape. Check that the temperature is read by the monitoring device (see Note 3) (Fig. 23.2h). 13. Connect the injection device to the catheter in the tail vein. Again check that the catheter is working by slowly injecting 0.1 mL NaCl solution containing heparin. 14. Position the animal inside the MR scanner (Fig. 23.2i). Ensure that the tubes of the injection device are not squeezed and thereby blocked when the rat is put inside the scanner. 3.3.5. MR Imaging of Experimental Gliomas

The parameters described in the protocols below should be regarded as starting values, and they should be optimised for each individual MR system. For tumour imaging, T2w images before contrast injection and T1w images before and after contrast administration are acquired. 1. Register data about the animal in the scanner software (see Note 4).

MRI of Experimental Gliomas

463

2. Perform a pilot scan, using, for instance, a 2D gradient echo FLASH sequence (called 1-TriPilot or 1-TriPilot-multi in the software and found in the location B_ANATOMY). The purpose of this scan is to perform auto-adjustments of the machine, to check and adjust the position of the rat inside the MR scanner and to obtain images for positioning the slices on the next scan (more about the auto-adjustments can be found in the ParaVision 5 manual). It is recommended to acquire multiple slices (3– 5 slices) in all three directions (coronal, sagittal and axial). The following protocol works fine for most cases: Repetition time (TR) 200 ms, echo time (TE) 5 ms, acquisition matrix (matrix) 128×256, 1 mm slice thickness, 1.5 mm slice separation, number of averages (NA) 1, flip angle (FA) 30◦ , 5 slices in each direction (coronal, axial and sagittal). 3. By inspecting the sagittal images, ensure that the brain is positioned in the centre of the magnet (Fig. 23.3a). Otherwise, reposition the animal and redo the overview scan. 4. Good anatomical images of the brain and the tumours are obtained by performing a T2-weighted sequence (we recommend a spin echo RARE sequence, for instance, TurboRARE- T2, found in location B_ANATOMY). Ensure that the scan field of view (FOV) is placed so that

Fig. 23.3. MR imaging of experimental gliomas. a Saggital view of the rat brain, showing correct positioning of the head inside the scanner. b T2-weighted image through the central part of the tumour. c T1-weighted image before contrast injection, with the same positioning as the T2-weighted image. d T1-weighted image after contrast injection.

464

Thorsen

the rat head is in the centre of the FOV. The following parameters are recommended as starting values: TR 4,200 ms, TE 36 ms, 256×256 matrix, 1 mm slice thickness, 1 mm slice separation, 20 slices, 4 cm field of view, NA 4, flip angle 180◦ , rare factor 8. Total scan time is around 6 min 30 s (Fig. 23.3b). 5. Perform a T1-weighted MR image series before the contrast agent is administered. For the experimental glioma models described here, either a multispin echo MSME or a spin echoe RARE sequence. Use the same FOV, the same positioning and the same number of slices as for the T2w sequences. The following parameters are recommended as starting values: TR 1,000 ms, TE 8.7 ms, 256×256 matrix, 1 mm slice thickness, 1 mm slice separation, 20 slices, 4 cm field of view, NA 2. Total scan time is around 6 min (Fig. 23.3c). 6. Set up a new T1w sequence, with exactly the same positioning and parameters as described in point 5. In ParaVision 5 this is done by choosing “Clone Scan” in the scan control window. Slowly inject the contrast agent to the animal over a time period of 20 s. Then inject 0.1 mL NaCl solution. Wait for 1 min and start the T1 scan (Fig. 23.3d). 3.3.6. MR Spectroscopy (MRS) of Experimental Gliomas

For the sake of clarity, the protocol for proton spectroscopy is described in a separate section. However, it is important that the spectroscopy is performed before any administration of contrast agents, as the gadolinium contrast will affect the tissue properties considerably, thus destroying the metabolite spectra completely. With the appropriate MR scanner hardware, MRS is relatively easy to perform. MRS can be performed using any nucleus with the proper spin number. The most commonly applied nuclei for medical and preclinical applications include hydrogen, phosphorus and carbon. The acquisition of proton spectra is discussed in the following section. There are two basic methods used to sample a given volume in MRS: A. Stimulated echo acquisition mode (STEAM), using three 90◦ pulses to obtain a stimulated echo. B. Point resolved spectroscopy (PRESS), using one 90◦ pulse and two 180◦ pulses to obtain a spin echo. The STEAM sequence can be performed with very short echo times, thereby increasing the number of metabolites that can be observed. However, the PRESS sequence is used by most authors today, because it provides double the signal to noise ratio

MRI of Experimental Gliomas

465

(SNR) when compared with STEAM. We have also been using the PRESS sequence with short echo times (down to 6 ms) and been able to discriminate all the commonly reported metabolites found in experimental brain tumours (12). Several brain metabolites can be studied by doing MR spectroscopy. The main peaks are N-acetylaspartate (NAA, peaks 2.02, 2.5 and 2.6 ppm), choline (cho, peak 3.22 ppm), creatine (cre, peaks 3.02 and 3.94 ppm), myo-inositol (mI, peaks 3.56 and 4.06 ppm), glutamine and glutamate (Glx, peaks 2.1–2.5 ppm) and lactate (lac, peak doublet centred at 1.33 ppm). The water peak is found at 4.7 ppm. Proton spectroscopy can be performed with either single voxel or multi-voxel techniques. The following procedure describes single voxel proton spectroscopy using the PRESS sequence and assumes that the user has some basic knowledge of moving the obtained spectrum into the analysis part of the Paravision software (TOPSPIN). The Bruker user manual describes the necessary procedures and commands used in TOPSPIN, and only the necessary commands for spectrum calibration and transformation are described in the following. More information about proton spectroscopy can also be found in Section 5.11 of the ParaVision 5 application manual. When routine spectroscopy experiments are performed, it is necessary to change very few parameters. The operator usually needs to set the repetition time TR, the echo time TE and the number of averages NA. Furthermore, the operator needs to verify or adjust the shim, the frequency and the RF gains. For such routine work, it is recommended to use the macro Localised_Spectroscopy_Guide, which is found in the location B_SPECTROSCOPY. The use of this macro is described in the following. 3.3.6.1. Localising the Volume of Interest

For in vivo imaging, accurate voxel positioning should be done on images with a high contrast, for example, RARE images. 1. Perform a 1-TriPilot-multi. 2. Perform a T2w scan (see point 4 in Section 3.3.4 above).

3.3.6.2. Adjusting the Field Homogeneity

A good homogeneity of the main magnetic field within your volume of interest is the most important requirement to obtain proton spectra of good quality. This will increase the signal obtained, and nearby resonance lines can be discriminated. 3. Choose the sequence Fastmap-4 mm in the B_SPECTROSCOPY location. Position your voxel of interest within the tumour area, using the T2w images as reference. For practical purposes, it is recommended to perform spectroscopy with a minimum voxel size of

466

Thorsen

3.0 mm, due to SNR limitations. The shim should be performed on a larger voxel, for instance, 3.5 mm. 4. Perform a local shim by starting the macro FastMapScout. 5. Start the shim by pushing the button “Adjust 1st&2nd Order”. 6. Check that the flatness of the measured homogeneity profile is adequate (the parameter “Evolution time” should be at least 20 ms in the normal rat brain). If the shimming is poor, it should be repeated. 3.3.6.3. Performing Spectroscopy of the Water Peak

The resolution of the spectrum should be quantified by measuring the FWHM of the water peak. For a 7 T scanner, a width of around 14 Hz is commonly measured in the normal rat brain and 15–20 Hz inside the tumour. 6. Prepare for spectroscopy of the water peak by choosing your own optimised, localised spectroscopy protocol, or use the sequence PRESS-waterline of the B_SPECTROSCOPY location. 7. Ensure that the band width is 1.5–2 times larger than the entire spectral range of resonance lines. In the PRESS sequences, the spectrum is by default 4006.41 Hz. 8. Choose a voxel size of 3.0 mm and carefully position the voxel inside the shimmed volume of 3.5 mm (see above). 9. Start PRESS-waterline by choosing Localised_Spectroscopy_Guide.

the

macro

10. Perform an extra-shim by pressing Shim-AutoAdj. Then the software tries to do the FastmapScout even better, and, if it succeeds, the new values will be used, otherwise they will be discarded. You can only see if it works by studying the water peak, which should become more narrow, and to look at the “fid” in the right part of the Acq/Reco Display window, which should become longer. 11. Set the frequency by clicking SF and then run AutoAdj. 12. Set eddy current correction off and frequency lock off (see Note 5). 13. Set number of acquisitions to 4. 14. Start the scan. Total scan time is around 10 s. 3.3.6.4. Analysing the Water Peak

15. Import the spectrum into the analysing software TOPSPIN by right clicking on the scan in the scan control window and choose “export to TOPSPIN” (see Note 6). 16. Open the TOPSPIN window and write the following commands into the bottom line (each command followed by pressing the Enter key): si 8 k, lb 0, em, ft, apk (see Note 7).

MRI of Experimental Gliomas

467

17. Measure the FWHM of the water peak (the necessary commands are found in the Paravision 5.0 user manual). 3.3.6.5. Performing Spectroscopy of the Brain Volume of Interest

17. Prepare for spectroscopy of the metabolites by choosing the sequence PRESS-1H. 18. Adjust the voxel size to 3.0 mm and place it in exactly the same position as the voxel used to measure the water peak. 19. Start PRESS-1H by choosing the macro Localised_ Spectroscopy_Guide. 20. Then again set the frequency by clicking SF and then run AutoAdj. Set the NA to 256 and TE to 6 ms. 21. Again do not use frequency lock and eddy current correction. 22. Start the scan. Total scan time is around 15 min. 23. You should now have a look at your spectrum: Import the spectrum into TOPSPIN by right clicking on the scan and choose export to TOPSPIN. Open the TOPSPIN window and write the following commands into the bottom line: si 8 k, lb 0, em, ft, apk (see Note 7). 24. If the spectrum is noisy (as you would expect when measuring tumour metabolites), repeat point 23 above and increase smoothing of the spectrum by using lb 2. Typical spectra from normal rat brain and rat brain tumour are seen in Fig. 23.4.

3.3.6.6. Quantification of Absolute Metabolite Concentrations

3.3.7. After the MR Scanning

The quantification of the absolute concentrations of the metabolites will not be described in this section. A suitable program for such analysis is, for instance, LCModel, described by Provencher (16). The reference manual to this analysis software describes in detail how to perform the quantification. 1. Turn off the gas anaesthesia unit. 2. Withdraw the animal bed and remove the surgical tape securing the animal. 3. Remove the rectal probe. 4. Turn off the anaesthesia. 5. Remove the animal from the bed and bring the animal back to its cage. 6. Monitor the animal until it wakes up, this usually takes 2–5 min. 7. Bring the rat back to the animal room.

468

Thorsen

Fig. 23.4. Proton spectra from animal brain. a Typical spectrum from the normal rat brain. b Typical spectrum from an experimental glioma, showing decrease in N-acetylaspartate, and increases in choline and myo-inositol. A decrease in N-acetylaspartate corresponds to loss of neurons. Choline is a component of phospholipid metabolism and a marker of cellular turnover. The increase in choline thus reflects increased cell proliferation. Myo-inositol is a glial marker, located in astrocytes, thus reflecting the glial nature of the tumour cells. To determine changes in the other metabolites, a more precise quantification has to be performed (see, for instance, Ref. 12). Abbreviations: Cre: creatine, Gua: guanine, mI: myo-inositol, Cho: choline, NAA: N-acetylaspartate, Glx: glutamine and glutamate, Lac: lactate, Lip: lipids.

MRI of Experimental Gliomas

469

4. Notes 1. Putting the catheter into the tail vein is a tricky procedure which takes a lot of practice. Alternatively, 1 mL of undiluted contrast agent (0.5 mmol/mL) can be injected subcutaneously into the hip area of the rat, 5 min before the T1-weighted MR sequence is started. To avoid taking the animal out of the MR scanner for the administration of contrast, a catheter should be placed subcutaneously in the hip area and secured with surgical tape. The tube in the injection device, as well as a 1.0 mL syringe, is then filled with undiluted contrast agent. 2. If the animals show signs of sickness due to, for instance, tumour burden, the respiration cycle should be monitored extra carefully. Sometimes when the breathing drops below 40 bpm, there is a relatively high risk that the animals completely stop breathing. Therefore, when the respiration cycle drops down to 40, the amount of isofluorane gas must be decreased by at least 0.2%, sometimes more, until the respiration again stabilises between 50 and 70 bpm. 3. The scanning procedure might take between 30 and 90 min, and the animals might drop considerably in body temperature. It is therefore necessary to constantly monitor the temperature and adjust the temperature of the water bath to ensure that a stable body temperature between 35 and 37◦ C is achieved. 4. For Paravision users: create a new patient, if it the first time you scan this animal, thereafter register the patient reference and a study name. Also note that patient reference is used to create directories containing the raw data, so if you are transferring the data to other computers after scanning, it is important to use a meaningful name which you will be able to recognise. 5. We have verified that the “frequency lock” option deteriorates the spectrum, using a 50 mL water phantom containing 1 mL absolute alcohol. We expected four peaks (CH3 group) and three peaks (CH2 group), but with “frequency lock on” we got eight peaks for CH3 and six peaks for the CH2 group. 6. TOPSPIN is an integrated software package for displaying NMR data, printing and plotting spectra, importing NMR data from various file formats, exporting displays and plots in various graphs and metafile formats, archiving spectra, and performing various data analyses on spectra. The software is well described in the Paravision manuals.

470

Thorsen

7. si: Size of spectrum. lb: Line broadening. em: Exponential window multiplication. ft: Fourier transform. apk: Automatic phase correction.

Acknowledgments Our own experimental studies referred to in this chapter have been supported by the Norwegian Cancer Society, the Norwegian Research Council, Innovest AS, Haukeland University Hospital, the Bergen Translational Research Programme, the Centre Recherche de Public Santé Luxembourg and a grant from the EU 6th Framework Programme (IP project “Angiotargeting”, contract no. 504743). The MR imaging has been performed at the Molecular Imaging Center (MIC) at the University of Bergen, Norway. References 1. Louis, D. N., Ohgaki, H., Wiestler, O. D. et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007;114:97–109. 2. Wen, P. Y., Kesari, S. Malignant gliomas in adults. N Engl J Med 2008;359:492–507. 3. Wrensch, M., Minn, Y., Chew, T., Bondy, M., Berger, M. S. Epidemiology of primary brain tumors: Current concepts and review of the literature. Neuro Oncol 2002;4:278–299. 4. Stupp, R., Mason, W. P., van den Bent, M. J. et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 2005;352:987–996. 5. Sakariassen, P. O., Prestegarden, L., Wang, J. et al. Angiogenesis-independent tumor growth mediated by stem-like cancer cells. Proc Natl Acad Sci USA 2006;103:16466– 16471. 6. McKnight, T. R. Proton magnetic resonance spectroscopic evaluation of brain tumor metabolism. Semin Oncol 2004;31:605–617. 7. Mulder, W. J., van der Schaft, D. W., Hautvast, P. A. et al. Early in vivo assessment of angiostatic therapy efficacy by molecular MRI. FASEB J 2007;21:378–383. 8. Thomale, U. W., Tyler, B., Renard, V. et al. Neurological grading, survival, MR imaging, and histological evaluation in the rat brainstem glioma model. Childs Nerv Syst 2009;25:433–441.

9. Wintersperger, B. J., Runge, V. M., Tweedle, M. F., Jackson, C. B., Reiser, M. F. Brain tumor enhancement in magnetic resonance imaging: Dependency on the level of protein binding of applied contrast agents. Invest Radiol 2009;44:89–94. 10. Thorsen, F., Enger, P. O., Wang, J., Bjerkvig, R., Pedersen, P. H. Human glioblastoma biopsy spheroids xenografted into the nude rat brain show growth inhibition after stereotactic radiosurgery. J Neurooncol 2007;82:1–10. 11. Huszthy, P. C., Goplen, D., Thorsen, F. et al. Oncolytic herpes simplex virus type-1 therapy in a highly infiltrative animal model of human glioblastoma. Clin Cancer Res 2008;14:1571–1580. 12. Thorsen, F., Jirak, D., Wang, J. et al. Two distinct tumor phenotypes isolated from glioblastomas show different MRS characteristics. NMR Biomed 2008;21:830–838. 13. Wang, J., Sakariassen, P. O., Tsinkalovsky, O. et al. CD133 negative glioma cells form tumors in nude rats and give rise to CD133 positive cells. Int J Cancer 2008;122:761–768. 14. Majos, C., Alonso, J., Aguilera, C. et al. Adult primitive neuroectodermal tumor: Proton MR spectroscopic findings with possible application for differential diagnosis. Radiology 2002;225:556–566.

MRI of Experimental Gliomas 15. Tong, Z., Yamaki, T., Harada, K., Houkin, K. In vivo quantification of the metabolites in normal brain and brain tumors by proton MR spectroscopy using water as an internal standard. Magn Reson Imaging 2004;22:1017– 1024. 16. Provencher, S. W. Estimation of metabolite concentrations from localized in vivo

471

proton NMR spectra. Magn Reson Med 1993;30:672–679. 17. Boulanger, Y., Labelle, M., Khiat, A. Role of phospholipase A(2) on the variations of the choline signal intensity observed by 1H magnetic resonance spectroscopy in brain diseases. Brain Res Brain Res Rev 2000;33:380–389.

wwwwwww

Chapter 24 MRI in Experimental Stroke Timothy Q. Duong Abstract Stroke is the third leading cause of death and the leading cause of long-term disability in the United States. Brain imaging data from experimental stroke models and stroke patients have shown that there is often a gradual progression of potentially reversible ischemic injury toward infarction. A central core with severely compromised cerebral blood flow (CBF) is surrounded by a rim of moderately ischemic tissue with diminished CBF and impaired electrical activity but preserved cellular metabolism, often referred to as the “ischemic penumbra.” Re-establishing tissue perfusion and/or treating with neuroprotective drugs in a timely fashion is expected to salvage some ischemic tissues. Diffusion-weighted imaging (DWI) based on magnetic resonance imaging (MRI) in which contrast is based on water apparent diffusion coefficient (ADC) can detect ischemic injury within minutes after onsets, whereas computed tomography and other imaging modalities fail to detect stroke injury for at least a few hours. Along with quantitative perfusion imaging, the perfusion–diffusion mismatch which approximates the ischemic penumbra could be defined non-invasively. This chapter describes stroke modeling, perfusion, diffusion, and some other MRI techniques commonly used to image acute stroke and, finally, image analysis pertaining to experimental stroke imaging. Key words: MRI, perfusion–diffusion mismatch, ADC, CBF, DWI, PWI, experimental stroke model, rodents, fMRI.

1. Introduction Stroke is the third leading cause of death and the leading cause of long-term disability. A stroke is caused by a disturbance in the blood supply to the brain, resulting in loss of brain functions. Stroke is a medical emergency. Earlier detection and earlier treatment would mean more brain tissue can be salvaged. There are two types of stroke. Ischemic stroke, which occurs as a result of an obstruction within a blood vessel accounts for M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_24, © Springer Science+Business Media, LLC 2011

473

474

Duong

about 85% of all stroke cases. Hemorrhagic stroke which occurs as a result of bleeds into the surrounding brain and subsequent increase intracranial pressure accounts for about 15% of stroke cases. According to the statistics published recently by the American Heart Association (1), someone in the United States suffers a stroke approximately every 40 s, and 5.8 million Americans have permanent neurological deficits from stroke with more than 71% of these stroke survivors unable to return to work. The American Heart Association projected that $70 billion will be expended on the care of stroke patients in 2009 (2). This cost is steadily rising because the conditions that put people at risk for stroke (such as heart disease, diabetes, and obesity) are also steadily on the rise. Magnetic resonance imaging (MRI) provides flexible and multiple clinically relevant information to image stroke in a single setting. In particular, diffusion-weighted imaging (DWI) (3) in which contrast is based on water apparent diffusion coefficient (ADC) is widely recognized as a useful imaging modality, because of its ability to detect stroke within minutes after onsets, whereas computed tomography and other imaging modalities fail to detect stroke injury for at least a few hours. Hyperintense regions on DWI correspond to tissues with a reduced apparent diffusion coefficient (ADC) of water. Although the biophysical mechanism(s) underlying ADC reduction remains poorly understood and controversial (3, 4) the ADC decline has been correlated with energy failure and breakdown of membrane potential in animal models (5–7). CBF can be measured by using an exogenous intravascular contrast agent or by magnetically labeling the endogenous water in blood (8, 9). The former is efficient, but it is incompatible with dynamic CBF fMRI as the long half-life of the contrast agent allows only one CBF measurement per bolus injection. Arterial spin labeling (ASL) techniques, on the other hand, are totally non-invasive, and the labeled water has a favorable short half-life (∼blood T1 ), making it possible to perform multiple repeated measurements that can be used to augment spatial resolution and/or signal-to-noise ratio. In humans, the “perfusion–diffusion mismatch” is presumed to approximate the “ischemic penumbra.” Although the strict definition of ischemic penumbra requires correlation with energy metabolism (5–7) and such a correlation is not feasible in humans, the “ischemic penumbra” and viability thresholds have been operationally defined based on DWI, PWI, and equivalent modalities. While the “perfusion–diffusion mismatch” is widely observed in acute human stroke (10–14) similar observations in animal stroke models have been limited and the temporal evolution of the perfusion–diffusion mismatch in animal models has yet to be systematically investigated. Animal models where focal

MRI in Experimental Stroke

475

ischemia can be reproducibly studied under controlled conditions would be important for identifying and predicting the severity of ischemic injury and for evaluating the efficacy of therapeutic intervention. In this chapter, we describe the stroke surgery procedures, a few common MRI protocols used in acute stroke imaging and, finally, image analysis pertaining to experimental stroke imaging in rats.

2. Materials 2.1. Stroke Modeling

1. Rats (200–250 g) (vendor: many) 2. Anesthetics (isoflurane or pentobarbital, etc.) (vendor: many) 3. Common surgical tools and supplies (vendor: many) 4. 4–0 monfilament nylon suture for occlusion (vendor: many) 5. PE-50 tubing (vendor: Fisher Scientific or Cole Palmer) 6. Warm pad, temperature feedback monitoring, and other monitoring equipment to ensure normal animal physiology (vendor: Fisher Scientific or Cole Palmer) 7. TTC (2,3,5-triphenyltetrazolium chloride) for histology (vendor: Sigma)

2.2. MRI

1. Bruker 7 T scanner (Billerica, MA) 2. 40-G/cm BGA12 gradient insert (ID = 12 cm, 120-μs rise time) 3. Animal holder 4. Custom-made RF transmitter and receiver coils for brain imaging 5. Custom-made RF transmitter coil for arterial spin labeling 6. Actively decoupled switch box to detune RF coils 7. Other magnet, gradient, RF coil configurations should also work

2.3. Peripheral MRI Compatible Monitor Equipment and Animal Supports

1. Oximetry (heart rate, arterial oxygen saturation) – (vendor: Mouse Ox) 2. Blood pressure (invasive with artery catheterization) – (vendor: Biopac/Acknowledge) 3. Respiration rate via Biopac/Acknowledge)

force

transducer

–

(vendor:

476

Duong

4. Forepaw stimulation device – (vendor: many) 5. Circulating warm water bath (Haack water bath, Cole Palmer) 6. Temperature feedback regulator (Digisense, Cole Palmer) 7. Anesthetic delivery, such as vaporizer – (vendor: many)

3. Methods 3.1. Stroke Surgery

1. Male rats (200–250 g) are anesthetized with isoflurane (∼2%). Weights of animals should be within 50 g to ensure consistent lesion volume. Other anesthetics can also be used. All anesthetics have some effects on stroke outcome (such as infarct volumes) and thus studies need to be designed and interpreted with this confound in mind. Male rats are often used to avoid the effects of female hormone on ischemic injury. Female rats are also widely studied and some female hormones have been found to have neuroprotective effects. 2. Aseptic preparations are strongly encouraged as infection and immunological responses could affect outcome. 3. Focal brain ischemia is induced using the intraluminal suture occlusion method, originally described by Koizumi et al. (15) and adapted by our group (16, 17). The right common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA) are exposed through a midline incision of the neck. The ECA will be permanently ligated distally, severed and flipped 180◦ around, such that it will be placed parallel to the common carotid artery. The occluder is made of a 4–0 monofilament nylon suture with its tip rounded by flame or coated with silicone (the latter yields less vascular damages in reperfusion studies upon withdrawal of the occluder). Occluder is advanced via the ECA to cause occlusion until a mild resistance is felt. The length is typically 18–20 mm from the CCA and ICA bifurcation. ECA access is preferred, because the occluder can be withdrawn to pass the bifurcation to allow blood flow to resume from the CCA to the MCA during reperfusion. 4. The right femoral artery is catheterized for blood–gas sampling, continuous blood pressure and heart rate monitoring. These physiological parameters are important, because devi-

MRI in Experimental Stroke

477

ations could affect stroke outcome, increasing statistical scatters. 5. Rats are secured in a supine position on an MR-compatible rat stereotaxic headset, anesthesia is reduced to ∼1.1% isoflurane. Rats breathe spontaneously. Mechanical ventilation can also be used. Rectal temperature should be maintained at 37.0 ± 0.5◦ C. It is strongly suggested that heart rate, respiration rate, mean arterial blood pressure, and oxygen saturation (from oximetry) are monitored. Blood gas should be sampled once during a break between imaging scans. All recorded physiological parameters are within normal physiological ranges. MRI data are acquired at 30, 90, and 180 min and again at 24 h post-ischemia. 6. For reperfusion studies, the animals are taken out of the scanner and the suture is withdrawn past the bifurcation to allow blood flow to resume from the CCA to the MCA which can be confirmed visually. Successful reperfusion can be confirmed by MR angiography (MRA) and CBF MRI later. The animal is then placed back in the scanner. Three principle planes of the localizer images and EPI images are carefully aligned with images obtained before taking the animal out the holder. The accuracy of placing the holder and image slices to those before taking out the holder is typically within one pixel or less. The entire procedure typically takes ∼15 min. For a permanent occlusion study, the ECA will be ligated permanently; the occluder will be inserted via the CCA to block blood flow to the MCA. 7. Rectal temperature is maintained at 36.5–37.5◦ C and respirations are recorded throughout the study. Body core temperature is critical, because it could affect stroke outcome. 8. At the end of the study, animals are euthanized properly. The brain should be taken out of the skull as soon as possible (within 15 min) and prepared for histology. Brain slices are cut with the same thickness and orientation as the MRI acquisition. Brain slices are incubated in TTC (2,3,5-triphenyltetrazolium chloride, 0.125% w/v) solution at 37◦ C for about 30 min. The brain slices are then transferred to 10% formalin solution and stored at 4◦ C for 24 h. For histological analysis, brain slices are photographed and analyzed using the software BioScan OPTIMAS (Edmonds, WA). A typical photoscanner can also be used and analysis can be done using NIH Image (free software). Edema correction needs to be applied (18) to correctly derive infarct volume. Alternatively, transcardial perfusion fixation may also be used for immunohistochemistry.

478

Duong

9. Finally, readers are encouraged to review the series of recommendations on good laboratory practice aimed at preventing the introduction of bias in experimental stroke investigation by MacLeod et al. (19) 3.2. MRI

• Pilot scan • T2 -weighted MRI • DWI • ASL PWI (alternatives: FAIR PWI or DSC) • fMRI Position of RF coil: Position the RF coil as central to the region of interest as possible. For a surface coil, avoid pressing the coil too hard on the animal’s head as it would increase “loading,” which decreases SNR. Tune and match RF coil: Tune and match RF coil by adjusting the capacitors to 1 H resonance frequency and 50 . Position scan: Position the scan on x, y, and z to ensure the subject is centered. Open up the FOV if needed. Shimming: Run autoshim or manual shim as needed. Calibrate RF pulses: Calibrate RF pulses for given pulse shapes and durations. This can be set up to be done automatically. Pilot scan: Perform a pilot scan, using a 2D gradient echo FLASH or RARE sequence (10–30 s). Based on the pilot scan, plan 5–8 1.5 mm coronal slices to cover the region of interest. T2 MRI: T2 -weighted images are acquired using the fast spinecho pulse sequence (echo time per echo = 6.5 ms) with two different effective echo times (52 and 104 ms), echo train length 16, and 16 signal averages. Typical parameters are spectral width is 30–50 kHz, TR = 2–3 s (90◦ flip angle), pulse shape Gaussian or Sinc3, pulse duration 1–2 ms. DWI: ADCav can be obtained by averaging three ADC maps with diffusion-sensitive gradients separately applied along the x-, y- or z-direction. An average of at least three axes is preferred to minimize anisotropic effects. Single shot, echoplanar images (EPI) can be acquired with matrix = 64 × 64, spectral width = 200 kHz, repetition time TR = 2 s (90◦ flip angle), echo time TE = 37.5 ms, b value = 4 and three directions of 1,170 s/mm2 , separate between diffusion gradient = 24 ms, diffusion gradient duration δ = 4.75 ms, field of view FOV = 2.56 cm × 2.56 cm, eight 1.5 mm slices, and 16 averages (total time ∼ 2.5 min). CBF: There are two methods to measure CBF, namely continuous arterial spin labeling (cASL) or dynamic

MRI in Experimental Stroke

479

susceptibility-enhanced MRI with Magnevist (Gd-DTPA) or Omiscan (another contrast agent). With the latter, measurement can only be made once every hour or so because intravascular half-life of MRI is on the order of 6 min. In stroke, the contrast agent is often trapped and longer wait time may be necessary. For the cASL technique, single-shot, gradient-echo, echoplanar-imaging (EPI) acquisition is used. Paired images are acquired alternately - one with arterial spin labeling and the other without (control). MR parameters were data matrix = 64×64, FOV = 2.56 cm × 2.56 cm, eight 1.5-mm slices, TE = 20 ms, and TR = 2 s (90◦ flip angle). Continuous arterial spin labeling employed a 1.78-s square radiofrequency pulse to the labeling coil in the presence of 1.0 G/cm gradient along the flow direction, such that the condition of adiabatic inversion is satisfied. The sign of the frequency offset is switched for control (non-labeled) images. Number of averages is typically 20–40, depending on the SNR needed. For the DSE technique, single-shot, gradient-echo, echoplanar-imaging (EPI) acquisition with matrix = 64×64, FOV = 2.56 cm × 2.56 cm, 3–5 slices of 1.5-mm, TE = 20 ms, and TR = 0.333 s (22◦ flip angle). Preload the iv line with 0.15–0.2 ml of Magnevist or Omiscan (typically 3 ft long of PE-50 tubing will hold such volume). Start the DSE acquisition of 1 min. About 20 s into the acquisition, deliver the contrast agent in a single bolus flush of saline. Continue DSE acquisition for another 40 s. Note that if DSE is used, GdDTPA has a non-negligible intravascular half-life. cASL and fMRI studies cannot be done immediately after Gd-DTPA injection. fMRI: Combined CBF and BOLD measurements are made using the continuous arterial spin-labeling technique with single-shot, gradient-echo, echo-planar-imaging (EPI) acquisition. Paired images are acquired alternately - one with arterial spin labeling and the other without (control). MR parameters were data matrix = 64×64, FOV = 2.56 cm × 2.56 cm, eight 1.5-mm slices, TE = 20 ms, and TR = 2 s (90◦ flip angle). Continuous arterial spin labeling employed a 1.78 s square radiofrequency pulse to the labeling coil in the presence of 1.0 G/cm gradient along the flow direction, such that the condition of adiabatic inversion is satisfied. The sign of the frequency offset is switched for control (non-labeled) images. For each set of CBF and BOLD measurements, 60 pairs of images (4 min) are acquired during baseline and 30 pairs (2 min) during hypercapnic challenge or forepaw stimulation.

480

Duong

• Hypercapnic challenges used a premixed gas of 10% CO2 with 21% O2 and balance N2 . • Forepaw somatosensory stimulation used the previously optimized parameters under identical isoflurane anesthetic condition in normal animals (20): 6 mA current with 0.3 ms pulse duration at 3 Hz. These stimulation parameters did not cause an increase in MABP. Needle electrodes are inserted under the skin of the two forepaws before surgery. The electrodes are connected in a series and the two forepaws are stimulated simultaneously. • Each trial consists of 4 min of data acquired during baseline and 2 min of data acquired during a hypercapnic challenge or forepaw stimulation. This is for combined BOLD and CBF measurements. If only BOLD fMRI is acquired, 2 min baseline and 1 min “stimulation” would be sufficient. 3.3. Image Analysis

Image calculation and co-registration can be done using codes written in Matlab (MathWorks Inc, Natick, MA) (20, 21). In addition to Matlab programs, we also use the STIMULATE (University of Minnesota) software for display and plotting. There are also many other free software programs available to calculate and display MRI images. (a) Map calculations: ADC maps with intensity in unit of mm2 /s are calculated pixel-by-pixel by using (22). ADC = − ln (S1 /S0 )/(b1 − b0 ) where bi = γ 2 Gi (2) δ 2 ( – δ/3), ln is the natural logarithm, S0 and S1 are the signal intensities obtained with b0 and b1 , respectively. The b-value is proportional to the gradient strength (G), magnetogyric ratio (γ ), duration of each gradient pulse (δ), and the time ( ) between applications of the two gradient pulses. ADC maps are calculated at each time point. For ASL images, CBF images (SCBF ) with intensity in units of mL/g/min are calculated (23, 24) pixel-by-pixel using SCBF = λ/T 1 • (Sc − SL )/(SL + (2 − 1)Sc ), where SC and SL are signal intensities of the control and labeled images, respectively. λ – 0.9 ml/g – is the partition coefficient (25), α is the labeling efficiency which is measured to 0.75–0.9 in animal models. For DSC-CBF calculation, the transverse relaxation rate ( R2∗ ) is calculated using the equation R2∗ (t) = – ln (S(t)/S0 )/TE, where S(t) is the signal intensity at time t, S0 is the precontrast baseline signal intensity, and TE is the sequence echo time. A CBF map is then generated by deconvolving the

MRI in Experimental Stroke

481

change in tissue concentration over the first pass of contrast agent with an arterial input function using singular value decomposition (26, 27). Mean transit time and cerebral blood volume can also be obtained with this analysis and they may be useful for stroke analysis. This software can be obtained from many sources. T2 maps can be calculated from at least T2 -weighted MRI with two echo times (TE). T2 = – ln (STE2 /STE1 )/(TE2 –TE1 ), STE2 and STE1 are the signal intensities obtained with TE2μ and TE1 , respectively. Thresholding ADC and CBF maps: To determine the ADC and CBF critical thresholds, ADC or CBF is separately lowered (via a Matlab program) until the CBF- and ADC-defined LV at 3 h numerically equal to the TTC infarct volume at 24 h. This method sets a fixed value below which the pixels within the ADC or CBF map are considered ischemic. The 3-h time point is chosen because the ADC-derived LV of this stroke model is shown previously to stop evolving by this time (28). The same thresholds can then used to calculate the LV for all time points. To utlize the critical thresholds, the ADC- and CBF-derived LV are determined using only the thresholds from Group I and without using TTC data. The ADC- and CBF-derived LV at 3 h are independently correlated with TTC-derived infarct volume at 24 h. Hypercapnic responses in different tissue types: Images obtained during the transition period between baseline and stimulus onset (30 s for CO2 challenge and 15 s for forepaw stimulation) are discarded. BOLD images are obtained from the control (nonlabeled) images of the CBF measurements. BOLD and CBF magnitude and percent changes relative to baselines are calculated: (1) on a pixel-by-pixel basis, (2) for the ISODATA-derived normal, mismatch and core clusters, and (3) for the ROI of the forepaw somatosensory cortices. Forepaw-stimulation responses in the forepaw cortices: Crosscorrelation analysis associated with the forepaw primary somatosensory cortices is performed. ROIs of the normal LH forepaw cortices are drawn based on the averaged crosscorrelation activation maps of all time points with references to the rat brain atlas and MRI anatomical images to avoid bias to any particular time point. The forepaw ROIs on the ischemic RH are obtained by symmetrically reflecting the LH ROIs along the midline to the RH. ADC, baseline CBF, and fMRI signals in the forepaw primary somatosensory cortex ROI are analyzed pixel by pixel, as well as by averaging pixels within the forepaw ROI. Magnitude baseline CBF and CBF changes are computed. 3.4. Additional Advanced Analysis

Evolution of “mismatch” pixels: The temporal and spatial evolution of the “mismatch” pixels, defined at 30 min after occlusion,

482

Duong

is evaluated as they migrated to different clusters. ADC, CBF, and BOLD under baseline and stimulated (CO2 or forepaw) conditions are analyzed for the pixels that subsequently migrated into the normal zone, core zone, or remained in the mismatch zone at 180 min post-ischemia. For permanent occlusion, ischemia stopped evolving 180 min post-occlusion which is taken as the imaging endpoint as demonstrated previously (17, 16). Pixel-by-Pixel Analysis: Pixel-by-pixel scatterplots of the CBF and ADC values are analyzed to evaluate the distribution of pixels over time. Only the center four slices are analyzed to minimize the misalignment between gradient-echo and spin-echo images at the ear canals. Four quadrants on the CBF-ADC scatterplots are derived using the TTC-derived ADC and CBF thresholds. The four zones are operationally defined as (i) the “normal” cluster where both ADC and CBF are above the thresholds, (ii) the “core” cluster where both ADC and CBF are below the thresholds, (iii) the “mismatch” cluster where the ADC is above the threshold but CBF is below the threshold, and (iv) “zone 4” where ADC is below the threshold, but CBF is above the threshold. Tissue volumes, means, and standard deviations of the ADC and CBF values of each cluster are evaluated at each time point. The history of the pixels that eventually became infracted is analyzed. The pixels from where the “core” (red) pixels came at the previous time points are colored blue in the CBF–ADC spaces. Projection profiles of the ADC and CBF distributions are also plotted at each time point. Iterative self-organizing data analysis (ISODATA): The ISODATA technique is an unsupervised segmentation method based on K-means clustering algorithm with additional iterative splitting and merging steps that allow statistical adjustment of the number of clusters and the cluster centers. Two major improvements based on Jacobs et al.’s algorithm (29) are incorporated, namely the use of Mahalanobis distance measure and spatial contiguity. In the original ISODATA method (30 31), Euclidean distance is used which did not take into account the variances of each feature parameter. Mahalanobis metric (32) removes several of the limitations of the Euclidean metric, namely (1) it automatically accounts for the scaling of the coordinate axes, (2) it corrects for correlation between the different features, and (3) it can provide curved or linear decision boundaries. Mahalanobis distance r can be written as r 2 = (x − mx ) Cx−1 (x − mx ), where r is the Mahalanobis distance from the feature vector x to the mean vector mx , and Cx is the covariance matrix for x.

MRI in Experimental Stroke

483

The surfaces on which r is constant are ellipsoids that are centered about the mean mx . In the special case where the features are uncorrelated and the variances in all directions are the same, these surfaces are spheres, and the Mahalanobis distance measure reduces to the Euclidean distance measure. Spatial contiguity incorporates spatial information when assigning clusters. Due to “noise” in the ADC and CBF measurements, a small fraction of (often single) pixels could be mistakenly assigned to another cluster. Consequently, a few scattered pixels of one class could be embedded in another class. The (dis) contiguity at a single pixel (j) is defined as the fraction of its spatial neighbors that are not in the same cluster: Dj =

Number of adjacent pixels i for which k(i) = k(j) Number of adjacent pixels

where k(j) is the cluster to which j belongs and k(i) is the cluster to which j ’s neighbors that i belongs. Eight neighbors are used in this study. For cluster re-assignment, another contiguity index of pixel j, Djl, is defined as Djl =

Number of adjacent pixels i for which k(i) = k(j) and k(i) = L , Number of adjacent pixels where L = 1 ∼ n_cluster and L = k(j) (n_cluster is the total number of clusters). Pixels are re-assigned if they had 6 or more out of 8 possible neighbors belonging to another class (Dj ≥ 6/8) and the class to which these pixels are to be re-assigned had to have 5 neighbors out of 8 possible neighbors (Djl ≥ 5/8). Both conditions needed to be satisfied; otherwise the pixel would not be assigned. The goal is to remove 1 or 2 “noisy” pixels only, avoiding erroneously re-assigning pixels, especially at 30 min post-occlusion, where there are small “islands” of normal tissues embedded in large abnormal ADC lesions.

Acknowledgments This work is supported in part by the National Institute of Neurological Disorders and Stroke, the National Institute of Health (R01 NS045879), the Scientist Development Grant (SDG-0430020N) and the Established Investigator Award (EIA 0940104N) from the American Heart Association. I thank my former and current colleagues who have participated on these stroke projects.

484

Duong

References 1. Rosamond, W., Flegal, K., Furie, K. et al. Heart disease and stroke statistics – 2008 update: A report from the American heart association statistics committee and stroke statistics subcommittee. Circulation 2008;117:125–146. 2. Lloyd-Jones, D., Adams, R., Carnethon, M. et al. Heart disease and stroke statistics – 2009 update: A report from the American heart association statistics committee and stroke statistics subcommittee. Circulation 2009;119:480–486. 3. Moseley, M. E., Cohen, Y., Mintorovitch, J. et al. Early detection of regional cerebral ischemia in cats: Comparison of diffusionand T2-weighted MRI and spectroscopy. Magn Reson Med 1990;14:330–346. 4. Duong, T. Q., Ackerman, J. J. H., Ying, H. S., Neil, J. J. Evaluation of extra- and intracellular apparent diffusion in normal and globally ischemic rat brain via 19F NMR. Magn Reson Med 1998;40:1–13. 5. Hoehn-Berlage, M., Norris, D. G., Kohno, K., Mies, G., Leibfritz, D., Hossmann, K. -A. Evolution of regional changes in apparent diffusion coefficient during focal ischemia of rat brain: The relationship of quantitative diffusion NMR imaging to reduction in cerebral blood flow and metabolic disturbances. J Cereb Blood Flow Metab 1995;15: 1002–1011. 6. Kohno, K., Hoehn-Berlage, M., Mies, G., Back, T., Hossmann, K. A. Relationship between diffusion-weighted MR images, cerebral blood flow, and energy state in experimental brain infarction. Magn Reson Imag 1995;13:73–80. 7. Back, M., Hoehn-Berlage, T., PhD, M., Kohno, M. D., Hossmann, K., PhD, M. D., Diffusion Nuclear, K. -A. Magnetic resonance imaging in experimental stroke correlation with cerebral metabolites. Stroke 1994;25:494–500. 8. Barbier, E. L., Lamalle, L., Decorps, M. Methodology of brain perfusion imaging. J Magn Reson Imaging 2001;13:496–520. 9. Calamante, F., Thomas, D. L., Pell, G. S., Wiersma, J., Turner, R. Measuring cerebral blood flow using magnetic resonance imaging techniques. J Cereb Blood Flow Metab 1999;19:701–735. 10. Albers, G. W. Expanding the window for thrombolytic therapy in acute stroke: The potential role of acute MRI for patient selection. Stroke 1999;30:2230–2237. 11. Heiss, W. D., Graf, R. The ischemic penumbra. Curr Opin Neurol 1994;7:11–19.

12. NINDS. Tissue plasminogen activator for acute ischemic stroke. The national institute of neurological disorder, and stroke rt-PA stroke study group. N Engl J Med 1995;333:1581–1587. 13. Rohl, L., Ostergaard, L., Simonsen, C. Z. et al. Viability thresholds of ischemic penumbra of hyperacute stroke defined by perfusion-weighted MRI and apparent diffusion coefficient. Stroke 2001;32: 1140–1146. 14. Schlaug, G., Benfield, A., Baird, A. E. et al. The ischemic penumbra: Operationally defined by diffusion and perfusion MRI. Neurology 1999;53:1528–1537. 15. Kiozumi, J., Yoshida, Y., Nakazawa, T., Ooneda, G. Experimental studies of ischemic brain edema: I: A new experimental model of cerebral embolism in rats in which recirculation can be introduced in the ischemic area. Jpn J Stroke 1986;8:1–8. 16. Meng, X., Fisher, M., Shen, Q., Sotak, C. H., Duong, T. Q. Characterizing the diffusion/perfusion mismatch in experimental focal cerebral ischemia. Ann Neurol 2004;55:207–212. 17. Shen, Q., Meng, X., Fisher, M., Sotak, C. H., Duong, T. Q. Pixel-by-pixel spatiotemporal progression of focal ischemia derived using quantitative perfusion and diffusion imaging. J Cereb Blood Flow Metab 2003;23: 1479–1488. 18. Tatlisumak, T., Carano, R. A. D., Takano, K., Opgenorth, T., Sotak, C. H., Fisher, M. A novel endothelin antagonist, A-127722, attenuates ischemic lesion size in rats with temporal middle cerebral artery occlusion: A diffusion and perfusion MRI study. Stroke 1998;29:850–858. 19. Macleod, M. R., Fisher, M., O’Collins, V. et al. Good laboratory practice: Preventing introduction of bias at the bench. Stroke 2009;40:e50–e52. 20. Liu, Z. M., Schmidt, K. F., Sicard, K. M., Duong, T. Q. Imaging oxygen consumption in forepaw somatosensory stimulation in rats under isoflurane anesthesia. Magn Reson Med 2004;52:277–285. 21. Shen, Q., Ren, H., Cheng, H., Fisher, M., Duong, T. Q. Functional, perfusion and diffusion MRI of acute focal ischemic brain injury. J Cereb Blood Flow Metab 2005;25:1265–1279. 22. Stejskal, E. O., Tanner, J. E. Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J Chem Phys 1965;42:288–292.

MRI in Experimental Stroke 23. Silva, A. C., Lee, S. -P., Yang, C., Iadecola, C., Kim, S. -G. Simultaneous blood oxygenation level-dependent and cerebral blood flow functional magnetic resonance imaging during forepaw stimulation in the rat. J Cereb Blood Flow Metab 1999;19:871–879. 24. Duong, T. Q., Silva, A. C., Lee, S. -P., Kim, S. -G. Functional MRI of calcium-dependent synaptic activity: Cross correlation with CBF and BOLD measurements. Magn Reson Med 2000;43:383–392. 25. Herscovitch, P., Raichle, M. E. What is the correct value for the brain-blood partition coefficient for water? J Cereb Blood Flow Metab 1985;5:65–69. 26. Ostergaard, L., Sorensen, A. G., Kwong, K. K., Weisskoff, R. M., Gyldensted, C., Rosen, B. R. High resolution measurement of cerebral blood flow using intravascular tracer bolus passages. Part II: Experimental comparison and preliminary results. Magn Reson Med 1996;36:726–736. 27. Ostergaard, L., Weisskoff, R. M., Chesler, D. A., Gyldensted, C., Rosen, B. R. High resolution measurement of cerebral blood flow using intravascular tracer bolus passages. Part I: Mathematical approach and statistical analysis. Magn Reson Med 1996;36:715–725.

485

28. Reith, W., Hasegawa, Y., Latour, L. L., Dardzinski, B. J., Sotak, C. H., Fisher, M. Multislice diffusion mapping for 3-D evolution of cerebral ischemia in a stroke model. Neurology 1995;45:172–177. 29. Jacobs, M. A., Zhang, Z. G., Knight, R. A. et al. A Model for multiparametric MRI tissue characterization in experimental cerebral ischemia with histological validation in rat: Part 1. Stroke 2001;32: 943–949. 30. Jacobs, M. A., Knight, R. A., SoltanianZadeh, H. et al. Unsupervised segmentation of multiparameter MRI in experimental cerebral ischemia with comparison to T2, diffusion, and ADC MRI parameters and histopathological validation. J Magn Reson Imag 2000;11:425–437. 31. Shen, Q., Ren, H., Bouley, J., Fisher, M., Duong, T. Q. Dynamic tracking of acute ischemic tissue fates using improved unsupervised ISODATA analysis of highresolution quantitative perfusion and diffusion data. J Cereb Blood Flow Metab 2004;24: 887–897. 32. Duda, R. O., Hart, P. E. Pattern Classification and Scene Analysis. New York, NY: Wiley; 1973.

wwwwwww

Chapter 25 Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of Parkinson’s Disease Anthony C. Vernon and Michel Modo Abstract Neurotoxin-based rodent models of Parkinson’s disease (PD) are widely used for pre-clinical evaluation of novel therapeutics for PD and have provided insights into mechanisms underlying motor dysfunction and nigrostriatal degeneration in PD. Predominantly, magnetic resonance imaging (MRI) studies in such models have focused on alterations in T2 water 1 H relaxation or 1 H MR spectroscopy (MRS), whilst potential morphological changes and their relationship to histological or behavioural outcomes have not been fully investigated. Identification of MR signal changes that are significantly related to behavioural and histological outcomes in pre-clinical PD models may identify useful non-invasive surrogate markers of nigrostriatal degeneration in vivo. Development of such in vivo imaging-based biomarkers may provide a simple, efficient and comprehensive means to study lesion progression and therapeutic interventions in rodent models of PD, which may also have translational value. Key words: Parkinson’s disease, T2 -weighted MRI, lactacystin, iron, proteasome.

1. Introduction Parkinson’s disease (PD) is a progressive neurodegenerative movement disorder characterised by a selective degeneration of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNc) (1). This is accompanied by formation of cytoplasmic inclusions in remaining neurons, termed Lewy bodies (LB), composed primarily of fibrillar aggregates of α-synuclein (2). Degeneration of nigral DA neurons results in significant depletion of striatal dopamine levels, which can be readily visualised in PD patients using positron emission tomography (PET) by specific radiotracers, such as 18-fluorodopa (18 F-DOPA) (3). However, whilst there is abundant PET data for PD, the results M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_25, © Springer Science+Business Media, LLC 2011

487

488

Vernon and Modo

of magnetic resonance imaging (MRI) morphometric studies regarding the basal ganglia (BG) in PD patients are still relatively scarce and inconsistent (4). Interestingly, recent voxel-based morphometry (VBM) studies have identified partial reductions in grey matter volume (GMV) including atrophy of the head of the left caudate nucleus and cortical GMV changes in both early and advanced PD (4, 5). Asymmetrical hypertrophy of the lateral ventricles (LV) has also been reported (6). Interestingly, these morphometric changes correlate robustly with patient disability scores on the Unified Parkinson’s Disease Rating Scale (UPDRS), raising the possibility that these may be potential surrogate markers of disease progression in PD (4–6). In addition to subtle volumetric changes, T2 water proton relaxivity rates are decreased in pathologically relevant areas in PD patients, including the SNc and putamen, which has been ascribed to regional iron accumulation (7–9). Interestingly, a significant relationship between T2 relaxivity, iron accumulation and clinical symptoms has been described, although only cross-sectional studies have been performed to date (9, 10). The possibility that morphometric changes or a combination of this with alterations in T2 relaxation time may be used to non-invasively monitor disease progression in PD, however, remains unclear (11). Crucial to a more mechanistic understanding of pathology and the evaluation of novel treatments for PD are animal models that more closely reflect the human disease. Rodent toxin-based models of PD, although subject to limitations, have provided useful insights into the pathophysiology of PD (12). Thus, combining animal models with non-invasive imaging, such as MRI, offers a powerful tool with which to investigate dynamic morphological and relaxivity changes due to degeneration of the nigrostriatal system (13). Importantly, detailed anatomical or relaxivity information can be correlated directly with behavioural phenotypes, raising the possibility that MRI could provide non-invasive surrogate markers predictive of the degree of functional impairment in individual animals (13). We have previously utilised MRI to scan in vivo brains from rodents bearing a nigrostriatal lesion induced by an intranigral injection of the proteasome inhibitor lactacystin (14). This relatively recently developed model has been suggested to be a novel, pathologically relevant model of PD (15). We here present the methodological framework in which to generate this animal model of PD and how to perform a simple locomotor behavioural test. We also provide a step-by-step guide to acquire serial T2 -weighted MR images from these animals for the serial assessment of morphological and relaxivity changes in vivo. Lastly, we provide guidance on how to perform basic histological analysis of pathology in this model and details of how to integrate behavioural, MR and histological measures by correlation analysis.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

489

2. Materials 2.1. Surgery

1. Animals (male, Sprague-Dawley rats, Harlan, Bicester, UK) 2. Stereotaxic frame (Kopf Instruments, Harvard Apparatus, Edenbridge, UK) 3. Leica M651 surgical microscope (Leica Microsystems, Milton Keynes, UK) 4. Hamilton syringe 701 series 10 μl (Fisher Scientific, Loughborough, UK) 5. 1 ml syringe (VWR, Lutterworth, UK) 6. 5 ml syringe (VWR, Lutterworth, UK) 7. 23 G needles (VWR, Lutterworth, UK) 8. 25 G needles (VWR, Lutterworth, UK) 9. Scalpel blades (10A, Harvard Apparatus, Edenbridge, UK) 10. Scalpel holder (Harvard Apparatus, Edenbridge, UK) 11. Surgical clamps (×2 Harvard Apparatus, Edenbridge, UK) 12. Surgical forceps (Harvard Apparatus, Edenbridge, UK) 13. Microdrill (FST, Interfocus, Cambridge, UK) 14. Vaseline (Boots Chemists, London, UK) 15. Sterile gauze swabs (Fisher Scientific, Loughborough, UK) 16. Industrial methylated spirits (70% v/v) (Fisher Scientific, Loughborough, UK) 17. Medetomidine hydrochloride (domitor, final concentration 0.25 mg/kg) (Pfizer, Sandwich, UK) 18. Ketamine hydrochloride (vetelar, final concentration 230 mg/kg) (Pfizer, Sandwich, UK) 19. Sterile water for injection (VWR, Lutterworth, UK) 20. Lactacystin (cat#L6785, final concentration 10 μg in 2.5 μl, Sigma-Aldrich, Poole, UK) 21. 0.9% saline solution (sterile) (Animal Care Ltd., York, UK) 22. Sutures Ethilon 4/0 (Ethicon, Edinburgh, UK) 23. Antibiotic powder (Battle, Hayward and Bover Ltd., Lincoln, UK) 24. Atipamezole hydrochloride (antisedan, final concentration 5 mg/kg) (Pfizer, Sandwich, UK) 25. Buprenorphine (final concentration 0.3 mg/kg) (Alstoe Animal Health, York, UK) 26. Automated rotameter system (TSE systems, Bad Homberg, GER).

490

Vernon and Modo

27. Apomorphine hydrochloride (cat#A4393, Sigma-Aldrich, Poole, UK) 28. Graphpad prism (v4.0, Graphpad software, La Jolla, USA) 2.2. MRI

1. Isoflurane (Abbott Laboratories Ltd., Queenborough, UK) 2. 7 T horizontal small-bore magnet (Varian systems, Palo Alto, CA, USA) 3. Head RF coil (custom made, David Herlihy) 4. Biopack physiological monitoring equipment (Linton Instrumentation, Norfolk, UK) 5. VnmrJ software (Varian systems, Yarnton, Oxford, UK) 6. JIM software (v5.0, Xinapse systems, Thorpe Waterville, UK)

2.3. Perfusion

1. Paraformaldehyde (cat#P6148; Sigma-Aldrich, Poole, UK) 2. Sodium phosphate dibasic (cat#S9763; Sigma-Aldrich, Poole, UK) 3. Sodium phosphate monobasic dihydrate (cat#71500; Sigma-Aldrich, Poole, UK) 4. Sodium chloride (cat#S9888; Sigma-Aldrich, Poole, UK) 5. Clamps/scissors (FST, Interfocus, Cambridge, UK) 6. Cardiac perfusion needle (FST, Interfocus, Cambridge, UK) 7. Perfusion pump (Watson-Marlow Bredel Pump, Falmouth, UK) 8. Plastic tubing (Fisher Scientific, Loughborough, UK) 9. Guillotine (Harvard Apparatus, Edenbridge, UK) 10. Sodium azide (final concentration 0.03% w/v) (cat#S2002; Sigma-Aldrich, Poole, UK) 11. Sucrose (final concentration 30% w/v) (cat#S9378; SigmaAldrich, Poole, UK)

2.4. Immunohistochemistry

1. Freezing microtome (Microm HM430; Thermo Fisher Scientific, Loughborough, UK) 2. PBS tablets (cat#P4417; Sigma-Aldrich, Poole, UK) 3. Anti-tyrosine hydroxylase antibody (AB151, Chemicon, Watford, UK) 4. Anti-neuron-specific nuclear protein antibody (MAB377, Chemicon, Watford, UK) 5. Hydrogen peroxide (cat#H-0904; Sigma-Aldrich, Poole, UK)

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

491

6. Triton X-100 (cat#T8787; Sigma-Aldrich, Poole, UK) 7. Normal goat serum (S-1000; Vector Labs, Peterborough, UK) 8. Biotinylated goat-α-rabbit secondary antibody (BA-1000; Vector Labs, Peterborough, UK) 9. Avidin–biotin horseradish peroxidase kit (PK-6100; Vector Labs, Peterborough, UK) 10. 3 3-Diaminobenzidinetetrahydrochloride Sigma-Aldrich, Poole, UK)

(cat#D5905;

11. Prussian blue (cat#03899; Sigma-Aldrich, Poole, UK) 12. Cresyl violet acetate (cat#C5042; Sigma-Aldrich, Poole, UK) 13. Superfrost microscope slides (Thermo Fisher Scientific, Loughborough, UK) 14. Ethanol (technical grade) (Fisher Scientific, Loughborough, UK) 15. Methanol (technical grade) (Fisher Scientific, Loughborough, UK) 16. Xylene (Fisher Scientific, Loughborough, UK) 17. Distyrene plasticizer xylene (DPX) mounting media (VWR, Lutterworth, UK) 18. Coverslips 24 × 60 mm (Thermo Fisher Scientific, Loughborough, UK) 2.5. Microscopy

1. Zeiss Axioimager with stereology kit (Carl Zeiss, Welwyn Garden City, UK) 2. Stereo investigator software (v7.0, MBF Bioscience, Magdeburg, Germany)

2.6. Correlations

1. SPSS software (v16.0, SPSS Inc., Woking, UK)

3. Methods 3.1. Proteasome Inhibitor Model 3.1.1. Animals

Male Sprague-Dawley rats (250±10 g, Harlan, UK) should be housed in groups of three at 21 ± 1◦ C on a 12 h light:dark cycle (lights on 07:00, lights off 19:00). Standard rat chow and drinking water should be available ad libitum. All animal experiments

492

Vernon and Modo

must be carried out with local ethical approval and appropriate national regulations. 3.1.2. Proteasome Inhibitor Preparation

1. Immediately prior to use on the day of surgery, dissolve lactacystin in 0.9% saline (pH 7.4) to give a final concentration of 10 μg in an injection volume of 2.5 μl. For this, dissolve 1 vial (approx. 200 μg) lactacystin in 50 μl of 0.9% saline solution, to give a 4 μg/μl solution. Thus, 2.5 μl of this solution contains 10 μg lactacystin, sufficient to lesion one animal. This dilution can be adjusted to increase or decrease the dose of toxin to be injected as desired. 2. Once prepared, it is not recommended by the manufacturer that the lactacystin solution be stored frozen, but should be prepared fresh for each surgical session. Using the dilution described in this protocol, a single 200 μg vial of lactacystin is sufficient to lesion approximately 15–18 animals, allowing for additional spare volume for spoilage or accidental spillage. 3. Once prepared, the lactacystin solution should be stored on ice and protected from light to minimise degradation during the course of surgery. Lower doses of lactacystin may be administered to produce smaller lesion sizes with less non-specific toxicity (as described previously by (16)). The appropriate dose of toxin should ideally be determined in pilot studies.

3.1.3. Stereotaxic Surgery

Lactacystin does not readily cross the blood–brain barrier, thus, to generate a nigrostriatal lesion, lactacystin must be injected intracranially using a stereotaxic surgical approach. The site of injection is dependent on the type of lesion required. Classically, to generate a nigrostriatal lesion, dopaminergic neurotoxins, such as 6-hydroxydopamine (6-OHDA), may be injected into one of three sites, the substantia nigra pars compacta (SNc), the medial forebrain bundle (MFB) or the striatum (STR). Each site has its advantages and disadvantages, which are summarised in Table 25.1 and in an excellent review (17). However, unlike 6-OHDA, lactacystin is not selectively toxic to dopaminergic cells alone and hence induces a general neuronal toxicity (16). Thus, direct injection of lactacystin into the SNc is preferable to minimise this non-specific toxicity. Nevertheless, lactacystin lesioning results in robust aggregation of α-synuclein, more closely mimicking the human disease (15, 16, 18).14 Furthermore, this model has also been used successfully in studies examining novel therapeutic agents for PD (19, 20). Furthermore, titrating the dose of lactacystin from 10 to 1 μg substantially reduces non-specific toxicity, but this results in a proportional reduction of the nigral lesion size (16). Therefore, we strongly recommend conducting

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

493

Table 25.1 Nigrostriatal damage overview. Advantages and disadvantages of the three typical lesion sites for modelling parkinsonism in rodents Lesion site

% Loss SNc neurons

% Loss striatal DA

MFB

60–97

SNC

CPU

Advantages

Disadvantages

>90

• Almost complete nigrostriatal lesion • Valuable model advanced PD • Robust parkinsonianlike behavioural deficits

• Cell loss in VTA (dependent on toxin dose) • Static lesion • Models advanced PD

80–90

80–100

• Cell loss pattern mimics more closely to human PD • DA depletion patterns mimic human PD • Valuable model of advanced PD • Robust parkinsonianlike behavioural deficits

• Small size of target may lead to accidental lesions of VTA and interanimal variations in lesion placement and size • Models advanced PD • Static lesion

30–80∗

20–70∗

• Generates partial lesion, • No consensus on dependent on toxin lesion site in striatum dose and injection site or number of injections • Model of early stage PD • Potential for general • Progressive retrograde neuronal toxicity loss nigral neurons with non-DA • Behavioural and selective toxins e.g. biochemical data closely rotenone or mimic human situation proteasome inhibitors (depending on dose)

∗ Depending on 1,2,3 or four-site lesioning paradigm (see Deumens et al., 2002 for detail)

pilot studies to determine the appropriate dose of lactacystin to minimise non-specific toxicity, whilst generating a robust nigrostriatal lesion. Alternatively, these studies can be carried out using a classical 6-OHDA lesion. In order to carry out this procedure, the following steps should be followed: 1. Randomise animals to either saline (control) or lesion (lactacystin) groups (n=10 per group). Weigh and anaesthetise animals by an i.p. injection of a mixture of medetomiR , 0.25 mg/kg) and ketamine dine hydrochloride (Domitor TM hydrochloride (Vetalar , 230 mg/kg) prepared in sterile water for injection. 2. Once the animal is sufficiently anaesthetised as determined by paw pinch reflex response, position the animal in a stereotaxic frame with the incisor bar set 3.3 mm below the

494

Vernon and Modo

interaural line. The correct positioning of ear bars in the ear canal may be observed by reflex blinking of the animal. Ensure the surface of the head is level and that ear bars are properly aligned. 3. Make a midline incision using a scalpel and expose the parietal bones for drilling. Ensure all bleeding is controlled. Fill the Hamilton syringe with toxin or saline solution as appropriate for injection and place it in the stereotaxic frame. Locate and position the needle above bregma, then move the needle to the following stereotaxic co-ordinates for an intranigral injection: anterio-posterior (AP): –5.2 mm, mediolateral (ML): +2.4 mm lateral from bregma (21) as previously described (15). Ensure good surgical and aseptic techniques are used at all times to minimise the risk of infection. Ensure all instruments are autoclaved and sterile before use. 4. Retract the needle slightly and make a burr hole at the co-ordinates given above in the skull using a microdrill. Care should be taken not to break the dura. Once completed, ensure the needle enters the burr hole cleanly without deviation. Control any bleeding and clean the skull of debris. 5. Place the tip of the needle, such that it just penetrates the dura. Read and record the final co-ordinate that is –7.6 mm ventral to dura (DV). Insert the needle to the correct depth in the brain, moving slowly to minimize mechanical injury. Lactacystin injection should then be performed at a rate of 0.5 μl/min using a motorized syringe pump. Slowly withdraw the needle no sooner than 5 min after completion of injection to minimize diffusion of toxin into the injection tract, preventing non-specific toxicity. 6. Remove the needle from the frame and clean carefully with ethanol and sterile saline. Control any bleeding, dust with antibiotic powder and close the scalp incision, using at least three sutures. 7. Reverse anaesthesia 1 h after induction by subcutaneous R , (s.c.) injection of atipamezole hydrochloride (Antisedan ◦ 5 mg/kg). Place the animal in a heated (37 C) recovery chamber until recovery from anaesthesia. Administer 0.3 mg/kg buprenorphine for analgesia, twice in the first 24 h post-surgery, plus 4 ml (2 ml per side) 0.9% saline solution i.p. for fluid replacement before returning the animal to its home cage. Animals may be given mashed high-nutrient food for 24 h post-surgery. Sham-lesioned animals should undergo identical surgery, but receive an injection of 0.9% saline. Saline and lactacystin groups

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

495

should be operated on in a randomised fashion in the same surgical session. Animals should be weighed and semi-quantitatively scored daily for neurological deficits using a general neurological rating scale, as previously described (22) until they have recovered pre-operative weights. 3.1.4. Rotational Asymmetry

Animals should be routinely handled prior to and after surgery in order to calm the animals and enhance reliability when testing. A simple behavioural test to measure the degree of nigrostriatal damage is drug-induced rotation by apomorphine injection. Following an injection of apomorphine, lesioned animals rotate in a contralateral direction to the lesioned hemisphere. The degree of rotation is directly proportional to the degree of nigral cell loss (23). Sham-lesioned animals should show no net increase in contralateral turning behaviour. 1. Harness lesioned and control animals into jackets tethered to an automated rotameter and allow animals to acclimatize for 30 min before administration of apomorphine hydrochloride (0.1 mg/kg dissolved in 0.9% saline, s.c.). 2. Measure the number of complete contraversive rotations each minute over 60 min using an automated rotameter. 3. To analyse the data, calculate the net number of contralateral turns using the formula: net contralateral turns = (contralateral rotation–ipsilateral rotation). This may be done in 5-min blocks for ease of presentation. 4. Net contraversive turns for lactacystin or saline-injected treated animals in response to apomorphine hydrochloride were plotted against time and the area under the curve (AUC) calculated in Graphpad Prism (v4.0 San Deigo, CA, USA). 5. The resultant means of AUC ± SEM for lactacystin or salineinjected animals were then compared using two-tailed paired student’s t-test. Representative results using this method are shown in Fig. 25.1.

3.2. MRI Acquisition 3.2.1. Preparation

T2W MR images from saline and lactacystin-injected animals were acquired using a 7.0 T horizontal small bore magnet and a custom-built head RF coil linked to a LINUX-based control console running VnmrJ acquisition software. Animals from each group should be randomised into the same scanning session. Choosing the time points at which MR images are acquired is dependent on the requirements of the study, for example, whether it is desired to examine acute or chronic changes in the brain following lesioning. An additional constraint is the time course of neuronal degeneration in toxin-based models of PD. Typically

496

Vernon and Modo

Fig. 25.1. Lactacystin-lesioned animals displayed significantly increased rotational asymmetry in response to apomorphine challenge (0.1 mg/kg s.c.). a Contraversive rotations over time following apomorphine injection in saline and lactacystin-lesioned animals. Data shown are mean contraversive rotations ± SEM. b Area under curve (AUC) analysis reveals a significant increase in contralateral rotations in lactacystin-lesioned animals compared to saline controls; ∗∗∗ P<0.001 (two-tailed students t-test).

with an intranigral or MFB lesion, neurodegeneration develops rapidly in the first 7 days following toxin injection, before becoming static (17). By contrast, intrastriatal lesions progress more slowly, such that nigral cell loss occurs progressively over a period of 3–4 weeks (17). In both models, the magnitude of the damage is dependent on the dose of toxin injected. Each of these should be considered when deciding on a time course for MR acquisition. In addition, MR scans should always be acquired at baseline prior to surgery in serial studies. In cross-sectional studies, this may not always be necessary, since comparisons will be made to a control group at the same time point. Regardless, prior to acquisition of MR images, these preparatory steps should be followed to place the animal in the magnet: 1. Anesthetise animals using an appropriate volatile agent for MR imaging (isoflurane, 5% induction, 1.5% maintenance) in an oxygen/medical air (30:70%) mixture, delivered at 1 L/min. 2. Centre the head of the animal inside the RF coil within a polytetrafluoroethylene (PTFE) MRI compatible stereotaxic head holder. Immobilise the head using head screws in the head mould to minimize movement artefacts. 3. Connect physiological monitoring equipment. Place a respiration rate pad under the animal’s body in the chest area and attach a pulse oxymeter clip to one of the animal’s hindpaws. Lubricate (vaseline) and insert the rectal temperature probe. Maintain animal body temperature at 37◦ C using a heated blanket.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

497

4. Check all connections are secure (particularly anaesthetic supply) and insert the RF coil apparatus into the centre of the magnet and secure. 5. Ensure physiological monitoring equipment is functioning normally, and animal blood pressure, heart rate, respiration rate and body temperature are within normal physiological ranges. These parameters should be monitored carefully for the duration of the scanning session. 3.2.2. MRI Parameters and Acquisition

Scanning parameters will depend on the scanner hardware. In this protocol, T2W images were acquired using a multi-echo, multislice spin-echo pulse sequence (MEMS), with the following scan parameters: TR = 4,200 ms; TE=10, 20, 30, 40, 50, 60, 70, 80 ms; 4 averages, FOV = 35 mm × 35 mm; matrix = 192 × 192; total scan duration 54 min. Although shorter T2W spinecho images can be acquired, a series of echo times is required to create a parametric T2 map to allow quantitative measurement of changes in tissue water proton relaxivity. This sequence offers a robust compromise between signal-to-noise ratio (SNR), image quality and time limitations for in vivo scanning with recovery from anaesthesia (∼60 min) to allow sufficient animal throughput in one continuous scanning session. Using this sequence, approximately eight animals may be scanned over a 10-h period. In our sequence, 50 contiguous 500-μm-thick coronal slices with an inplane resolution of 189 × 189 μm were acquired such that the entire brain of each animal was covered. Although protocols may vary between different MR systems and software, the following steps should act as a guide to successfully acquiring MR images for users of a Varian system, running VnmrJ software. 1. Acquire a scout image using a fast gradient-echo sequence to confirm accurate positioning of the animal inside the scanner bore. If necessary, remove the RF coil apparatus from the bore and adjust the animal position within the head holder and repeat the scout sequence to confirm correct positioning. Special care should be taken to ensure the animals’ head is not tilted in the magnet bore, thus preventing acquisition of unaligned images. 2. Shim the magnet to obtain optimal SNR. Since all subjects should be of similar size and weight, the original shim settings may be used for all subsequent scans in a single MRI session, although some users may prefer to shim the magnet for each individual animal. Importantly, however, shimming should be repeated at the beginning of each new session. 3. Ensure slice acquisition covers the whole brain and that positioning of the slices is consistent across animals. For example, on the scout image, always position the slice acquisition guide box such that the 50 slices are acquired from the base

498

Vernon and Modo

of the cerebellum, extending to the base of the olfactory bulb. 4. Initiate the T2W pulse sequence and carefully monitor physiological readouts for the duration of the scan. 5. Once scanning is completed, check that the images have saved correctly and perform a quick quality control to check for motion or intensity artefacts. Should these be present, particularly severe motion artefacts, it may be necessary to repeat the scan following steps 1–4 to ensure good quality images are obtained, as this will greatly facilitate subsequent quantitative analysis. Ensure the acquired .FID files are backed up in a separate location to the scanner console itself at intervals during the scanning session, when scans are not being acquired to prevent accidental loss of data. 6. Once the data are saved, remove the RF coil apparatus from the bore disconnect all monitoring equipment and the anaesthetic supply. Remove the head holder from the centre of the RF coil and remove the animal from the head holder. Place the animal in a heated recovery chamber in a separate holding room and administer 4 ml 0.9% saline solution (i.p., 2 ml per side). Following full recovery from anaesthesia, return animals to their home cages. 3.2.3. MRI Post-processing

Once MR images have been successfully acquired, some postprocessing of images is necessary prior to quantitative analysis. The reasons for this are to ensure quality control of images prior to analysis, conversion of MR images into a format that may be read by JIM v5.0 software for volumetric measurements and to generate parametric T2 maps for quantitative relaxivity measurements. Again, although protocols may vary between different MR systems and software, the following steps should act as a guide to post-processing MR images for users of a Varian system, running VnmrJ software. 1. Transfer brain images from the console to a workstation, for a detailed visual inspection for motion or intensity artefacts prior to post-processing. Scans displaying severe motion or intensity artefacts should be noted and discarded from further analysis. 2. Interpolate MR images from the acquired 192 × 192 matrix corresponding to a 189 μm in-plane linear pixel size to 256 × 256 pixel grid (136 μm voxel size) in VnmrJ. 3. It may be desirable to run images through the N3 algorithm (24) to correct image intensity inhomogeneity caused by scanner drift between different scanning sessions, which could affect T2 relaxivity analysis. If substantial image artefacts are present, it may be necessary to check the pulse sequence and RF coil tuning.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

499

4. For quantitative manual volumetric measurements, images corresponding to TE = 10–80 ms can be summed using the math function in VnmrJ. The resultant images may then be saved and exported as flexible data format (FDF) files and subsequently converted to analyse 7.5 format (Mayo Clinic, Rochester, USA) using the image converter tool in JIM v5.0. 5. Quantitative T2 relaxation maps may be obtained by applying a mono-exponential fit to the eight multi-echo images (TE =10–80 ms) using the math function in VnmrJ software. Resultant images should then be saved as .FID files for subsequent analysis in VnmrJ. 3.3. Quantitative Analysis of MR Images 3.3.1. Volumetric Analysis

For whole brain and regional volumetric analyses, brain structures can be delineated manually on MR images in each hemisphere, from each subject, in each treatment group by the following method. 1. Ideally, two independent reviewers blinded to animal surgical status should delineate individual brain structures. 2. Brain structures should be delineated manually on a slice-byslice basis in the coronal plane using the region-of-interest (ROI) tool in JIM v5.0 software. The number of slices will depend on the size of the structure being measured. 3. ROIs should be traced in both the ipsilateral and the contralateral hemispheres of each subject at low magnification followed by manual correction of borders at higher magnification. 4. Clear anatomical landmarks and reference to the standard rodent brain atlas should be used to define ROIs for each region analysed in each slice. Some examples of these are given in Fig. 25.2 for common brain regions, which may be measured reliably with the level of contrast available in these in vivo MR images. Once generated, ROIs should be saved for future reference and quality control procedures. 5. Volumes may then be calculated by multiplying the sum of the areas of a given structure on all slices by the slice thickness for each animal. 6. Inter- and intra-rater reliability should be assessed using the inter-class-correlation coefficient (ICC). An ICC value of ≥0.80 suggests robustly reliable segmentation of brain structures between reviewers. Some representative ROIs and results based on this method are shown in Table 25.1 and Fig. 25.3, respectively. Note that it is

500

Vernon and Modo

Fig. 25.2. Anatomical landmarks and criteria used for MRI volume measurements and sample ROI tracings for each brain region. Abbreviations: CSF, cerebrospinal fluid.

Fig. 25.3. a Representative data from ROI-based MRI volume measurements acquired 3 weeks post-surgery demonstrate a significant decrease in whole brain volume and the volume of the ipsilateral ventral midbrain and corpus striatum in lactacystin-lesioned animals compared to saline controls. No change in the absolute volume of the cerebellum could be detected. Data shown are mean volume in mm3 ± SEM; ∗ P<0.05 saline vs. lactacystin (two-tailed students t-test). In addition to volumetric changes, T2 relaxivity measurements (b) reveal a significant decrease in T2 in the substantia nigra, but not corpus striatum of lactacystin-lesioned animals compared to saline controls. Data shown are relative T2 (ms) expressed as the ratio of T2 values measured in the contralateral and ipsilateral hemispheres for each region, respectively. ∗ P<0.05; saline vs. lactacystin (two-tailed students t-test).

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

501

not possible to measure the volume of the SN itself accurately or indeed other small sub-cortical nuclei at the level of contrast available in our MR images. 3.3.2. T2 Relaxometry

Intensity values for T2 should be measured in both the contralateral and the ipsilateral hemispheres of saline and lactacystinlesioned animals in the striatum and the SN from quantitative T2 maps using VnmrJ software. To measure T2 using this protocol, carry out the following steps: 1. Draw an ROI encompassing the striatum or SN using the ROI tool in VnmrJ on representative coronal slice(s) containing these structures. 2. A rodent brain atlas should be used to interpret the MR images and define ROI borders. To identify the SN robustly, atlas plates containing the SN may be manually overlaid onto MR images to aid identification of this nucleus. Importantly, using this method no difference can be made between the pars reticulata and the pars compacta of the SN. Thus, T2 measurements in this area refer to the whole SN. 3. Values for T2 can be expressed as the ratio between the contralateral and the ipsilateral striatum and SN (relative T2 ).

3.4. Histological Analysis 3.4.1. Tissue Collection

To validate MR signal changes and analyse neuropathological changes in the brain, perfusion-fixed brain tissue must be collected. Note paraformaldehyde (PFA) is highly tumorigenic, so adequate personal protective equipment (PPE) must be worn at all times (lab coat, gloves, face mask). In addition, preparation of PFA and perfusion must be carried out in a fume hood. The procedure requires the following steps: 1. Prepare an appropriate volume of 8% PFA and phosphate buffer, which are mixed in equal volume to make a 4% PFA solution. Perfusion requires approximately 50 ml of 4% PFA per 100 g body weight and scale up the volume required accordingly. In addition, prepare an appropriate volume of 0.9% saline solution. 2. Weigh the animal and deeply anaesthetise with sodium pentobarbital (60 mg/kg i.p.). After the loss of righting and respiration reflexes, confirm depth of anaesthesia by paw or tail-pinch test. Do not proceed before these are completely absent. 3. Prime the perfusion pump with 0.9% saline solution, ensuring that no air bubbles are present in the tubing. Locate the xiphoid cartilage and cut with scissors. Lift the muscle tissues and open the abdominal cavity and cut the diaphragm. Clamp the xiphoid cartilage and open up the rib cage.

502

Vernon and Modo

4. Locate the descending aorta and clamp. Insert a cardiac perfusion needle into the left ventricle until it is visible in the ascending aorta. Clamp the needle in place and snip the right ventricle to allow blood to escape. Switch on the pump and flush with saline until the solution runs clear from the right ventricle. 5. Once the exiting fluid is clear, switch off the pump and swap the tube to the 4% PFA vessel without introducing bubbles. Restart the pump and perfuse with fixative with 50 ml/100 g rat. Once completed, switch off the pump and remove clips. Remove animal’s head with a guillotine. Open the cranium and carefully remove the brain, placing it in a pot filled with 4% PFA. Allow the brain to post-fix for 24 h, before rinsing twice in phosphate buffer and cryoprotection in 30% sucrose solution (w/v). 3.4.2. Immunohistochemistry

Prior to histological analysis, prepare the brain tissue by removing the cerebellum and sectioning serial slices through the whole brain (1 in 12, 40 μM thickness) on a freezing microtome. It is necessary to cut thicker sections for stereology experiments. Thus, the sections should be collected and stained free-floating to facilitate antibody penetration and give robust immunohistochemical staining. To evaluate nigrostriatal damage, one complete series per animal from each treatment group should be stained for the DA neuronal marker tyrosine hydroxylase (TH). To evaluate neuronal toxicity, the general neuronal marker, neuronspecific protein N (NeuN) should be used instead. Ideally, both stains should be performed (on separate tissue series) as changes in TH expression do not necessarily reflect neuronal loss. A standard immunoperoxidase method may be used as previously described (25): 1. Wash sections 3 × 5 min in 0.1 M phosphate buffered saline (PBS) 2. Quench endogenous peroxidase activity by incubation for 15–30 min in 70% methanol/PBS (v/v) solution containing 0.03% H2 O2 . 3. Permeabilise the tissue sections by incubation in PBS solution containing 0.3% triton X-100 for 15 min. 4. Block non-specific binding by incubating tissue sections in 10% normal sera (species depending on species primary antibody was raised in) diluted in PBS containing 0.3% triton for 30 min to 1 h. 5. Incubate with appropriately diluted α-TH or α-NeuN primary antibodies overnight at 4◦ C. Antibodies should be diluted in PBS containing 0.3% triton X-100 and 10% normal serum.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

503

6. Wash sections 3 × 5 min in PBS and incubate with appropriately diluted biotinylated secondary antibodies (dilute in PBS) at room temperature for 2 h. 7. Wash sections 3 × 5 min in PBS. Incubate sections in avidin– biotin-conjugated horseradish peroxidase kit for 1 h. This solution must be prepared 30 min before use, left at 4◦ C to mix and protected from light. 8. Wash sections 3 × 5 min in PBS prior to visualisation of immunostaining using 3 3-diaminobenzidine (DAB) as a chromogen. Mix 1 DAB tablet (10 mg) with 12 ml of PBS and filter. To this, add 12 μl of 30% H2 O2 in PBS. Add this solution to individual wells (one at a time) for approx. 30 s to 1 min, or until brown precipitate staining is observed. Quickly remove the DAB solution and quench the staining by washing the sections in PBS. Ensure incubation time in DAB is consistent across wells to give equal staining intensity. Wash sections a further 3 × 5 min in PBS once staining completed. 9. Mount sections in anatomical series on appropriately labelled charged slides and allow air-drying. Rinse sections in distilled water for 2 min. Rapidly dehydrate sections through a series of graded alcohol solutions (70, 90, and 100%) for 1 min each before clearing in neat xylene (2 min). Apply DPX mounting solution and coverslip. Leave coverslip slides to dry. Note this section must be performed in a fume hood as DPX is an irritant. Once dried, place slides in a labelled slide box and store in a cool, dry place prior to microscopic analysis. Representative photomicrographs of TH-immunostained sections from saline and lactacystin-injected animals are shown in Fig. 25.4. 3.4.3. Perl’s Prussian Blue Staining

As changes in T2 relaxivity in the SN, or indeed other brain areas, may be linked to iron accumulation, this may be evaluated by Perl’s stain for iron-containing cells or deposits. For this: 1. Prepare Perl’s solution by adding 1.0 g of potassium hexacyanoferrate (ferrocyanide) to 25 ml distilled water and 25 ml of 13% hydrochloric acid. This solution should be freshly prepared and not reused. Scale up volumes as appropriate if more solution is required. In addition, prepare an appropriate volume of 10% cresyl violet acetate solution. 2. Mount tissue sections (one series, per animal, from each group) onto labelled superfrost-coated slides and allow sufficient air-drying that sections will not float off slide in solution. Wash tissue sections in distilled water and incubate in Perl’s solution for 20–30 min.

504

Vernon and Modo

Fig. 25.4. Representative photomicrographs of the TH-immunostained sections from (a) the substantia nigra (SN) and (b) corpus striatum reveal the extent of nigrostriatal damage induced by lactacystin lesioning compared to saline controls at 3 weeks postlesion. Note the extensive loss of TH+ cells in the ipsilateral SN and almost complete denervation of TH+ fibres in the ipsilateral corpus striatum.

3. Wash sections in distilled water (two changes) and counterstain with cresyl violet acetate solution for 1–2 min. 4. Rinse with tap water and dehydrate with graded alcohols (70, 90 and 100%), clear in xylene and coverslip with DPX mounting medium (in fume hood). 5. Ferric iron will appear blue; use cresyl violet staining and morphology to identify iron-containing cells. 3.4.4. Stereological Quantification of Nigral Cell Loss

To estimate the remaining number of TH-positive (TH+) neurons within the SN in sham and lactacystin-lesioned animals,

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

505

cell counts should be performed using an unbiased stereological method. For this a three-level fractionator sampling method on a computerized image analysis set-up (Zeiss Axioscope, Carl Zeiss, Gottingen, GER) running Stereo investigator software (v7.0, MBF Bioscience, Chicago, IL, USA) is appropriate, as previously reported by ourselves and others (26). For this: 1. Sample every sixth section throughout the rostro-caudal extent of the SN systematically using an unbiased counting frame (the “optical fractionator”) of known area superimposed on the field of view by the software. 2. Assess section thickness empirically in each tissue section analysed, and use guard zones of 2 μm thickness at the top and bottom of each section. 3. Manually outline the SN at ×2.5 magnification. Counting frames are then systematically distributed with known x and y steps throughout the region from a random starting point. At least 20 counting frames should be sampled per section. All cell counts should be performed under ×40 magnification, from 3 to 5 sections per animal from each treatment group. 4. Calculate the coefficients of error (CE) according to the procedure of West and colleagues (27) only accepting values <0.10 (27). Additional cell counts should be performed in parallel for NeuN-positive (NeuN+) cells in the SN using the method described above. 3.5. Correlation Analysis

In order to investigate how dynamic changes in the brain measured in vivo by MRI are related to behavioural measures and pathological changes measured by post-mortem immunohistochemistry and histology, it is desirable to run a correlation analysis. These types of analysis are extremely useful as, for example, if one finds a strong correlation (which may be negative or positive) between, say, volume measurements by MRI and motor behavioural deficits, it would indicate that MRI volume measurements could be predictive of the degree of functional impairment in individual animals. Thus, MRI volumes could act as a surrogate marker of behavioural measures as previously reported (26). At the simplest level, this can take the form of a standard Pearson product moment model. This may be used to find a correlation between at least two continuous variables. The value for a Pearson’s may fall between 0.00 (no correlation) and 1.00 (perfect correlation). Other factors such as group size will determine if the correlation is significant. Generally, r values >0.80 are considered to indicate a very strong correlation. Note that the r value indicates the degree of similarity between the two independent measurements, whereas the P value indicates if this link is significant. A more complex method may involve a multiple

506

Vernon and Modo

regression analysis, which may be especially useful when multiple time points and groups are involved in the study (e.g. (28)). Both methods may be carried out using generally available statistical software such as SPSS (v16.0). In this instance, we describe the methodology for a standard Pearson product moment correlation analysis. For this: 1. Create an SPSS data file containing your variables. It may be easier to enter all your variables: MR data, behaviour and histology into one data file to facilitate running several different Pearson correlations. 2. Also include in your variables “group” and “animal.” In the group variable, for each measure (MR, behaviour, histology) assign each animal a number to separate the two, for example, saline-injected animals could be labelled “1” and lactacystin-injected animals “2.” This allows for correlations to be tested between and within groups. 3. From the “analyze” drop-down menu, select “correlate, bivariate.” This displays a sub-menu in which you can select which variables you wish to correlate. Variables may be correlated one at a time or all together at once. To add a variable, simply select it from the left column and click the arrow button. Repeat until all the variables to be analysed are selected. 4. In the box at the bottom of the menu, ensure the box marked “Pearson” is ticked, along with “two-tailed” and “flag significant correlations.” Click the button marked “options” and check the “means and standard deviations” box in the “statistics” area so descriptive statistics are displayed in the output, although these options are not needed to run the actual Pearson’s correlation. Click on continue and click on OK to run the correlation. 5. Once the analysis is complete an output matrix showing descriptive statistics, the mean, standard deviation and number of cases for each variable, is generated. Variables are arranged in a matrix such that where their columns/rows intersect the statistic informs about the interaction between the variables. Three pieces of information are provided in each cell – the Pearson correlation, the significance, and number of cases. Where a variable interacts with itself, the correlation will obviously have r= 1.00. No significance will be given in these cases. Any significant interactions will be highlighted as ∗ (significant to 0.05 level) or ∗∗ (significant to the 0.01 level). An example correlation is shown in Fig. 25.5 between striatal volume measured on MR images and apomorphine rotation.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

507

Fig. 25.5. Correlation analysis (independent of treatment) between striatal volumes as measured from in vivo MR images, TH+ fibre density measured post-mortem and the degree of contraversive rotation induced by apomorphine challenge. a There is an overall robust negative correlation between in vivo striatal volume and the degree of

508

Vernon and Modo

4. Notes 1. In this particular pre-clinical model of PD, the dose of the proteasome inhibitor (lactacystin or similar) is critical. If the dose is too high (>10 μg), there will be substantial general neuronal toxicity and thus it is difficult to ascribe any subsequent MR changes as specifically related to the degeneration of the nigrostriatal system, thus limiting their usefulness. It is highly recommended therefore that an initial dose–response pilot study be carried out to determine an appropriate dose of toxin that produces robust, selective nigrostriatal degeneration. Care should be taken in deciding the route of administration also as outlined in Section 3.1.3. 2. Care should be taken when handling animals after exposure to apomorphine as they may become aggressive. 3. Whilst acquiring MR images, place the slice acquisition box at the base of the olfactory bulb to the top of the cerebellum to ensure comparable sampling between subjects. In addition, note the 90 and 180◦ pulse power settings. If these change substantially between scanning sessions this could affect the quality of your T2 measurements. 4. Ensure any ROI used for volumetric or T2 measurement is saved for later quality control or re-analysis. In particular for T2 measurements where it is hard to identify structures such as the substantia nigra due to the low anatomical resolution of the T2 map, it may be desirable to trace an ROI comprising a box of a defined voxel size (e.g. 5 × 5) using the anatomical guides as suggested in Section 3.3.2, save it and re-apply it to all subjects in each group for consistent measurements.

 Fig. 25.5. (continued) contralateral rotation when both groups are compared (solid black line). When each group is analysed separately, however, no significant correlation between striatal volume and apomorphine rotation for saline (dashed line) or lactacystin-injected animals (dotted line) is observed. This is consistent with an allor-nothing effect, which one might expect from this behavioural analysis. Interestingly, these data also show a narrow threshold in striatal volume beyond which, significant rotational behaviour is apparent (approx. 30 mm3 ). b An identical pattern is observed for TH+ fibre density in the striatum correlated to the degree of apomorphine rotation. Again note the lack of correlation within groups and the narrow threshold at which point significant rotational behaviour occurs. c An overall positive correlation is observed between in vivo striatal volume and of TH+ fibre density postmortem, suggesting that the change in striatal volume in vivo is related to the density of TH+ fibres remaining.

Non-invasive MR Imaging of Neurodegeneration in a Rodent Model of PD

509

5. When carrying out Prussian blue staining, thorough washing of sections is vital to remove unbound dye and allow robust identification of iron deposits. 6. To ensure optimal DAB staining of free-floating sections make sure that the sections are in a suitable volume to allow sufficient exposure to antibodies and other reagents. For 40μM thick sections, 1–1.5 ml in a 12-well plate is acceptable. Always prepare ABC reagent 30 min before use and DAB immediately before use to get best results. Ensure sections are exposed to DAB for approximately the same time before thorough washing in PBS to ensure comparable staining intensity. 7. In using the optical fractionator probe to count the numbers of TH+ cells in the SN, a grid size of 170 × 170 μm using a counting frame of 100 × 100 μM with an average of 20–25 sampling sites at ×40 magnification (29) results in consistent Gunderson CE of less than 0.1 as recommended by (27).

Acknowledgements Our studies are supported by a grant from the Edmond J. Safra philanthropic foundation, which we thank for their generous financial assistance. We also thank the British Heart Foundation for supporting the 7 T MRI scanner (Preclinical Imaging Unit, Kings College, London). ACV is supported by an Edmond J Safra fellowship. MM is supported by an RCUK fellowship. References 1. Fearnley, J. M., Lees, A. J. Ageing and Parkinson’s disease: Substantia nigra regional selectivity. Brain 1991;114(Pt 5):2283–2301. 2. Spillantini, M. G., Schmidt, M. L., Lee, V. M., Trojanowski, J. Q., Jakes, R., Goedert, M. Alpha-synuclein in Lewy bodies. Nature 1997;388:839–40. 3. Brooks, D. J. The role of structural and functional imaging in parkinsonian states with a description of PET technology. Semin Neurol 2008;28:435–445. 4. Reetz, K., Gaser, C., Klein, C. et al. Structural findings in the basal ganglia in genetically determined and idiopathic Parkinson’s disease. Mov Disord 2009;24:99–103. 5. Camicioli, R., Gee, M., Bouchard, T. P. et al. Voxel-based morphometry reveals

extra-nigral atrophy patterns associated with dopamine refractory cognitive and motor impairment in parkinsonism. Parkinsonism Relat Disord 2009;15:187–195. 6. Lewis, M. M., Smith, A. B., Styner, M. et al. Asymmetrical lateral ventricular enlargement in Parkinson’s disease. Eur J Neurol 2009;16:475–481. 7. Brar, S., Henderson, D., Schenck, J., Zimmerman, E. A. Iron accumulation in the substantia nigra of patients with Alzheimer disease and parkinsonism. Arch Neurol 2009;66:371–374. 8. Kosta, P., Argyropoulou, M. I., Markoula, S., Konitsiotis, S. MRI evaluation of the basal ganglia size and iron content in patients with Parkinson’s disease. J Neurol 2006;253:26–32.

510

Vernon and Modo

9. Martin, W. R., Wieler, M., Gee, M. Midbrain iron content in early Parkinson disease: A potential biomarker of disease status. Neurology 2008;70:1411–1417. 10. Wallis, L. I., Paley, M. N., Graham, J. M. et al. MRI assessment of basal ganglia iron deposition in Parkinson’s disease. J Magn Reson Imaging 2008;28:1061–1067. 11. Pavese, N., Brooks, D. J. Imaging neurodegeneration in Parkinson’s disease. Biochim Biophys Acta 2009;1792:722–729. 12. Jenner, P. Functional models of Parkinson’s disease: A valuable tool in the development of novel therapies. Ann Neurol 2008;64(Suppl 2):S16–S29. 13. Sanchez-Pernaute, R., Brownell, A. L., Jenkins, B. G., Isacson, O. Insights into Parkinson’s disease models and neurotoxicity using non-invasive imaging. Toxicol Appl Pharmacol 2005;207:251–256. 14. Vernon, A. C., Johansson, S. M., Modo, M. M. Non-invasive evaluation of nigrostriatal neuropathology in a proteasome inhibitor rodent model of Parkinson’s disease. BMC Neurosci 2010;11:1. 15. McNaught, K. S., Bjorklund, L. M., Belizaire, R., Isacson, O., Jenner, P., Olanow, C. W. Proteasome inhibition causes nigral degeneration with inclusion bodies in rats. Neuroreport 2002;13:1437–1441. 16. Miwa, H., Kubo, T., Suzuki, A., Nishi, K., Kondo, T. Retrograde dopaminergic neuron degeneration following intrastriatal proteasome inhibition. Neurosci Lett 2005;380:93–98. 17. Deumens, R., Blokland, A., Modeling Parkinson’s, P. J. disease in rats: An evaluation of 6-OHDA lesions of the nigrostriatal pathway. Exp Neurol 2002;175: 303–317. 18. Fornai, F., Lenzi, P., Gesi, M. et al. Fine structure and biochemical mechanisms underlying nigrostriatal inclusions and cell death after proteasome inhibition. J Neurosci 2003;23:8955–8966. 19. Pan, T., Kondo, S., Zhu, W., Xie, W., Jankovic, J., Le, W. Neuroprotection of rapamycin in lactacystin-induced neurodegeneration via autophagy enhancement. Neurobiol Dis 2008;32:16–25. 20. Zhu, W., Xie, W., Pan, T. et al. Prevention and restoration of lactacystin-induced nigros-

21. 22.

23.

24.

25.

26.

27.

28.

29.

triatal dopamine neuron degeneration by novel brain-permeable iron chelators. Faseb J 2007;21:3835–3844. Paxinos, G., Watson, C. The Rat Brain in Stereotaxic Co-ordinates, 6 ed. San Diego, CA: Academic Press; 2007. Modo, M., Stroemer, R. P., Tang, E., Veizovic, T., Sowniski, P., Hodges, H. Neurological sequelae and long-term behavioural assessment of rats with transient middle cerebral artery occlusion. J Neurosci Methods 2000;104:99–109. Ungerstedt, U., Arbuthnott, G. W. Quantitative recording of rotational behavior in rats after 6-hydroxy-dopamine lesions of the nigrostriatal dopamine system. Brain Res 1970;24:485–493. Sled, J. G., Zijdenbos, A. P., Evans, A. C. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 1998;17(1): 87–97. Vernon, A. C., Zbarsky, V., Datla, K. P., Croucher, M. J., Dexter, D. T. Subtype selective antagonism of substantia nigra pars compacta Group I metabotropic glutamate receptors protects the nigrostriatal system against 6-hydroxydopamine toxicity in vivo. J Neurochem 2007;103:1075–1091. Vernon, A., Johansson, S. J., Modo, M. M. Non-invasive evaluation of nigrostriatal neuropathology in a proteasome inhibitor rodent model of Parkinson’s disease. BMC Neuroscience 2009;11:1. West, M. J., Slomianka, L., Gundersen, H. J. Unbiased stereological estimation of the total number of neurons in the subdivisions of the rat hippocampus using the optical fractionator. Anat Rec 1991;231: 482–497. Modo, M., Beech, J. S., Meade, T. J., Williams, S. C., Price, J. A chronic 1 year assessment of MRI contrast agentlabelled neural stem cell transplants in stroke. Neuroimage 2009;47(Suppl 2): T133–T142. Koprich, J. B., Reske-Nielsen, C., Mithal, P., Isacson, O. Neuroinflammation mediated by IL-1beta increases susceptibility of dopamine neurons to degeneration in an animal model of Parkinson’s disease. J Neuroinflammation 2008;5:8.

Chapter 26 Detecting Amyloid-β Plaques in Alzheimer’s Disease Christof Baltes, Felicitas Princz-Kranz, Markus Rudin, and Thomas Mueggler Abstract One of the major neuropathological changes characteristic of Alzheimer’s disease (AD) is deposits of beta-amyloid plaques and neurofibrillary tangles in neocortical and subcortical regions of the AD brain. The histochemical detection of these lesions in postmortem brain tissue is necessary for definitive diagnosis of AD. Methods for their in vivo detection would greatly aid the diagnosis of AD in early stages when neuronal loss and related functional impairment are still limited and would also open the opportunity for effective therapeutic interventions. Magnetic resonance imaging (MRI) theoretically provides the spatial resolution needed to resolve amyloid-β plaques. Although currently limited for clinical applications due to unfavorable long acquisition times, MRI has been used to visualize Aβ plaques in AD mouse models. The ability to detect amyloid-positive brain lesions in vivo using non-invasive imaging would allow to track disease progression and to monitor the efficacy of potential therapies in disease-modifying studies using transgenic models resembling AD pathology. Here, we provide MRI protocols for in vivo (mouse) and ex vivo (AD tissue samples) amyloid plaque imaging and the procedure for correlating these with thioflavin-S and iron-staining histology. Current challenges and limitations are discussed. Key words: Beta-amyloid, Aβ plaques, amyloid precursor protein (APP), Alzheimer’s disease (AD), magnetic resonance imaging (MRI), MRI contrast, contrast agent, radiofrequency (RF) coil, histology.

1. Introduction Senile amyloid-beta (Aβ) plaques representing the cardinal feature in Alzheimer’s disease (AD) accumulate over the course of decades in the brains of patients with AD. A fundamental pillar of the amyloid hypothesis of AD is that the deposition of Aβ precedes and induces the neuronal abnormalities that underlie M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_26, © Springer Science+Business Media, LLC 2011

511

512

Baltes et al.

dementia (1) conceptually challenged by the hypothesis that alterations in axonal trafficking and morphological abnormalities precede and lead to senile plaques. A general consensus has been reached recently, insofar that disordered metabolism of Aβ protein is central to the pathological cascade that ultimately leads to clinical AD. Consequently, Aβ reduction in humans is a major therapeutic objective, and transgenic mouse models of AD allow controlled study of this process. Nuclear imaging methods, such as positron emission tomography (PET), hold great promise to follow senile plaque onset and progression in AD patients and thereby allowing evaluating potential therapeutic effects. The 11 C-labeled Pittsburgh compound-B (11 C-PIB) is currently in use in human studies at a number of academic sites, and commercial development of 18 F-PIB is in progress. This should make the tracer even more accessible in coming years. PET imaging, however, suffers from a limited spatial resolution providing a semiquantitative plaque load value. Moreover, preclinical micro-PET studies, evaluating the labeling of amyloid plaques in transgenic mouse models of AD with PIB, have not been successful to date (2). Magnetic resonance imaging (MRI), on the other hand, theoretically provides the spatial resolution needed to resolve single neuritic Aβ deposits. Although currently limited for clinical applications due to unfavorable long acquisition times, MRI has been used to visualize Aβ plaques in AD mouse models. The additional value to follow non-invasively the course of Aβ deposit pathology at the single plaque level has been recently highlighted by the finding that plaques form extraordinarily quickly, over 24 h with progressive neuritic changes ensuing, leading to increasingly dysmorphic neurites over the next days to weeks (3). Both mouse and human possess compact and diffuse plaques. The mouse compact plaques display a fibrillar internal structure radiating outward from a central core that is not visible in human compact plaques. Thus some authors have classified the thioflavine S-positive deposits as cored plaques (4). These plaque structures are significantly larger in AD mice (80–100 μm), being on average about twice the diameter of human classical plaques (5). See Fig. 26.1 for the histology of human and PS2APP transgenic mouse brain sections. Histological data from previous studies using postmortem human AD tissue (6) have confirmed focal iron load in Aβ plaques. Ferric iron (Fe+3 ) in brain tissue acts as a natural contrast agent causing faster proton T2 and T2 ∗ relaxation. It is important to note here that the cause of plaque contrast in mouse model tissue may partly differ from the contrast mechanism suggested for human plaques. The iron content revealed by Perl’s Prussian Blue staining seems to be much lower in AD mouse brain tissue than in the human AD tissue samples suggesting a different cause of the hypo-intense appearance on T2 - or T2 ∗ -weighted MR images in

Detecting Amyloid-β Plaques in Alzheimer’s Disease

513

Fig. 26.1. Histology of human and PS2APP transgenic mouse brain sections. Brain sections from PS2APP mice (a, e) and from human AD (b–d, f–h) specimens were double stained with thioflavine S and an antibody specific for the amino terminal amino acids 2–8 (BAP-2) in order to differentiate between the three plaque types. Arrowheads indicate diffuse plaques (thioflavine S negative and BAP-2 positive), arrows indicate compact plaques (thioflavine S positive and BAP2 positive). a Typical example of a cored plaque is given in c and g (core: thioflavine S positive and BAP-2 positive; corona: thioflavine S negative and BAP-2 positive). Diffuse plaques were found occasionally in transgenic mice and more frequently in human brain. Compact plaques were rare in humans, but predominate in PS2APP mice. Human vessels with CAA were visualized by both methods (d, h) (reprinted from (4) Fig. 26.1).

the absence of ferric iron (7). The ability to delineate Aβ plaques with MRI has been demonstrated ex vivo (some authors refer to this as in vitro) with human AD brain samples (7–9) or samples derived from AD mouse models (10–14). Visualization of plaques using MRI in the living AD mouse (15–18) puts additional challenges on the MR acquisition due to brain motion originating, i.e., from respiration and blood pressure waves. Furthermore, the total scan time is limited by the total anesthesia time. Jack and colleagues (15) in their pioneering work were able to visualize plaques with diameters as small as 50 μm based on a native contrast as hypo-intense structures on T2 - or T2 ∗ -weighted images in aged AD mice at a scanning time of less than 2 h, i.e., approaching a time frame acceptable for in vivo imaging. Further strategies to improve the specificity in Aβ plaque imaging are based on labeling strategies using positive T1 contrast agents (10, 19) or by detecting the 19 F MR signal in the brain of living APP transgenic mice after intravenous injection of benzene derivative (FSB) compound (20). All these studies have used genetically engineered mouse models of AD that either express transgenic mutated amyloid precursor protein (APP) alone or in combination with transgenic mutated presenilin (PS1 or PS2) (21) leading to plaque formation. As outlined above the overall aim is to follow Aβ deposit pathology at the single plaque level under in vivo conditions. Following a description of the most commonly used AD mouse lines and their preparation for MR imaging (Sections 2.1 and 3.1), the chapter’s main focus is on MR acquisition hardware (covering main magnetic field strength, gradient system, and radiofrequency coils), MR contrast agents, as well as protocols for the

514

Baltes et al.

in vivo visualization of single Aβ plaques (Sections 2.2 and 3.2). In order to provide a comprehensive overview the subchapters include also protocols for in vitro MRI acquisition. The method part is completed by giving a step-by-step protocol for brain sample preparation for in vitro MRI and histology (Section 3.4) and by discussing approaches of correlating MR images with histological sections stained for Aβ deposits (Section 3.5). Having motivated our aim and outlined the chapter we now want to draw the readers’ attention on the three main aspects which will serve as a red line throughout the entire chapter: • How to achieve the needed resolution with a sufficiently high signal-to-noise ratio (SNR). • How to achieve optimal contrast of Aβ plaques. • How to improve specificity and thereby avoid false positives. For a more detailed discussion on the topic of targeting strategies to improve the specificity in Aβ plaque MR imaging, the interested reader is referred to recent work that extensively cover different labeling and delivery approaches for in vivo imaging in AD mouse models (22, 23).

2. Materials 2.1. In Vivo Models – Mouse Models of Cerebral Aβ Amyloidosis 2.1.1. Nomenclature and Availability of Strains

The identification of familial forms of AD (FAD mutations) provided potential means of generating rodent models of the disease. The last 12–15 years have seen the emergence of a significant number of transgenic mouse models of cerebral Aβ amyloidosis (Table 26.1). As a central resource for archiving and distributing AD models the Alzheimer’s disease mouse model repository (ADMMR) was established at The Jackson Laboratory (TJL) (www.jax.org/jaxmice/models/alzheimers). Current models carrying mutations in 13 different genes directly relevant to AD are available, including apolipoprotein E (ApoE), amyloid-beta precursor protein (APP), and presenilins (PS1 and PS2). A comprehensive overview on models and strains and can also be found on http://www.alzforum.org/res/com/tra/.

2.1.2. Study Design – Choice of Appropriate Strain and Cohort

Single- and double-transgenic mice provide a useful model of AD sharing the cardinal pathological feature of extracellular Aβ containing senile plaques. It has to be noted, that in terms of severity, distribution, and onset, they do show a heterogeneous picture.

APP695 (Swedish)

APP751 (Swedish)

APP695 (Swedish), PS1 (M146L)

APP751 (Swedish), PS2 (N141I)

APPSwe (2576)

APP23

PSAPP

PS2APP

a In vitro and in vivo MRI using native contrast.

APP770 (Swedish) and APP695 (London, V642I)

Transgenea (FAD mutation) APP695,751,770 (V717F)

APPLd

PDAPP (line109)

Line

Mouse line genotype

(39)

(38)

(37)

(36)

(35)

Original report (34)

4–6 months

8–9 months

12 months

12 months

Plaques (age of onset) 6–9 months

C57Bl/6 × DBA/2.

C57Bl/6 J

Contact: Laurence Ozmen Laurence.ozmen@ roche.com

Contact Matthias Staufenbiel at Novartis; matthias. [email protected]

Not reported

(17, 28)

(11)

(25)

C57Bl/6 × SJL

Plaque MRI reference(s) Not reported

(16)

B6; SJL-Tg(APPSWE)2576 kh www.tacona.com (#001349) or backcrossed: 129S6 (#002789)

Not commercially available; contact Dora Games or Dale Schenks at Elan Pharmaceutical

Source/availability

FVB/N

Background (origin) C57Bl/6 × DBA/2

Plaque phenotype and strain availability

Table 26.1 Phenotype and availability of APP-based mouse models of AD

Detecting Amyloid-β Plaques in Alzheimer’s Disease 515

516

Baltes et al.

A careful study design is therefore recommended comprising not only the appropriate FAD model but also furthermore consideration needs to be given to strain background, age, and sex of the population, as well as the correct controls. 1. Aβ deposition in single APP transgenic mice is seen from around 9 months of age. 2. On average, substantial plaque load is observed between 12 and 24 months of age. Ceiling effects have been reported, thus a given upper plaque load is reached in aged mice. 3. Double-transgenic mice (APP combined with PS mutations) show accelerated Aβ deposition with an onset at around 4–6 months of age. 4. In both types of models (APP and APP/PS), although governed by specific promoters that predominantly drive expression to the cortex and hippocampus, Aβ deposition spreads throughout the brain, including regions with low transgenic APP expression. 5. The deposition is initially as dense core plaques; diffuse Aβ only appears in the later stages of the animal’s life. 6. Mouse littermates should be used as controls, as they do not only provide the correct strain background but also control for environmental conditions. 7. Control for differences between male and female. 2.1.3. Materials for In Vivo MRI Preparation

R 1. Isoflurane and gaseous carrier or other non-terminal anesthesia (see Note 1).

2. Mouse cradle (see Note 2) with an in-built warm water circulating system or warm air and with in-built gas supply (for gas anesthesia), tooth-bar and/or holder for ear-bars. 3. Trigger device based on either electrocardiogram (disposable, adhesive electrodes) or respiratory monitoring device (e.g., SA Instruments, Stony Brook, NY). 4. Rectal temperature probe (see Note 3). 5. Catheter with a 30 G needle (0.3 mm × 13 mm) and R PE tubing (e.g., BD Microlance, Smith Medical). Portex Optional: 6. Cold light source (e.g., Intralux, Harvard Apparatus). 7. Small animal ventilator (e.g., KTR5, Harvard Apparatus). 8. Neuromuscular blocking agent, e.g., gallamine triethiode or pancuronium bromide (Sigma-Aldrich). 2.2. MR Acquisition

Visualization of small structures in the order of 50 μm puts high demands on the sensitivity of the MR signal detection. In order to increase the inherent SNR of the MR acquisition two approaches

Detecting Amyloid-β Plaques in Alzheimer’s Disease

517

have been pursued: increasing MR signal intensity by moving to higher magnetic field strength or/and decreasing noise contributions by developing radiofrequency (RF) coils specifically dedicated for mouse brain imaging. Besides these approaches specific labeling of Aβ plaques using external contrast agents has been successfully demonstrated in the living mouse using standard MR hardware operating at the proton resonance frequency (24). An alternative approach has been followed by Higuchi et al. (20) using a 19 F-labeled benzene derivative (FSB). 2.2.1. Main Magnetic Field Strength and Gradient System

2.2.2. Radiofrequency Coils

Essential for high spatial resolutions at small field-of-views are 1. High main magnetic field strengths. 2. State-of-the-art gradient systems (in the order of 400 mT/m at 80 μs rise time). MR microscopy of human ex vivo specimen (8) and of histological sections of the rodent brain (7) was performed at 7 T to investigate and optimize the underlying contrast mechanism between Aβ plaques and brain parenchyma. While in the ex vivo case scan time is in general unrestricted and the acquisition can be optimized for contrast, the scan time for in vivo Aβ plaque imaging is limited by the duration of anesthesia and the physiological stability of the animal resulting in reasonable scan times of about 1 h. Individual Aβ plaques in transgenic mouse models were visualized in vivo both at 7 T (16) and at 9.4 T (15). While increasing the MR signal by moving to higher main magnetic field strength is a promising option, the effective gain in SNR is limited due to increasing T1 relaxation times and decreasing T2 /T2 ∗ relaxation times for increasing field strengths making adaptations of the MR imaging parameters necessary. Depending on the specific application of interest different radiofrequency coils are commercially available or have to be specifically designed. Besides technical hallmarks, also availability and costs are summarized in the overview listed in Table 26.2. The homogeneous RF field of volume resonators, for example, makes them suitable for both histological measurements of ex vivo samples and anatomical in vivo acquisitions. Furthermore they can be used for quantitative mapping of T1 and T2 relaxation times. However, due to their large volume coverage they provide only poor SNR which is usually compensated for by signal averaging resulting in long acquisitions times, which might be acceptable for ex vivo imaging, but is a limiting factor in in vivo applications. In order to visualize individual Aβ plaques in transgenic mouse models a volume resonator was successfully applied at high-field strength of 9.4 T (25) while at 7 T a cross-coil setup with a surface receive coil (16) was used to achieve sufficient

++

Hom

Large (=10 mm)

+++

++

(10)

SNR

B1 -field

Sensitive volume

Availability

Costs

References

a See Note 4.

+

+

Operation mode

(7, 26)

Small (two parallel 24×30 mm2 continuous copper layers)

Hom

+++

Transceive

Transceive

Nucleus

1H

Histological coil

1H

Application Ex vivo

Volume resonator

(25)

++

+++

Large (ø=25 mm)

Hom

+

Transceive

1H

In vivo

Volume resonator

(20)

+++

+

Large (ø=25 mm)

Hom

++

Transceive

19 F

Volume resonator

Inhom

+++

quadrature, transceive

1H

Surface coil

(16)

+++

+++

(15)

++

+

Small Small (ø=1 cm (ø=20 mm) per element)

Inhom

+++

Cross-coil

1H

Surface coil

(33)

+++++

++

Small (20×17 mm2 per element)

Inhom

++++

Cryogenic, quadrature, transceive

1H

Surface coila

Table 26.2 Summary of radiofrequency coils used for both ex vivo (human and rodent brain samples) and in vivo imaging of Aβ plaque depositions

518 Baltes et al.

Detecting Amyloid-β Plaques in Alzheimer’s Disease

519

SNR. For further SNR enhancement Jack et al. (15) combined high magnetic field strength (9.4 T) with dedicated detection hardware (quadrature surface transceive coil) designed for mouse brain imaging. However, increasing SNR by using RF surface coils comes at the expense of an inhomogeneous B1 -field distribution. This leads, in the case of signal reception, to descending detection sensitivity for increasing distances from the coil. In the case of a surface transceive coil, both signal excitation and reception are inhomogeneous and require substantial adaptations of the MR pulse sequence (based on adiabatic RF pulses) to account for this restriction (15). While volume resonators and standard receive surface coils are commercially available, dedicated quadrature RF transceive coils as presented in (15) require advanced knowledge in RF coil design. Detecting MR signal from plaque-bound 19 F-labeled FSB compounds requires detection hardware tuned to the fluorine resonance frequency. For the examination of ex vivo histological sections dedicated to histological coils have been developed which allow for a direct comparison of MR microscopy images and immunohistological stainings (26). 2.2.3. Targeted MR Contrast Agents

Recently, the application of MR contrast agents specifically designed for labeling of Aβ plaques has been reported. As Gadolinium (Gd)-based contrast agents lead to a local shortening of T1 relaxation times, Gd-labeled plaques can be detected using T1 -weighted proton (gyromagnetic ratio of 1 H: γ =42.576 MHz/T) imaging sequences using conventional RF coils. In general, contrast agents targeting Aβ plaques in vivo need to fulfill certain criteria: (i) high in vivo stability, (ii) nondestructive crossing of the blood–brain barrier (BBB) after i.v. administration, (iii) high affinity to bind to Aβ plaques, and (iv) generation of high tissue contrast to be detectable by MRI. Poduslo and colleagues (24) developed the Gd-based contrast agent: putrescine–gadolinium–amyloid-β peptide (PUT–Gd–Aβ). After i.v. administration of the compound into APP-PS1 mice, the animals were sacrificed, and the MR detectability and the selective labeling of Aβ plaques have been thoroughly confirmed in subsequent experiments. Although the in vivo application has not been demonstrated yet, the approach holds great potential for visualizing the fate of individual plaques in longitudinal studies. An alternative approach has been followed by Higuchi et al. (20) using a 19 F-labeled benzene derivative (FSB). MR imaging of 19 F-nuclei is advantageous as its sensitivity (83%) is almost as high as for protons, and 19 F-background signal is expected to be very low as biological tissue contains essentially no 19 F. However, as conventional MR imaging systems and hardware are designed for proton imaging dedicated MR components for

520

Baltes et al.

excitation and detection of the 19 F signal (gyromagnetic ratio of 19 F: γ =40.075 MHz/T) need to be available. 2.3. Image Analysis Tools

2.4. Sample Preparation for Ex Vivo MRI and Histological Validation

Initial attempts to visualize Aβ plaques focused on the MR acquisition protocol, accordingly image analysis tools were mainly based on commercially available software provided by the vendor of the MR system or on generic image analysis software. More sophisticated analysis tools are based on customized software written in, e.g., IDL (Research Systems, Inc., Boulder, CO, USA) which are available only to a small community. A more detailed description of some implemented functionalities is given in Section 3.3. 1. Sodium pentobarbital 2. Perfusion pump for perfusion-fixation 3. PBS, pH 7.4 (heparin 1 U/g body weight can be added) 4. 4% paraformaldehyde in PBS 5. 30% sucrose in 1× PBS 6. 10% dimethyl sulfoxide (DMSO) in 1× PBS 7. 20% glycerol/2% dimethyl sulfoxide in 0.1 M phosphate buffer 8. GdDOTA (0.1 mmol in 5% formalin solution) R R (perfluorinated polyether) or Fluorinert (flu9. Fomblin orocarbon) or agar

10. Freezing microtome or cryostat 11. Gelatin-coated slides (see Note 5) and coverslips R 12. Xylene or CitriSolv

13. Congo red solution (2.5 g of Congo red/250 ml of 50% ethyl alcohol) 14. 1% thioflavin-S (in distilled water, filtered through Whatman no. 1) 15. Lithium carbonate solution 16. Fluorescence microscope 17. 1–2% potassium ferrocyanide (Perl reagent for Prussian blue [PB] staining in 1% HCl and 1% Triton X-100 in distilled water) 18. DAB solution (0.5 mg/ml 3.3 -diaminobenzidine tetrahydrochloride (Sigma) and 2 μl/ml of 30% hydrogen peroxide (Sigma) in 0.05 M Tris-HCl, pH 7.6) R kit (Vector Laboratories) 19. MOM

20. Aβ-antibodies 21. Normal serum (rat, goat, donkey, etc.) to block nonspecific antibody binding

Detecting Amyloid-β Plaques in Alzheimer’s Disease

521

3. Methods 3.1. Animal Preparation for In Vivo MRI

3.1.1. Freely Breathing: With Optional Gated Acquisition

Structural brain imaging of mice at high spatial resolution (<100 μm isotropic) requires a stable preparation of the animal. During the inherently long acquisition time, motion due to respiration and/or blood pressure waves leads to compromised spatial resolution. Proper, reproducible positioning of the animal’s head can be achieved by using stereotactic fixation. Furthermore respiratory artifacts can be minimized by intubation, artificial ventilation, and paralysation using a neuromuscular blocking agent. Two different non-invasive experimental protocols are provided here allowing a longitudinal study design. Several authors reported on aged (18–30 months) transgenic APP animals being physiologically fragile and thus not able to withstand excessively long anesthesia sessions. In general, though, scan time is optimally limited to 1 h (see Note 6). R 1. Mice are anesthetized using 1.0–2.0% isoflurane in a mixture of O2 /N2 O (usually 30:70) or O2 /air (usually 50/ 50–30/70) and positioned in a cradle for imaging.

2. The mouse is positioned on the cradle with its abdomen/thorax on top of a respiration pressure transducer of a respiratory monitoring device. 3. MR compatible ECG leads (e.g., SA Instruments Inc. Stony Brook, US) are fixed to the left forelimb and right hind limb to monitor the electrocardiogram. 4. Body temperature is monitored by a rectal probe (e.g., fiberoptic QUASYS AG, Cham, Switzerland or standard) and maintained at 37±0.5◦ C using an in-built warm water circulation system or warm air. 5. The tail vein is cannulated with a 30 G needle (0.3 mm × 13 mm) connected to a PE tube for on-line infusion of e.g. a contrast agent. 3.1.2. Artificial Ventilation and Stereotactic Fixation

R 1. Mice are initially anesthetized with 2.5–3% isoflurane (same carrier as in Section 1).

2. Endotracheal intubation is performed with a polyethylene tube (ID/OD 0.58/0.96 mm) under direct visualization of the larynx using an external cold light source. 3. Mice are placed on a cradle in prone or supine position with the intubation tube connected to the Y-piece of a small animal ventilator (e.g., KTR5, Harvard Apparatus) and artificially ventilated (90 bpm, stroke volume 0.4 ml). R level is then reduced to 1.0–2.0%. 4. Isoflurane

5. Ear bars are mounted for reproducible positioning and reduction of motion artifacts.

522

Baltes et al.

6. The tail vein is cannulated and body temperature is monitored throughout the MR experiment following Section 3.1.1. steps 4–5. 7. Body temperature is monitored throughout the MR experiment. 8. Shortly before MR acquisition, a neuromuscular blocker is injected (e.g., gallamine triethiode 10–30 mg/kg or pancuronium bromide, Sigma-Aldrich, Steinheim, Germany) for transient paralysis (see Note 7). Further experimental information on how to perform in vivo MRI in mice with focus on physiological monitoring (devices) can be found in (27). 3.2. MR Acquisition 3.2.1. MR Sequences for Optimal Contrast of Aβ Plaques

The visualization of individual Aβ plaques requires reasonable contrast between the plaques and the surrounding tissue (e.g., brain parenchyma). For this reason the MR properties of Aβ plaques, such as T1 , T2 , T2 ∗ relaxation times and proton density, need to be known. Histological investigations revealed increased focal iron deposits in Aβ plaques resulting in a shortening of T2 and T2 ∗ relaxation times. According to these findings, Aβ plaque imaging was performed using T2 -weighted spin echo (SE) or T2 ∗ weighted gradient echo (GE) sequences (Table 26.2). While the operation of conventional MR hardware (volume resonator or cross coil setup) allows for applying standard T2 -weighted SE (25) or T2 ∗ -weighted GE (16) sequences, respectively, the application of dedicated detection devices requires the development of advanced MR pulse sequences to take full advantage of the increased SNR (15, 28). Only the incorporation of adiabatic RF pulses and the reduction of the field-of-view by selective refocusing pulses enabled the in vivo application of the T2 -weighted SE sequence for Aβ plaque imaging in combination with the dedicated quadrature transceive coil. In a comprehensive study using ex vivo specimen, the MR properties of Aβ plaques were investigated by Chamberlain and colleagues (28) revealing similar values for T1 and proton density of cortical plaques and normal cortical tissue, while differences in T2 and T2 ∗ relaxation times were observed supporting the histological findings. Accordingly, a multi-echo GE sequence with susceptibility weighting was found to provide optimal contrast for ex vivo imaging when quantitatively comparing various MR sequences (Figs. 26.2 and 26.3). While large susceptibility differences at air-to-tissue and/or fat-to-skull interfaces make GE sequences impractical for in vivo applications, a multiple asymmetric spin echo (mASE) sequence holds promise for well defined in vivo visualization of Aβ plaques (Table 26.3).

Detecting Amyloid-β Plaques in Alzheimer’s Disease

523

Fig. 26.2. Magnified images from all nine sequences in a 9-month-old ex vivo APP/PS1 brain. The images are ordered by increasing plaque CNR. The multi-echo imaging sequences all produced increased plaque CNR compared to their single-echo counterparts. Similarly, all SWI sequences produced increased plaque CNR compared to the magnitude-only counterparts (reprinted from (28) Fig. 3).

Fig. 26.3. Representative in vivo images from the GRE, SE, and mASE sequences in a 12-month-old APP/PS1 mouse. a The GRE image showed cortical signal loss due to the susceptibility mismatch interfaces at TE – 23 ms. Sinuses inferior to the image border also caused a rippling artifact in the image. Since plaque contrast is maximal near 40 ms, this effect makes in vivo plaque imaging with GRE-based sequences impractical. b The SE sequence that was previously demonstrated to visualize plaques in vivo. c The mASE sequence provides increased SNR and image quality compared to the SE sequence (reprinted from (28) Fig. 6).

3.2.2. Targeted Imaging of Aβ Plaques

Targeted labeling of Aβ deposits in the brain using a Gd-based compounds (10) leads to shortening of T1 relaxation times. Accordingly, MR sequence parameters are adjusted to acquire T1 weighted images (Table 26.4). In the case of Aβ plaque labeling using 19 F-based FSB MR sequence parameters are adjusted not only for T1 weighting but also to excite and receive the

Quad. transceive surface coil

(28)

RF coil setup

References

TR, repetition time; TE, echo time; FOV, field-of-view.

1 h 30 min

Scan time

2D slice thickness: 60 μm

3D volume

Resolution

(7)

Histological coil

1 h 32 min

128

179×179 μm2

60×60×120 μm3

1

128×128

256×256×32

Matrix

# averages

8 23×23 mm2

8

15.4×15.4×3.8 mm3

–

FOV

–

TE effective

–

12 ms

2,500 ms

SE

T2 weighted

# echoes

15 ms

8 ms

659 ms

TR

Echo spacing

GE

Seq. type

Min. TE

Ex vivo

Application

Susceptibility weighted

(15)

Quad. transceive surface coil

1 h 40 min

1

3D volume

60×60×120 μm3

256×96×32

15.4×3.8 mm2

1

–

52 ms

2,000 ms

SE

In vivo

T2 weighted

(25)

Volume resonator

~25 min

4

2D slices, 20–30 thickness, 200 μm gap, 200 μm

78×78 μm2

256×256

20×20 mm2

4

22.45 ms

–

10.6 ms

5,000–6,000 ms

SE

T2 weighted

(16)

Cross coil setup

68 min

1

3D volume

78×156×234 μm3

256×128×64

20×20×15 mm3

1

–

–

8 ms

500 ms

GE

T2 ∗ weighted

Table 26.3 Summary of MR sequence parameters for T2 - and T2 ∗ -weighting applied for ex vivo and in vivo visualizations of Aβ plaques

524 Baltes et al.

Detecting Amyloid-β Plaques in Alzheimer’s Disease

525

Table 26.4 Summary of MR sequence parameters for targeted imaging of Aβ plaques (E, E)-1-fluoro-2,5-bis (3-hydroxycarbonyl4-hydoxy) styrylbenzene (FSB)

Labeling compounds

Putrescine–gadolinium–amyloid-β peptide (PUT–Gd–Aβ)

Application

Ex vivo

In vivo 19 F

Nucleus

1H

Seq. type

T2 -weighted SE

T1 -weighted SE

SE

TR

3,000 ms

400 ms

2,000 ms

TE

100 ms

8 ms

5.5 ms

# echoes

1

32

FOV

16×8×8 mm3

20×20×16 mm3

Matrix

256×128×128

64×64×8

Resolution

62.5×62.5×62.5 μm3

156×156×2,000 μm3

3D volume

3D volume

# averages

1

8

225

Scan time

13 h 52 min

14 h 42 min

RF coil setup

Volume resonator

Volume resonator

References

(10)

(20)

2h

TR, repetition time; TE, echo time; FOV, field-of-view.

MR signal at the 19 F resonance frequency (20). This includes preparation sequences required for power optimization and B0 shimming. 3.3. Image Analysis

Reliable and user-independent image analysis requires standardized and semi-automatic evaluation procedures. For the assessment of individual plaques, changes in image brightness between different acquisitions can be corrected for by normalizing the images with respect to a reference tissue, such as skeletal muscle assumed to remain constant. In a next step, Aβ plaques are identified as signal intensity below a certain threshold which is constant for all acquisitions. Number, size, and percentage of plaque load can then be evaluated using automated algorithms. In recent reports (8, 15, 16, 25) size and number of individual Aβ plaques were evaluated using a manual ROI selection. While this approach is valid for initial proof-of-concept studies, the large number of individual Aβ plaques in the mouse brain and the necessity of a user-independent analysis require an automatic evaluation procedure. In longitudinal studies assessing the fate of individual plaques over an extended period, the image data need to be co-registered to allow for a comparison between the different imaging time

526

Baltes et al.

points. In human fMRI studies, several tools, such as the open source software statistical parametric mapping (SPM; Wellcome Trust Centre for Neuroimaging, London, UK) or FSL (FMRIB Oxford, UK), have been developed to identify activated brain regions. In recent reports such co-registration algorithms have been presented based on mouse and rat standard brain atlases (29). The main critical issue, when carrying out plaque MR imaging based on a native contrast, is the specificity of the observed focal hypo-intense areas in T2 - and T2 ∗ -weighted images. Several sources may account for the “dark spots” appearing on the MR images. For example, areas bearing micro-hemorrhages or heavily myelinated fiber tracts may also appear dark on T2 -weighted images (10). Moreover, under in vivo conditions, vessels orientated radially to/through cortical structures appear as similar hypo-intense structures on coronal image slices. Clearly, for final falsification of the observable hypo-intense areas histological validation and ideally a 1:1 correlation is required which has been recently addressed making use of a histological RF coil design to directly image histological samples (7, 26). In order to allow for an automated, user-independent evaluation texture analysis approaches have been recently investigated in human AD (30). Three main branches of texture analysis can be discriminated: (a) statistical approaches based, e.g., on gray-level co-occurrence matrices or image entropy, (b) signal processing techniques using Fourier or wavelet analysis, and (c) geometrical approaches subdividing images in smaller texture elements. 3.4. Sample Preparation for Ex Vivo MRI and Histological Validation

3.4.1. Mouse Brain Perfusion Fixation

In vitro MRI can be used to optimize MR acquisition parameters as one is not limited in total acquisition time as under in vivo conditions, however keeping in mind different contrast conditions due to preparation and fixation of the tissue. The approach has even been extended by optimizing RF detection (coil setup) for brain sections in order to compare resulting MR images one-toone with the histological slice (26). The second section provides a protocol, which focuses on methods applied to fixed mouse brain sections. Appropriate tissue preservation is fundamental for any of the following steps: systematic sectioning and storage of tissue sections. Additionally, monoclonal mouse antibodies are increasingly employed to detect pre-amyloid and amyloid deposits and their use has led to the development of kits that reduce nonspecific binding of these antibodies to mouse tissue. 1. The mouse has to be deeply anesthetized using, for instance, an intraperitoneal injection of sodium pentobarbital (50 mg/kg) and subsequently placed over a collection container.

Detecting Amyloid-β Plaques in Alzheimer’s Disease

527

2. After exposing the chest cavity a perfusion cannula is then inserted through the left ventricle into the ascending aorta. The right atrium is then cut and the perfusion pump started. 3. The initial perfusion should be performed with 20 ml (see Note 8) of a physiological solution such as 0.1 M sodium/potassium phosphate buffer or PBS, pH 7.4 (heparin 1 U/g body weight can be added). 4. If one would like to avoid perfusion fixation (see Note 9), the brain can then be snap frozen on dry ice at this step. 5. Otherwise the perfusion tubing is transferred to a buffered 4% paraformaldehyde in PBS and perfusion is then resumed with about 50–60 ml of fixative. 6. The brain is removed from the skull and post-fixed for 3–4 h (or overnight at 4◦ C) (see Note 10) in the same fixative. 7. After postfixation, brains are rinsed in 1× PBS and transferred to cryoprotectant solution: (a) 30% sucrose in 1× PBS for 24–48 h at 4◦ C (until brain sunk) or (b) 10% DMSO in 1×PBS at 4◦ C for at least 3 h (max. 15 h). Caution: fast cryoprotection using DMSO is not suited for extreme cold (dry ice). Can be used if immediate transfer to cryostat or microtome sectioning is required. 3.4.2. Mouse Brain Sample Preparation for In Vitro MRI

Whole brain sample In view of the large variety of reported protocols of whole mouse brain sample preparation for in vitro MRI, the following protocol serves rather as a red thread with optional steps than as step-by-step protocol. 1. The brain is perfusion fixed according to Section 3.4.1 (steps 1–6) 2. PFA is removed and the sample tissue is equilibrated in 0.1 M sodium phosphate, pH 7.4, for 24 h 3. The fixed specimens can be pretreated with GdDOTA (0.1 mmol in 5% formalin solution for o/n incubation) 4. For MR acquisition, the sample is embedded in a tube filled R ), a fluorocarbon with a perfluorinated polyether (Fomblin R  (Fluorinert ), or in 2% agar (see Note 11) Brain sections for histological MRI [following (26)] 1. The brain is perfusion fixed according to Section 3.4.1. 2. Sections are cut at 60 μm on a freezing microtome or in a cryostat. 3. 60 μm tissue samples are floated in PBS for 30 min to allow any residual formalin and sucrose in the tissue to leach out.

528

Baltes et al.

4. Sections are then encased between the two coverslips and placed into the histological RF coil. 3.4.3. Human AD Brain Samples Preparation for In Vitro MRI

So far, only a limited number of in vitro MRI studies on human AD brain samples with neurofibrillary changes staged according to Braak and Braak (31) can be found in the literature, restricted with regard to small sample sizes and mainly aiming to show the principle feasibility, i.e., to achieve the required resolution (8, 9). The hippocampal specimen used was embedded in Fomblin and covered by a glass coverslip. Only recently, Meadocrowft and colleagues (26) investigated Aβ plaque MR with regard to potential differences between the human and mouse AD tissue samples, thereby investigating the underlying contrast mechanism. Here, the brain tissue derived from human AD patients (see Note 12) was immersion fixed and prepared for histological MRI according to Section 3.4.2.

3.4.4. Histology

Congo red Staining or thioflavin-S staining (tissue sections mounted on slides) 1. The brain is perfusion fixed and sectioned according to Section 3.4.1. 2. The brain is then sectioned at 40 μm on a freezing microtome (or 15 μm on a cryostat), and selected series are then mounted onto gelatine-coated glass slides and air dried. R ) is hydrated by 3. The defatted tissue (in xylene or CitriSolv taking it through a series of ethyl alcohol solutions (100, 95, 80, and 70%, 1–2 min in each) before staining in the Congo red solution for 1 h or 1% thioflavin-S for 30–60 min.

4. The slides are subsequently dipped into a saturated lithium carbonate solution for 10–20 s and then rinsed for a similar period in distilled water. 5. Slides are subsequently transferred through 80 and 95%, and two sets of 100% ethanol (15 s–2 min in each) before placing them in xylene or CitriSolv for 5–10 min. 6. The slides can now be coverslipped with mounting media and stored in a cool, dark place, such as a refrigerator. Congo red staining can then be viewed under plain polarized light. Amyloid plaques should give apple-green birefringence, usually as a Maltese cross, whereas the neurons that are nonspecifically stained will not emit birefringence. Thioflavin-S staining is viewed under green fluorescence (480/525 nm) using the filter set recommended for that particular microscope. Perls’ Prussian blue reaction with DAB intensification 1. Mounted sections are hydrated in PBS and incubated in 1–2% potassium ferrocyanide for 30 min.

Detecting Amyloid-β Plaques in Alzheimer’s Disease

529

2. For diaminobenzide (DAB)-enhanced PB staining, the sections are subsequently incubated in hydrogen peroxideactivated DAB solution for 15 min in the dark. Slides are then washed again in PBS and counterstained with nuclear fast red and coverslipped for imaging. Immunohistochemistry: Staining for amyloid plaques and preamyloid deposits (on free-floating sections) Aβ-immunohistochemistry is performed using standard proR ) or the cedures, such as the Vector Mouse on Mouse (MOM VECTASTAIN Elite ABC system (Vector Laboratoires) and antiAβ antibodies (e.g., 6E10 mouse monoclonal antibody from Signet that is specific for human Aβ; 1:1,000-fold dilution). Initially, certain controls should be included, such as preadsorption of the primary antibody, with the antigen and omitting of the primary and secondary antibodies, as well as the avidin–peroxidase complex. For mouse-on-mouse detection, it is appropriate to include omission of the primary antibody in each run. For optimal immunodetection of plaques, it is not necessary to pre-treat the sections with heat or formic acid as is preferable for human sections. For detailed staining procedure and recommendation on antibodies, the reader is referred to Sigurdsson (32). 3.5. Correlation with Histological Analysis

As outlined above, it is impossible to spatially register an entire in vivo MRI volume to an ex vivo MRI or histological volume with accuracy at the tens-of-micrometers level. Furthermore, when comparing histology and MR images, one needs to take into account the slight expansion of the histological slices during preparation. For this reason, the images need to be matched based on common anatomical landmarks, such as ventricles, corpus callosum, hippocampal fissure. In Jack and colleagues (17) a pragmatic approach is described to correlate histological sections with MRI, briefly: 1. Medium resolution images of the individual histological sections are combined to create a digitized three-dimensional volume of the specimen (32 images per stain per mouse). 2. Appropriate multiples of the 30 μm histological sections were used to match the 120 μm through-plane resolution of MRI. 3. Adjacent 30 μm sections were stained with thio-S and DABenhanced Prussian blue; therefore, matching between the MRI, thio-S-, and iron-stained sections contained inherent partial volume averaging approximations. 4. The digitized histological volume was then spatially matched to the in vivo MRI volume using anatomical landmarks common to both volumes as described above. See Fig. 26.4 for images of AD mice.

530

Baltes et al.

Fig. 26.4. a Nine-month-old AD mouse. In vivo MRI, ex vivo MRI, thio-S-stained, and iron-stained images have been precisely spatially registered over a circumscribed area of the cortex, indicated by the box. The boxes in the right column (scale bar, 100 μm) represent 3× magnified portions of the adjacent parent image in the left column (scale bar, 1.0 mm). The numbered arrows indicate individual plaques visualized in each of the four different image types that matched with the linked cursor system. b Twelve-month-old AD mouse. In vivo MRI, ex vivo MRI, thio-S-stained, and iron-stained images have been precisely spatially registered over a circumscribed area of the cortex, indicated by the box. The boxes in the right column (scale bar, 100 μm) represent 3× magnified portions of the adjacent parent image in the left column (scale bar, 1.0 mm). The numbered arrows indicate individual plaques visualized in each of the four different image types that matched with the linked cursor system (reprinted/adapted (17), Figs. 2 and 4).

4. Notes 1. Isoflurane is recommended due to its easiness of usage (i.e., fast induction and termination of anesthesia, adjustable depth of anesthesia) and the ability of repetitive application. The anesthetic procedure has to be adapted according to local Animal Care Guidelines. 2. Some commercial systems comprise respiration pressure transducers (e.g., SA Instruments, Stony Brook, NY, USA). 3. Some suppliers provide fiber-optic devices (e.g., QUASYS AG, Cham, Schweiz). 4. As promising option to increase the MR sensitivity, cryogenic RF coils have recently become commercially available. It was demonstrated that cooling of RF probe (30 k) and preamplifier (77 k) leads to a gain in SNR of a factor of 2.5 (33). 5. For immuno-histochemical procedures, commercially available coated slides are sufficient, but for Congo red

Detecting Amyloid-β Plaques in Alzheimer’s Disease

531

staining of slide-mounted sections, it is preferable to use gelatin-coated slides. 6. In vivo protocols have to be adapted according to the local Animal Care Guidelines. 7. Pancuronium can be antagonized using, e.g., neostigmine. 8. Some protocols recommend up to 50 ml. However, for an adult mouse, a volume of 20 ml is enough – the longer the interval between flushing the blood and introducing the fixative, the more the artifacts one can produce. 9. Insufficient epitope accessibility due to paraformaldehyde fixation for certain antibodies or non-specific labeling of the murine vasculature using mouse monoclonal antibodies might necessitate other methods to preserve the sample tissue. 10. 3–4 h postfixation is recommended (see Note 8: overfixation is a common cause for non-specific antibody immunoreactivity). If the experimental setup requires o/n fixation, the tissue should be placed into a 1% PFA solution. 11. Fomblin, Fluorinert, or agar is used as an embedding medium to limit tissue dehydration, as well as susceptibility effects at the surface of the specimen. 12. For a comparison of brain samples derived from different AD, as well control patients, attention has to be drawn on the delay time between the tissue harvesting at the time of death and the time point of the MR acquisition, which can be significantly different.

Acknowledgments The author would like to thank Dr. Irene Knüsel and Dr. Martin Herzig for careful reading of the manuscript and providing their comments on AD mouse models and histology. References 1. Selkoe, D. J. Alzheimer’s disease: Genes, proteins, and therapy. Physiol Rev 2001;81: 741–766. 2. Klunk, W. E., Lopresti, B. J., Ikonomovic, M. D., Lefterov, I. M., Koldamova, R. P. Binding of the positron emission tomography tracer Pittsburgh compoundB reflects the amount of amyloid-beta in

Alzheimer’s disease brain but not in transgenic mouse brain. J Neurosci 2005;25: 10598–10606. 3. Meyer-Luehmann, M., Spires-Jones, T. L., Prada, C. et al. Rapid appearance and local toxicity of amyloid-beta plaques in a mouse model of Alzheimer’s disease. Nature 2008;451(7179):720–724.

532

Baltes et al.

4. Guentert, A., Doebeli, H., Bohrmann, B. High sensitivity analysis of amyloid-beta peptide composition in amyloid deposits from human and PS2APP mouse brain. Neuroscience 2006;143:461–475. 5. Kuo, Y. M., Kokjohn, T. A., Beach, T. G. et al. Comparative analysis of amyloid-beta chemical structure and amyloid plaque morphology of transgenic mouse and Alzheimer’s disease brains. J Biol Chem 2001;276: 12991–12998. 6. Connor, J. R., Menzies, S. L., St. Martin, S. M., Mufson, E. J. A histochemical study of iron, transferrin and ferritin in Alzheimer’s diseased brains. J Neurosci Res 1992;31: 75–83. 7. Meadowcroft, M. D., Connor, J. R., Smith, M. B., Yang, Q. X. MRI and histological analysis of Beta-Amyloid Plaques in both human Alzheimer’s disease and APP/PS1 transgenic mice. J Magn Reson Imaging 2009;29:997–1007. 8. Benveniste, H., Einstein, G., Kim, K. R., Hulette, C., Johnson, G. A. Detection of neuritic plaques in Alzheimer’s disease by magnetic resonance microscopy. Proc Natl Acad Sci USA 1999;96:14079. 9. Dhenain, M., Privat, N., Duyckaerts, C., Jacobs, R. E. Senile plaques do not induce susceptibility effects in T2∗ -weighted MR microscopic images. NMR Biomed 2002;15:197–203. 10. Poduslo, J. F., Wengenack, T. M., Curran, G. L. et al. Molecular targeting of Alzheimer’s amyloid plaques for contrastenhanced magnetic resonance imaging. Neurobiol Dis 2002;11:315–329. 11. Rudin, M., Mueggler, T., Allegrini, P. R., Baumann, D., Rausch, M. Characterization of CNS disorders and evaluation of therapy using structural and functional MRI. Anal Bioanal Chem 2003;377:973–981. 12. Helpern, J. A., Lee, S. P., Falangola, M. F. et al. MRI assessment of neuropathology in a transgenic mouse model of Alzheimer’s disease. Magn Reson Med 2004;51(4):794–798. 13. Lee, S. P., Falangola, M. F., Nixon, R. A., Duff, K., Helpern, J. A. Visualization of betaamyloid plaques in a transgenic mouse model of Alzheimer’s disease using MR microscopy without contrast reagents. Magn Reson Med 2004;52:538–544. 14. Zhang, J., Yarowsky, P., Gordon, M. N. et al. Detection of amyloid plaques in mouse models of Alzheimer’s disease by magnetic resonance imaging. Magn Reson Med 2004;51:452–457. 15. Jack, C. R., Jr, Garwood, M., Wengenack, T. M. et al. In vivo visualization

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

of Alzheimer’s amyloid plaques by magnetic resonance imaging in transgenic mice without a contrast agent. Magn Reson Med 2004;52:1263–1271. Vanhoutte, G., Dewachter, I., Borghgraef, P., Van Leuven, F., Van der Linden, A. Noninvasive in vivo MRI detection of neuritic plaques associated with iron in APP[V717I] transgenic mice, a model for Alzheimer’s disease. Magn Reson Med 2005;53(3):607–613. Jack, C. R., Jr, Wengenack, T. M., Reyes, D. A. et al. In vivo magnetic resonance microimaging of individual amyloid plaques in Alzheimer’s transgenic mice. J Neurosci 2005;25(43):10041–10048. Jack, C. R., Jr, Marjanska, M., Wengenack, T. M. et al. Magnetic resonance imaging of Alzheimer’s pathology in the brains of living transgenic mice: A new tool in Alzheimer’s disease research. Neuroscientist 2007;13(1):38–48. Wadghiri, Y. Z., Sigurdsson, E. M., Sadowski, M. et al. Detection of Alzheimer’s amyloid in transgenic mice using magnetic resonance microimaging. Magn Reson Med 2003;50:293–302. Higuchi, M., Iwata, N., Matsuba, Y., Sato, K., Sasamoto, K., Saido, T. C. 19F and 1H MRI detection of amyloid b plaques in vivo. Nat Neurosci 2005;8:527–533. Higgins, G. A., Jacobsen, H. Transgenic mouse models of Alzheimer’s disease: Phenotype and application. Behav Pharmacol 2003;14:419–438. Ramakrishnan, M., Wengenack, T. M., Kandimalla, K. K., Poduslo, J. F. et al. Selective contrast enhancement of individual Alzheimer’s disease amyloid plaques using a polyamine and Gd-DOTA conjugated antibody fragment against fibrillar Abeta42 for magnetic resonance molecular imaging. Pharm Res 2008;25(8):1861–1872. Wengenack, T. M., Jack, C. R., Jr., Garwood, M., Poduslo, J. F. MR microimaging of amyloid plaques in Alzheimer’s disease transgenic mice. Eur J Nucl Med Mol Imaging 2008;35(Suppl 1):S82–S88. Poduslo, J. F., Ramakrishnan, M., Holasek, S. S. et al. In vivo targeting of antibody fragments to the nervous system for Alzheimer’s disease immunotherapy and molecular imaging of amyloid plaques. J Neurochem 2007;102(2):420–433. Braakman, N., Matysik, J., van Duinen, S. G. et al. Longitudinal assessment of Alzheimer’s β-Amyloid plaque development in transgenic mice monitored by in vivo magnetic resonance microimaging. J Magn Reson Imaging 2006;24:530–536.

Detecting Amyloid-β Plaques in Alzheimer’s Disease 26. Meadowcroft, M. D., Zhang, S., Liu, W. et al. Direct magnetic resonance imaging of histological tissue samples at 3.0T. Magn Reson Med 2007;57:835–841. 27. Hildebrandt, I. J., Su, H., Weber, W. A. Anesthesia and other considerations for in vivo imaging of small animals. ILAR J 2008;49(1):17–26. 28. Chamberlain, R., Reyes, D., Curran, G. L. et al. Comparison of amyloid plaque contrast generated by T2-Weighted, T∗ 2-weighted, and susceptibility-weighted imaging methods in transgenic mouse models of Alzheimer’s disease. Magn Reson Med 2009;61: 1158–1164. 29. Paxinos, G., Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates. San Diego, CA: Academic Press; 2001. 30. Muskulus, M., Scheenstra, A. E. H., Braakman, N. et al. Prospects for early detection of Alzheimer’s disease from serial MR images in transgenic mouse models. Curr Alzheimer Res 2009;6:503–518. 31. Braak, H., Braak, E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991;82:239–259. 32. Sigurdsson, E. M. Histological staining of amyloid-beta in mouse brains. Methods Mol Biol 2005;299:299–308. 33. Baltes, C., Radzwill, N., Bosshard, S., Marek, D., Micro, R. M. MR imaging of the mouse brain using a novel 400 MHz cryo-

34.

35.

36.

37.

38.

39.

533

genic quadrature RF probe. NMR Biomed 2009;22(8):834–842. Games, D., Adams, D., Alessandrini, R. et al. Alzheimer-type neuropathology in transgenic mice over-expressing V717F beta-amyloid precursor protein. Nature 1995;373: 523–527. Moechars, D., Dewachter, I., Lorent, K. et al. Early phenotypic changes in transgenic mice that overexpress different mutants of amyloid precursor protein in brain. J Biol Chem 1999;274(10):6483–6492. Hsiao, K., Chapman, P., Nilsen, S. et al. Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. Science 1996;274:99–102. Sturchler-Pierrat, C., Abramowski, D., Duke, M. et al. Two amyloid precursor protein transgenic mouse models with Alzheimer disease-like pathology. Proc Natl Acad Sci USA 1997;94:13287–13292. Holcomb, L., Gordon, M. N., Accelerated, Mc. G. E. Alzheimer-type phenotype in transgenic mice carrying both mutant amyloid precursor protein and presenilin 1 transgenes. Nat Med 1998;4:97–100. Richards, J. G., Higgins, G. A., Ouagazzal, A. M. et al. PS2APP transgenic mice, coexpressing hPS2mut and hAPPswe, show agerelated cognitive deficits associated with discrete brain amyloid deposition and inflammation. J Neurosci 2003;23:8989–9003.

wwwwwww

Chapter 27 Assessing Subtle Structural Changes in Alzheimer’s Disease Patients Jennifer L. Whitwell and Prashanthi Vemuri Abstract Magnetic resonance imaging (MRI) allows the assessment of structural changes in subjects with Alzheimer’s disease (AD). Early studies used visual assessments of MRI or manual measurements of structures of interest, although these methods were limited by inter-rater variability. Techniques have now been developed which allow automated analysis of both cross-sectional and longitudinal MRI data and have provided valuable information concerning the patterns and progression of atrophy in subjects with AD. It is also now possible using machine learning-based techniques to provide individual-level diagnostic information from MRI scans. Various analysis techniques have been applied to validate the use of MRI to capture subtle structural changes due to atrophy in AD and its usefulness in providing diagnostic and prognostic information, as well as tracking the disease progression in AD. Key words: Structural MRI, atrophy, Alzheimer’s disease, registration, segmentation, longitudinal.

1. Introduction Structural magnetic resonance imaging (MRI) has been increasingly recognized as an important tool both in the study and diagnosis of degenerative dementias, particularly Alzheimer’s disease (AD). Structural changes in the brain in AD include gray matter atrophy which is related to the loss of neurons, synapses, and dendritic de-arborization that occurs on a microscopic level and white matter atrophy related to loss of structural integrity of white matter tracts presumably resulting from the dying back of axonal processes. Structural MRI, specifically T1-weighted sequences, allow accurate quantitation of atrophy of cerebral structures due M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_27, © Springer Science+Business Media, LLC 2011

535

536

Whitwell and Vemuri

to its ability to acquire high-resolution images with good soft tissue contrast properties. Studies using structural MRI in AD have found severe atrophy of the hippocampus and temporoparietal neocortex, as well as changes in the cingulate cortices and frontal lobes (1–3). A number of clinicopathological studies have in fact shown that the volumetric changes on MRI match well with the distribution of underlying pathological changes that occur in the brain (1, 4). In addition, since MRI is readily available, noninvasive and safe to perform over multiple time-points, it has been used to assess longitudinal changes over time in subjects with AD. MRI studies have demonstrated that progressive atrophy in the hippocampus and neocortex correlates well with clinical decline (5). MRI has also been shown to capture subtle changes in the brain which occur even before the onset of symptoms in AD (6); often while subjects have a diagnosis of amnestic mild cognitive impairment (aMCI) which is considered a prodromal stage of AD (7, 8). Measurements from structural MRI have therefore become well established as excellent biomarkers of disease progression in AD and are routinely included as outcome measures in clinical trials. Different techniques have been developed to analyze structural MRI data. Early studies employed simple visual assessments of MRI or manual measurements of structures of interest. However, these measurements are associated with large interrater variability and manual measurements are time consuming. In addition, region-of-interest-based analyses are spatially limited and do not make use of all the available information contained in a 3D image data set. To address these issues numerous methods have been developed over the past decade which allow the automated assessment of MRI scans. These techniques make it possible to analyze large cohorts of data and can assess changes at the voxel level (9, 10). While specific techniques vary and use different software algorithms the majority share some common processing steps, including a registration step to put scans in the same spatial framework, and some form of segmentation of the brain tissue into different tissue types. We will concentrate on describing well-established techniques that analyze volume changes across the whole brain, although the assessment of cortical thickness has also become popular. Statistical analysis is generally performed at the group level across every voxel of the brain, although techniques have recently been developed to assess atrophy at the individual level and provide diagnostic information based on the MRI scan (11). This chapter provides a framework for designing, acquiring, and analyzing structural MRI data to detect subtle structural changes due to AD. However, an exhaustive list of considerations for designing an observational study with MRI can be found in (8). We will first touch on the various aspects of obtaining

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

537

high-quality MRI data in subjects with AD, including subject recruitment and study design, image acquisition, quality control, and preprocessing. Later, we will focus on discussing the steps necessary to analyze both cross-sectional and longitudinal volumetric data using automated image processing pipelines.

2. Materials 1. MRI scanner: Available 1.5 or 3 T MRI scanners (most of the MRI AD studies have been validated on these field strength scanners). 2. Software programs: One may choose a software package depending on the study design and the analysis questions asked. Manual and visual assessment of images can be done using any number of image display software programs available, such as the commonly used Analyze program (http:// www.analyzedirect.com/). The popular programs for automated assessment of images are a. Tissue-based segmentation: SPM (http://www.fil.ion.ucl. ac.uk/spm/) and FSL (http://www.fmrib.ox.ac.uk/ fsl/). SPM additionally requires MATLAB (http://www. mathworks.com/). b. Cortical thickness estimation: FreeSurfer (http://surfer. nmr.mgh.harvard.edu/). c. Additional image processing tools: ITK (http://www.itk. org/) and LONI (http://www.loni.ucla.edu/Software/ index.php). 3. Computer requirements: Running an automated image processing pipeline requires a computer with a lot of disk space, memory, and processor speed. Machines with 2.0 GHz, or faster, processor with at least 2 GB RAM, 3D graphics card for image visualization, and sufficient disk space (at least 300 GB memory which may vary depending on the size of the study) are recommended.

3. Methods 3.1. Subject Recruitment

Subject selection and recruitment are an important first step of any MRI study. The conclusions that are drawn from a study depend specifically on the patient cohort and how representative

538

Whitwell and Vemuri

that cohort is of the sampled population. Here we provide some factors to consider while selecting subjects for the study: 1. Specify the aims of the study and choose a study design based on the relevant questions you plan to investigate. Careful consideration should be given to the appropriate number of subjects to recruit in order to address the study aims (see Note 1). Power calculations can be performed to help ensure the study will be well powered. 2. Potential subjects should be screened for factors that could potentially affect brain volume, such as previous history of chemotherapy, head injury, radiation therapy, alcoholism, epilepsy, and stroke. They should also be screened for factors that could prevent the acquisition of an MRI, such as claustrophobia, the presence of a pace maker, or metal objects in the head. 3. Establish clinical inclusion and exclusion criteria for the study: It is important to use established clinical criteria for AD (such as the commonly used NINDCS-ADRDA criteria (12)) for consistency across studies. Similarly, clinical criteria should be used to exclude the other neurodegenerative disorders that can often overlap clinically with AD, such as vascular dementia, dementia with Lewy bodies, and frontotemporal dementia. In addition, patterns of atrophy may vary according to factors such as age at onset, disease severity, family history, and the presence of genetic risk factors and so you may wish to consider these features in group selection. Population studies should select their cohort to be representative of the underlying population in terms of clinical features and demographics. 4. Select appropriate control populations: A group of cognitively normal (CN) subjects are often selected as a control comparison group in order to adjust for atrophy associated with normal aging. Ideally, this group would be selected to match the disease group on demographics, such as age, gender, and education, dependent on the study aims. Matching could be on a one-to-one (or individual) basis or frequency matched. If matching is not possible then these potential confounders should be dealt with in the statistical analysis. In general, the power of the study will increase with larger control groups, although the benefit of extra subjects will diminish drastically after the number of CN is twice as many as the patients. In studies that deal with multiple different disease groups, it is advantageous to keep the sample sizes similar across disease groups. 3.2. Image Acquisition and Preprocessing

Designing an appropriate MRI protocol plays a key role in the study, specifically if data will be processed through automated pipelines which expect high signal-to-noise ratio (SNR), good

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

539

contrast, and a stable acquisition. The methods required for obtaining high-quality imaging sequences in subjects with AD have recently been assessed and optimized by many experts in the field as part of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (www.adni-info.org) (13). Here are the important factors that need to be considered while planning an MRI acquisition: 1. Choice of MRI scanners: While choosing the field strength of the scanner bear in mind that even though SNR in tissue increases roughly linearly with field strength, RF penetration and standing wave effects which can cause severe intensity distortions in the acquired images also increase with field strength. That being said, measurement of subtle changes in structural images has been extensively validated on both 1.5 and 3 T MRI scanners (see Note 2). If using more than one MRI scanner in the study, it is advisable to use similar field strength scanners, since the tissue signal properties are a function of the field strength. In addition, when combining data from multiple MRI scanners, it is better to run all the images through preprocessing steps listed in the subsection below to ensure the intensity uniformity of images across different scanners. 2. Choice of MRI imaging protocol: For morphometric imaging of AD, the key feature of the imaging sequence is good contrast between gray matter, white matter, and CSF which is typically found in 3D T1-weighted structural images. The most commonly used imaging sequences are the spoiled gradient recalled (SPGR) sequence and the magnetization prepared rapid acquisition gradient echo (MPRAGE) at 1.5 and 3 T. While both pulse sequences have similar properties, MPRAGE has been shown to perform slightly better on perceived indices of image quality at 3 T (when compared to SPGR acquired in the same acquisition time). Suggested scanning parameters for the MPRAGE sequence can be found in (13). 3. Standardized imaging protocols: It is important to standardize the imaging parameters in order to reduce quantification errors due to a variable protocol. 4. Calibration of scanners using phantom imaging: Calibration scans are required to monitor scanner performance over time and become especially crucial when conducting longitudinal studies and also multi-scanner studies. Scanner instability may artificially either increase or decrease volume between different groups or different time-points on the same subjects. Therefore, scaling (for variability of linear geometry) of images using concurrent phantom acquisitions or registration algorithms is recommended in order to standardize the observed voxel sizes for different acquisitions.

540

Whitwell and Vemuri

Here are some preprocessing procedures that can be applied to the scans in order to further standardize the acquired images before analysis: 1. Inhomogeneity correction: Image intensity in-homogeneity can arise from the use of multi-array receiver coils, which are often used for increased SNR instead of single channel birdcage coils, resulting in a bright to dark outside-to-in pattern. Additionally at high field strength there is intensity distortion due to RF penetration and standing wave effects (called “dielectric resonance” artifact). Since many image analysis tools rely on similar tissues having homogenous intensity it is important to minimize these artifacts. Many algorithms exist to correct these effects, although one of the most common solutions is the non-parametric non-uniformity normalization (N3) algorithm which is employed to estimate and remove the smooth multiplicative bias or image nonuniformity (14). 2. Grad-warp correction: A major cause of spatial distortion of anatomical images is gradient nonlinearity; the deviation of the gradient field from ideal linear function of position. Gradient nonlinearity can be corrected by a process known as “grad warp.” This involves obtaining the spherical harmonic coefficients for a particular gradient design configuration from the manufacturer and applying an inverse warping procedure to images acquired with that particular gradient system. This procedure was applied to all MRI in the ADNI (13). 3.3. Quality Control

All imaging studies should have a rigorous process of quality control (QC) in which each image is checked to ensure that it meets an acceptable standard. There are three recommended aspects to the QC procedure: 1. Ensure protocol compliance: Ensure that the scan was performed using the correct protocol and parameters, as well as on the correct scanner. 2. Check for features that could confound volume measurements, such as metallic artifacts, head trauma, hemispheric infarctions, previous surgery, hemorrhage, space occupying lesion, or cerebral edema. Scans should be excluded from the study if these features disrupt brain structure. Subjects should also be excluded if the MRI scan indicates the presence of another disorder, such as normal pressure hydrocephalus. 3. Check the quality of the MRI: Scans should be excluded from the study if they have excessive movement, flow or susceptibility artifacts, inhomogeneity, nose or ear wrap that affects the brain, or incomplete brain coverage. In some

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

541

instances, the scalp may be required for image analysis in which case scans could be excluded for incomplete head coverage. 3.4. Automated Cross-Sectional Image Analysis

Here we present the broad framework by which automated crosssectional analysis of MRI scans can be performed (9, 10). A number of processing steps are required in order to get the data ready for statistical analysis. However, the algorithms used to perform these steps will differ between each of the popular software packages typically used for cross-sectional analysis. 1. Spatial normalization: For an automated cross-sectional comparison of MRI scans, all the MRI images from different subjects need to be in a common template space (e.g., single subject reference image or custom templates). This step ensures that a location in one subject’s MRI corresponds to the same location in another subject’s MRI before a quantitative cross-sectional comparison is made. The complexity of spatial normalization used to align scans varies from 12 degrees of freedom (dof) affine transformation (which involves rotation, translation, scaling, and shearing of the images to align) to sophisticated high-dimensional warping algorithms (which allow every segment of the MRI scan to deform freely to match the common template). 2. Segmentation: Images are segmented into different tissue compartments (For T1-weighted images – gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) or cortical surfaces), and analysis is performed on the structure of interest. There are a number of ways to perform the segmentation which vary from simple intensity-based segmentation algorithms (such as k-means clustering) to sophisticated algorithms (which might incorporate additional neighborhood information and tissue probability maps of already segmented images to guide the segmentation procedure). Automated segmentations could be further improved by manual editing or masking if necessary. 3. Modulation: Image intensities of the different structures are scaled by the amount of deformation that is applied during step 1 to normalize it to a common template, so that the total amount of structural information remains the same in the normalized image as in the original image. This step is called modulation (15). 4. Smoothing: The aim of smoothing the data (i.e., use an approximating function to interpolate the regional data in order to reduce noise) is twofold: (1) reduce intersubject variability; (2) some of the statistical models assume normally distributed structural data. Typically an 8 mm smoothing kernel is sufficient. Larger smoothing kernels

542

Whitwell and Vemuri

will increase the likelihood of identifying changes, but will reduce the ability to localize the changes anatomically. 5. QC: Quality control of the processed images in steps 1 and 2 is very important to ensure that analyses described in the next section are valid. The registration and/or segmentation steps often fail on some of the scans due to gross misalignment of the subjects MRI scan to the template space. The solution to correct this is a simple rotation and translation of the scan (this can be done manually or automatically) to center the ventricles of the subject’s scan to those of the templates. See Notes 4.3, 4.4 and 4.5 for discussion of specific problems associated with these procedures in assessing AD patients. 3.5. Longitudinal Image Analysis

Assessing subtle structural changes over multiple serial MRI scans in subjects with AD is an important area of study. Ideally, serial scans for an individual should be performed using the same scanner and same acquisition protocol approximately 12 months apart. Measurement error can account for a large proportion of the observed signal at short intervals (<6 months), and a linear approximation of rates is not as appropriate over large intervals (>3 years). Longitudinal data can be analyzed both at the whole brain and at the voxel level with the choice of appropriate technique depending on the specific questions being addressed in the study. To analyze whole brain rates of atrophy: 1. Registration: The repeat scan from an individual is matched to the baseline scan using a linear registration algorithm that applies exactly the same set of parameters to all voxels. This is typically a 9 dof rigid-body registration, although shears can be included by increasing the dof to twelve for a full affine registration. 2. Brain segmentation: The registration will often require a brain region to be outlined on at least one of the serial scans. Brain segmentations can be performed automatically using a number of different software programs, such as SPM. 3. Atrophy quantification: Atrophy can be quantified from a registered scan pair either by calculating the brain volume on both scans and then subtracting the difference or more accurately by using an algorithm called the brain boundary shift integral (BSI) (16). The BSI determines the total volume through which the boundaries of the brain have moved and, hence, the volume change, directly from voxel intensities. The boundaries of a brain will have a characteristic intensity profile, with brighter voxels in the brain to dark voxels in CSF. The BSI compares the intensity profile of the baseline brain to the corresponding intensity profile on a registered repeat scan and calculates the difference in

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

543

Fig. 27.1. a Color map of BSI-computed change in volume superimposed on the baseline MRI scan for an AD patient with a scan interval of 1 year. Red indicates shrinkage and green indicates expansion. b Plot showing change in brain volume over time in subjects with pathologically confirmed AD and frontotemporal lobar degeneration with ubiquitin-positive immunoreactive inclusions (FTLD-U). Individual subjects and a model estimated best fit slope are shown (Reproduced with permission from Ref. (17)).

the two boundaries over a specific intensity window. This simple theory is then applied in three dimensions throughout the whole brain boundary. Another variant of BSI called SIENA is available in the FSL package. Figure 27.1a illustrates a typical brain volume change detected by BSI in an AD patient superimposed on the baseline MRI scan. Figure 27.1b shows how brain volume changes measured using BSI can be used to track change over time in individuals with AD and other neurodegenerative conditions (17). Similar techniques can also be used to calculate rates of atrophy for particular regions of the brain, for example, the ventricles or hippocampi. Techniques for calculating rates of atrophy at the voxel level can vary greatly between research groups. Here we describe one implementation, referred to as tensor-based morphometry, which can be run using SPM software: 1. A rigid-body registration is first performed to align the serial scans. 2. A high-dimensional non-linear registration is then applied to the registered scan pair which provides more accurate matching of gyral anatomy by providing more freedom to warp one image onto the other (see Note 6). Non-linear registrations are generally more computationally expensive than linear registration techniques. The nonlinear registration aims to deform one image to provide an exact match with the other serial image. The amount of volume change can be quantified by taking the determinant of the deformation field at each voxel, producing what is termed the Jacobian determinant. 3. The baseline image (or alternatively an image that can be created by averaging the two warped scans) is segmented to create a baseline gray matter image.

544

Whitwell and Vemuri

4. The baseline gray matter image is then multiplied by the Jacobian determinant from the nonlinear transformation creating a product image. 5. The baseline gray matter image for each subject in the study is then normalized to a customized gray matter template using the methods described in Section 3.4, and the normalization parameters are applied to the product image. 6. The normalized gray matter baseline and product are smoothed. 7. Rates of atrophy can then be calculated by assessing differences between the baseline and the product scans (see Section 3.6). Alternatively, some investigators compare the Jacobian maps directly between groups without having to create a product image (18). This implementation involves normalizing the baseline scans and Jacobian maps to a template and then performing statistical analysis on the Jacobian maps. Nonlinear registrations can be very sensitive to factors such as poor scan resolution, intensity inhomogeneity, large signal-to-noise ratio, or artifacts appearing on the image. For this reason, it is always important to check the accuracy of the registration. 3.6. Statistical Analysis of Pattern Differences

Once the image processing steps described in Sections 3.4 and 3.5 have been performed the resultant smoothed images can be analyzed statistically at the voxel level to compare regional volumes between groups. Gray and white matter images are usually analyzed separately, although changes in white matter may be assessed more accurately using diffusion tensor imaging (DTI). Similar statistical methods can be used to analyze both crosssectional and longitudinal data. The steps discussed can be performed easily in SPM, although they could also be implemented in other software packages. 1. Select a statistical test: Statistical analysis can be performed with parametric statistics using the general linear model and the theory of Gaussian random fields to ascertain significance, although non-parametric testing can also be applied. For parametric statistics, simple group comparisons typically use a t-test, paired t-test (particularly for longitudinal analysis), or ANOVA-type design, although regression analyses can also be performed to investigate regions of the brain that correlate with specific clinical indices. The null hypothesis is that there is no difference in tissue volume between the groups in question. 2. Enter appropriate confounding variables as covariates in the analysis: These variables will depend on the analysis in question and how well the subject groups are matched; however,

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

545

it is often recommended to include age and gender. The total intracranial volume (TIV) which provides a measure of premorbid brain volume could also be included to correct for any differences in head size across groups. Analyses correcting for the total gray matter volume, i.e., global normalization, can also be performed. 3. Masking: In order to reduce the number of false positives in the data analysis, it is important to restrict the analysis to the regions of interest. The most common masking strategy is threshold masking, such that voxels below a threshold in all the subjects are set to zero. Alternatively, an explicit mask that has been generated by the user can be applied. The restriction of the analysis to specific structures of the brain can increase the chance of identifying significant results, but should ideally be hypothesis driven and based on previous work. Exploratory studies should analyze the entire cortex. 4. Correction for multiple comparisons: Since statistical tests are performed across a very large number of voxels, it is important that studies correct for multiple comparisons in order to prevent the occurrence of false positives. There are a couple of typical methods used to perform such a correction, for instance, the family-wise error (FWE) correction and the more lenient false discovery rate (FDR) correction that both reduce the chance of false-positive results (www.fil.ion.ucl.ac.uk). The FWE correction controls the chance of any false positives (as in Bonferroni methods) across the entire volume, whereas the FDR correction controls the expected proportion of false positives among suprathreshold voxels. These analyses generate statistical maps showing all voxels of the brain that refute the null hypothesis and show significance to a certain, user selected, p value. The results shown in Fig. 27.2 demonstrate that the methods discussed here and in Section 3.4 can detect subtle patterns of atrophy in subjects even before they receive a diagnosis of AD (see Note 7). However, instead of performing voxel level analysis, region level structural information, e.g., hippocampal volume, can also be obtained by normalizing the MRI images to a template space (see Section 3.4, step 1) where the regions of interest are already defined on an atlas. 3.7. Automated Individual Subject Diagnosis

Voxel-based analytic techniques are agnostic of disease-specific information, since they are designed to perform only group-wise comparisons and thus are unsuitable for evaluating the disease state of an individual subject. Because of this disadvantage, investigators have recently turned their attention toward multivariate analysis and machine learning-based algorithms for distinguishing AD patients from CN subjects. These techniques use the entire

546

Whitwell and Vemuri

Fig. 27.2. Patterns of gray matter loss in a group of 30 subjects compared to controls, shown at three different timepoints in their disease: approximately 3 years before progression to AD while carrying the diagnosis of mild cognitive impairment (MCI), approximately 1 year before progression to AD while still MCI, and at the time of the first AD diagnosis. Data for each time-point were processed using the steps described in Section 3.4 and analyzed as described in Section 3.6. The results show that atrophy starts in the medial temporal lobes and fusiform gyrus at least 3 years before subjects reach a diagnosis of AD and then spreads to the posterior temporal lobes and parietal lobes and then by the time of the first AD diagnosis the pattern becomes more severe and widespread involving the medial temporal lobe, temporoparietal cortices, and the frontal lobe. Results are shown corrected for multiple comparisons using the FDR at p<0.01. L = left; R = right (Reproduced with permission from Ref. (8)).

3D structural MRI to form a disease model against which individual subjects may be compared. The model here refers to a supervised machine learning algorithm which is “trained” to differentiate an AD patient’s structural MRI scan from that of a CN subject. Here are the main steps involved in the development of an algorithm for the diagnosis of an individual patient based on the subject’s MRI scans: 1. Feature extraction: Obtain region level, voxel level or cortical surface structural information from the MRI images by normalizing images to a common template as listed in Section 3.4. Feature here refers to the information extracted from each subject’s MRI scan that will be compared across all the subjects. 2. Feature reduction or selection: If the number of input features is much larger compared to the number of patients available to develop the model, the developed model will be complex and might not generalize well to new patient data. This problem is referred to as over-fitting. In order to avoid over-fitting, the aim in this step is to eliminate features that are irrelevant for distinguishing CN from AD or have a large variance across all the subjects. 3. Model selection and optimization: A model (or supervised algorithm, e.g., linear discriminant analysis, principal component analysis, support vector machine, etc.) is selected and the parameters of the model are optimized to maximally discriminate one group from another.

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

547

4. Model validation: At the end of step 3 we have a model that can be applied to discriminate AD scans from CN scans. It is easy to design an optimistically biased (low error, overtrained) algorithm. Therefore it is important to know if the model or the classifier works well enough to be useful for the application. The best way to validate a model is by testing the model on a new independent test data set (which has the correct classification information available) that was not used for training or using cross-validation techniques to report the average performance of the model. Figure 27.3A shows the regions (voxel level features) that were automatically selected to be important for the classification model in step 3 for one such model developed in (11). Most of these locations are in the medial–basal temporal lobe which is affected earliest and most severely in AD. Using these regions, modelbased scores were generated for 101 subjects who had both antemortem MRI and postmortem Braak neurofibrillary tangle (NFT) staging. These scores correlated strongly with the gold standard of pathology (Fig. 27.3b) and support the fact that optimally extracted information from MRI scans can be used as an independent approximate surrogate marker for in vivo pathological staging in AD.

Fig. 27.3. a. Voxels automatically selected to be important for developing model-based scores for diagnosis of AD and CN superimposed on the corresponding custom T1 template. b Model-based scores (generated based on voxels shown above) categorized according to pathological Braak NFT stage (Reproduced with permission from Ref. (11)).

548

Whitwell and Vemuri

4. Notes 1. Longitudinal studies on subjects with AD can suffer from patient drop-out. It is important to take into account expected rates of attrition when planning the study. 2. Scanner upgrades (either hardware or software) can have a significant impact on volume measurements from MRI scanners, due to changes in contrast, SNR or scaling factors. Longitudinal data that spans an upgrade should be used with caution, and mechanisms should be put into place to monitor acquisition stability over time (i.e., using a phantom or control subject). 3. The quality of the normalization discussed in Section 3.4 and Section 3.5 may vary greatly depending on the selected template. For example, scans with atrophic brains may register poorly to a template brain without atrophy. This is a particular problem for older cohorts of subjects with AD. Customized templates that are created using the study cohort or a cohort that is matched to the study cohort in terms of age, disease status, scanner field strength, and scanning parameters are recommended to minimize these problems. 4. The cross-sectional analysis (Section 3.4 and Section 3.6) can be particularly affected by variability in the data. Power to detect group differences will be greater in regions that normalize well across subjects. The use of customized templates, as described above, can help to improve the normalization and hence power across the brain. 5. The misclassification of brain tissue during segmentation (Section 3.4) is especially likely in atrophic brains, both because there is a greater potential for partial volume effects between gray matter and CSF and tissue pathology may be associated with reduced gray/white matter contrast (15). Displacement of tissue due to atrophy can also present a problem. This particularly affects structures surrounding the ventricles, such as the caudate nucleus. Caution is recommended in interpreting findings in these regions. The accuracy of the segmentation will also depend upon the quality of the normalization. Iterative normalization and segmentation methods have been developed which aim to optimize both procedures concurrently to improve the final segmentations (19). 6. Nonlinear registrations described in Section 3.5 may have trouble converging in subjects that have a lot of atrophy between baseline and follow-up scan. In these instances, the registration could be run for a large number of iterations.

Assessing Subtle Structural Changes in Alzheimer’s Disease Patients

549

7. We have not considered MRI acquisitions which are used to detect other changes in structure, for example, due to cerebrovascular disease and DTI which is used to measure change in the structural integrity of white matter fiber tracts. However, most of the discussed methods are generalizable to both kinds of MR acquisitions. References 1. Jack, C. R., Jr., Dickson, D. W., Parisi, J. E., Xu, Y. C., Cha, R. H., O’Brien, P. C., Edland, S. D., Smith, G. E., Boeve, B. F., Tangalos, E. G., Kokmen, E., Petersen, R. C. Antemortem MRI findings correlate with hippocampal neuropathology in typical aging and dementia. Neurology 2002;58:750–757. 2. Baron, J. C., Chetelat, G., Desgranges, B., Perchey, G., Landeau, B., de la Sayette, V., Eustache, F. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer’s disease. Neuroimage 2001;14:298–309. 3. Frisoni, G. B., Testa, C., Sabattoli, F., Beltramello, A., Soininen, H., Laakso, M. P. Structural correlates of early and late onset Alzheimer’s disease: Voxel based morphometric study. J Neurol Neurosurg Psychiatry 2005;76:112–114. 4. Whitwell, J. L., Josephs, K. A., Murray, M. E., Kantarci, K., Przybelski, S. A., Weigand, S. D., Vemuri, P., Senjem, M. L., Parisi, J. E., Knopman, D. S., Boeve, B. F., Petersen, R. C., Dickson, D. W., Jack, C. R., Jr. MRI correlates of neurofibrillary tangle pathology at autopsy: A voxel-based morphometry study. Neurology 2008;71:743–749. 5. Jack, C. R., Jr., Shiung, M. M., Gunter, J. L., O’Brien, P. C., Weigand, S. D., Knopman, D. S., Boeve, B. F., Ivnik, R. J., Smith, G. E., Cha, R. H., Tangalos, E. G., Petersen, R. C. Comparison of different MRI brain atrophy rate measures with clinical disease progression in AD. Neurology 2004;62:591–600. 6. Ridha, B. H., Barnes, J., Bartlett, J. W., Godbolt, A., Pepple, T., Rossor, M. N., Fox, N. C. Tracking atrophy progression in familial Alzheimer’s disease: A serial MRI study. Lancet Neurol 2006;5:828–834. 7. Jack, C. R., Jr., Weigand, S. D., Shiung, M. M., Przybelski, S. A., O’Brien, P. C., Gunter, J. L., Knopman, D. S., Boeve, B. F., Smith, G. E., Petersen, R. C. Atrophy rates accelerate in amnestic mild cognitive impairment. Neurology 2008;70:1740–1752. 8. Whitwell, J. L., Przybelski, S. A., Weigand, S. D., Knopman, D. S., Boeve, B. F., Petersen, R. C., Jack, C. R., Jr. 3D maps

9. 10.

11.

12.

13.

14.

15.

from multiple MRI illustrate changing atrophy patterns as subjects progress from mild cognitive impairment to Alzheimer’s disease. Brain 2007;130:1777–1786. Ashburner, J., Friston, K. J. Voxel-based morphometry – the methods. Neuroimage 2000;11:805–821. Davatzikos, C., Genc, A., Xu, D., Resnick, S. M. Voxel-based morphometry using the RAVENS maps: Methods and validation using simulated longitudinal atrophy. Neuroimage 2001;14:1361–1369. Vemuri, P., Whitwell, J. L., Kantarci, K., Josephs, K. A., Parisi, J. E., Shiung, M. S., Knopman, D. S., Boeve, B. F., Petersen, R. C., Dickson, D. W., Jack, C. R., Jr. Antemortem MRI based STructural Abnormality iNDex (STAND)-scores correlate with postmortem Braak neurofibrillary tangle stage. Neuroimage 2008;42:559–567. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., Stadlan, E. M. Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 1984;34:939–944. Jack, C. R., Jr., Bernstein, M. A., Fox, N. C., Thompson, P., Alexander, G., Harvey, D., Borowski, B., Britson, P. J., J, L. W., Ward, C., Dale, A. M., Felmlee, J. P., Gunter, J. L., Hill, D. L., Killiany, R., Schuff, N., Fox-Bosetti, S., Lin, C., Studholme, C., DeCarli, C. S., Krueger, G., Ward, H. A., Metzger, G. J., Scott, K. T., Mallozzi, R., Blezek, D., Levy, J., Debbins, J. P., Fleisher, A. S., Albert, M., Green, R., Bartzokis, G., Glover, G., Mugler, J., Weiner, M. W. The Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. J Magn Reson Imaging 2008;27:685–691. Sled, J. G., Zijdenbos, A. P., Evans, A. C. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging 1998;17:87–97. Good, C. D., Johnsrude, I. S., Ashburner, J., Henson, R. N., Friston, K. J., Frackowiak, R. S. A voxel-based morphometric study of

550

Whitwell and Vemuri

ageing in 465 normal adult human brains. Neuroimage 2001;14:21–36. 16. Freeborough, P. A., Fox, N. C. The boundary shift integral: An accurate and robust measure of cerebral volume changes from registered repeat MRI. IEEE Trans Med Imaging 1997;16:623–629. 17. Whitwell, J. L., Jack, C. R., Jr., Pankratz, V. S., Parisi, J. E., Knopman, D. S., Boeve, B. F., Petersen, R. C., Dickson, D. W., Josephs, K. A. Rates of brain atrophy

over time in autopsy-proven frontotemporal dementia and Alzheimer disease. Neuroimage 2008;39:1034–1040. 18. Thompson, P. M., Hayashi, K. M., Dutton, R. A., Chiang, M. C., Leow, A. D., Sowell, E. R., De Zubicaray, G., Becker, J. T., Lopez, O. L., Aizenstein, H. J., Toga, A. W. Tracking Alzheimer’s disease. Ann NY Acad Sci 2007;1097:183–214. 19. Ashburner, J., Friston, K. J. Unified segmentation. Neuroimage 2005;26:839–851.

Chapter 28 Pharmacological Application of fMRI Mitul A. Mehta and Owen G. O’Daly Abstract Magnetic resonance imaging (MRI) allows the assessment of functional changes consequent to drug administration. Two main approaches have been used: changes in functional MRI signal following drug injection compared to the signal prior to injection and changes in task-related brain networks on drug compared to placebo. Here we describe the additional constraints drug studies place on subject selection, study designs and additional technical requirements. Critical issues in the design of statistical analysis routines are described, including the incorporation of peripheral markers of drug action, such as heart and respiration rate, as well as pharmacokinetic data. Finally, we address methods to minimise the potential influence of non-specific drug effects and side effects on the MRI signal allowing interpretation more closely aligned to the precise research questions. Key words: phMRI, psychopharmacology, pharmacokinetics, BOLD, arterial spin labelling, experimental design, physiological confounds.

1. Introduction Functional brain imaging is a powerful technique to map changes in the recruitment of brain regions during different processes and different contexts (e.g. after drug administration). The technique relies on the blood oxygenation level-dependent (BOLD) effect. Neuronal activity leads to an increase in blood flow to the local brain area resulting in an increase in oxygenated haemoglobin relative to deoxyhaemoglobin. Since deoxygenated haemoglobin is paramagnetic, it causes inhomogeneities in the magnetic field leading to increased dephasing of proton spins. The increase in blood supply to active areas is thus indexed by an increased image intensity of T2∗ -weighted MRI images. Psychopharmacological M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_28, © Springer Science+Business Media, LLC 2011

551

552

Mehta and O’Daly

agents are the mainstay treatments across a number of disorders in neurology and psychiatry. While the mechanisms of action at a molecular level may be well understood for many treatments (e.g. levodopa for Parkinson’s disease, selective serotonin reuptake inhibitors for depression), the precise neurophysiological underpinnings of the resulting changes in symptoms and behaviours are poorly elucidated. Molecular assays and PET studies can be performed to define the affinity and occupancy of compounds at different target receptors, but these do not easily translate into descriptions of the alteration in brain information processing networks. This is exemplified by antipsychotic drugs, which can modulate functions of the prefrontal cortex, although the dopamine D2 receptor (the main target of these compounds) is present at much lower concentrations here than within the striatum (1). Thus, systemic drug administration can produce changes in brain function that are distal to sites of high receptor density. This comes as no surprise given that changes in brain activation result from complex and interacting effects within active networks (2). The main challenges of a pharmacological application of fMRI are (1) the signal change is typically very small and thus the power to detect drug-related changes is lower than the power to identify brain networks; (2) the BOLD signal is known to be contaminated by low-frequency noise (1/f noise) presenting problems in the detection of slowly evolving effects of drugs; (3) pharmacological agents can influence the BOLD signal at a neuronal or vascular level, hence complicating interpretation of observed changes. These points are elaborated below. The requirement to maximise sensitivity is paramount in any imaging study design, although the inclusion of an additional factor, such as a drug, is usually associated with the effect to be tested being an interaction effect rather than a main effect. That is the effect of a drug is typically measured on a change in imaging signal (e.g. control task versus active task) rather than on the raw imaging signal. Typically fMRI studies combined with drug administration include a minimum of 15 subjects, whereas many contemporaneous functional activation studies in single groups with no intervention typically include a minimum of 10 subjects. Two main study designs are employed in the pharmacological application of fMRI. The first is the so-called phMRI methodology (3, 4). In this design, the subject is scanned in a constant state (usually at rest, or in the case of experimental animals, anaesthetised), with the drug of interest injected or infused following collection of baseline data. All analyses compare the postdrug volumes with pre-drug volumes which can be BOLD or images of cerebral blood flow or blood volume. In humans an example of this approach is the study of the effects of cocaine

Pharmacological Application of fMRI

553

(5). In a landmark study cocaine was administered 5 min into an 18 min BOLD fMRI scan. Scans from 10 subjects given cocaine were analysed by regressing the temporal profile of subjective changes over the 13 min post-infusion scans showing distinct networks associated with the positive (e.g. rush) and negative (e.g. craving) effects. Profiles of plasma concentration of drug can also be used as a regressor in the analysis e.g. (6) or model-free approaches can be adopted, which have the advantage that they do not presuppose uniformity of a response across the brain e.g. (7). The existence of low-frequency noise in BOLD data acquisition presents the main challenge for this approach. The main solutions are to remove very low frequency components of the signal (e.g. twice the period of the scan time) or to employ linear de-trending prior to analysis. A better approach may be to utilise emerging perfusion imaging techniques which are not susceptible to low frequency noise (8). The second design employed in the pharmacological application of fMRI uses specific contexts in which to study drug effects by appropriating cognitive paradigms aligned to the research questions of interest. Such designs address the issue of brain activation differences between drug and placebo (or control drug). This is exemplified by the study of Sperling and colleagues (9) who compared the effects of scopolamine (an amnesic agent) with lorazepam (included as a sedation-inducing control drug) during a memory encoding task. Activation in the hippocampus was markedly reduced by scopolamine, but not lorazepam. The effects of drugs on the interrelationship between brain areas can also be examined using fMRI, although this has only sparsely been deployed to date. Drug administration in phMRI or cognitive designs may not only influence neural activity and subsequent BOLD signal changes but also influence the vascular signal via vasodilation/constriction (10). The pattern of drug effects typically observed in phMRI and fMRI studies is not consistent with a vasoactive interpretation which would be predicted to produce global changes in signal e.g. (6). Additionally, the outcome for fMRI tasks is the difference in activation between control and active components of sensorimotor and cognitive tasks, thus intrinsically controlling any static vasoactive effects of drugs. The ability to describe and control more selectively for vascular effects of drugs will improve as BOLD calibration capabilities develop (10). The design issues for pharmacological application of MRI data using cognitive tasks overlap with design issues in MRI and fMRI research in general. Although details are provided here, this is a rapidly changing area and thus the reader is also referred to excellent web-based resources that are freely available (e.g. www.mrc-cbu.cam.ac.uk/Imaging).

554

Mehta and O’Daly

2. Materials 1. To a large extent the equipment requirements for functional MRI are the same as for structural studies (is advised that for the most up to date information your scanner manufacturer and software publishers are consulted): ◦ MRI scanner: available 1.5 T or 3 T MRI scanners. ◦ A fast computer with significant RAM and a large storage capacity. 2. Software programs: a number of excellent packages are available. We list the four most commonly used here: ◦ SPM (http://www.fil.ion.ucl.ac.uk/spm/). SPM additionally requires MATLAB (http://www.mathworks. com/). ◦ FSL (http://www.fmrib.ox.ac.uk/fsl/). ◦ Brain Voyager (http://www.brainvoyager.de). ◦ AFNI (http://afni.nimh.nih.gov/afni). 3. Additional equipment required for successful fMRI and pharmacological application of fMRI. ◦ FMRI studies commonly involve participants performing some tasks while in the scanner. This necessitates the ability to present stimuli in the scanner environment. While auditory and visual stimuli are most commonly used, tactile, olfactory and gustatory stimulation are possible. Participants are also commonly required to respond during scan sessions and a wide array of manual (e.g. joysticks or button presses) and verbal response can be collected including the completion of sensitive measures of subjective experience (e.g. visual analogue scales). A number of commercial systems are available, for example, Invivo (http://www.invivocorp.com/fmri/ifis.php) and Nordic NeuroLabs (http://www.nordicimaginglab. com/Products_and_Solutions/nordic_fMRI_solution/ index.aspx), although many laboratories use in-house solutions to allow for greater flexibility in experimental design and align the technology to the precise needs of the research questions. ◦ The BOLD signal (the indirect index of neural activity that underlies fMRI) is known to be extremely sensitive to changes in a number of physiological parameters (i.e. blood pressure, pulse, oxygen saturation, respiration rate and depth) and there is a growing appreciation of the need to account for these influences. Currently, MRI-compatible pulse-oximeters are commonly used and

Pharmacological Application of fMRI

555

in many cases are integrated into the scanner hardware. A number of commercial systems are available, for example Nonin (http://www.nonin.com/PulseOximetry) and NewMatic (http://www.newmaticsound.com/ccp0catshow/PulseOx.html). Many scanners also include belts (bellows) which measure breathing-related changes in abdominal and/or chest circumference to track both respiration rate and depth. Furthermore, while special sphygmomanometers can be deployed within the scanner field they are less commonly used. Careful measurement of these parameters permits controlling for their influence during statistical analysis (11). ◦ Similarly, eye-tracking pupillometry measures can ensure that participants are alert, actively engaging in task performance and obeying instructions (i.e. fixating).

3. Methods 3.1. Subject Recruitment

The general procedures for recruiting participants for MRI studies involve ensuring participants align with local MRI safety policies (i.e. presence of metal from surgery, accidents) and can tolerate the scanner environment (i.e. screened for claustrophobia). When recruiting subjects for pharmacological fMRI a number of additional parameters need to be considered: 1. Prior to administering any compound all participants must be rigorously screened to ensure that they are physically healthy and that the chance of adverse events are minimised. This will include a general physical examination, an electrocardiograph and the collection of blood and urine samples to screen for potential liver, kidney or cardiovascular indicators of poor health; particularly for abnormalities which may influence the absorption, metabolism and clearance of drugs. 2. It is common practice to not recruit individuals with a personal history of any psychiatric or neurological illness. However, when drugs are to be administered it may be necessary to also exclude people with a family history (i.e. first degree relative) of psychiatric or neurological illness. This is primarily due to their potentially high loading of risk alleles. Such latent risk, while having little notable impact on cognition or health, this could render individuals more sensitive to a drug or vulnerable to certain side effects (e.g. psychotomimetic action). 3. Similarly, genetic polymorphisms can also strongly influence a drug’s pharmacokinetics and pharmacodynamics. For

556

Mehta and O’Daly

example, a number of agents (e.g. atomoxetine and paroxetine) are metabolised by a Cytochrome P450 CYP2D6 pathway. A polymorphism in the gene encoding this enzyme leads to poor metabolism of these compounds in approximately 5–10% of the Caucasian population and thus a significantly elevated plasma concentration for a much longer duration than normal. A participant who tests positive for this polymorphism may not, however, need to be excluded as such information can be used to help explain and perhaps control for significant sample outliers. 4. Prior or concurrent drug exposure also needs to be assessed. While all subjects should be screened for illicit substance use (a urine drug screen prior to dosing at every session), it is also important to control for individuals who are currently using legal drugs (i.e. nicotine, ethanol, caffeine) to excess (see Note 1). In this case, there are three concerns. First, these individuals may be dependent upon these substances, and the associated neuroadaptions may influence the neural response to the compound of interest. Second, they may begin to experience withdrawal from these compounds during study sessions. Third, these individuals may fail to abstain on study days and thus alcometers and respiratory carbon monoxide measures can be collected on each session. Generally, these issues can be addressed at initial screening and by excluding individuals who regularly drink more than the recommended weekly alcohol allowance or who binge drink. Individuals who smoke more than five cigarettes per day or who drink more than four cups of coffee daily can also be excluded (12). 5. For pharmacological studies, it is common practice to exclude those who have recently participated in trials of “investigational compounds.” This is justified because these drugs are not commercially available and little information regarding their actions is available. To avoid participants simultaneously taking part in multiple studies, centres in the UK can check a national register of clinical trial volunteers (http://www.nationalvolunteerregister.org/) simply by entering a potential participant’s details (including passport details). In the case of pharmacological fMRI studies involving medicated patients, the potential for drug–drug interactions warrants caution for three reasons: First, adverse side-effects may arise from such interactions. Second, they may confound the results of interest, particularly with a sensitive, but indirect index of neural activity, such as BOLD. And third, the additional metabolic load of removing two compounds may influence the pharmacodynamics

Pharmacological Application of fMRI

557

of both, the consequence of which may be difficult to ascertain without additional control sessions. 3.2. Study Design

Pharmacological fMRI studies commonly use both withinsubjects (repeated measures) and between-subject designs. However, while the former is significantly more popular in human studies (exceptions include phase 3 clinical trials that contrast effects against other available treatments), the necessity to coadminister anaesthetics for animal studies, particularly in rodent studies, has led to favouring of cross-sectional designs. Furthermore, if the between-subjects variance is likely to be significantly greater than the within-subject variability then a within-subjects design is advised. A number of additional concerns can motivate the choice of study design. For example, repeated measures designs are potentially susceptible to session confounds and order effects related to learning, particularly when tasks are being employed and habituation to scanner exposure (i.e. reduced anxiety/stress responses or diminution novelty). Similarly, residual psychological or pharmacological (metabolite-related) effects must also be minimised. There are a number of approaches to deal with these issues: 1. Use suitably long wash-out periods (i.e. minimum of 1 week, although longer periods may be necessitated due to long drug half-lives). 2. Habituate individuals to as many aspects of the study protocol as possible at a screening vist. This can include training on all paradigms exposure to the scanner (or a scanner simulator). 3. Repeated measures designed should be fully counter balanced. 4. For cross-sectional designs, it is important that both groups are matched for potentially confounding variables (e.g. age, IQ). In pharmacological studies, it is important to pick an appropriate control treatment against which to test the effects of your compound of interest. Most commonly an inert compound (e.g. ascorbic acid) is administered as a placebo. However, if the compound of interest is known to cause significant psychological (e.g. sedative or psychostimulant) or systemic (e.g. gastrointestinal disturbances), it may be desirable to control for these effects which may (a) unblind the subject as to which compound they were receiving and (b) confound the brain maps (conflating the neural effects and central representation of these peripheral/psychological effects). One approach to this is to use a lower dose of the compound of interest and a separate placebo (three study arms), so that a dose–effect may permit dissociation of neuronal and confounding effects. Alternatively, a second drug which

558

Mehta and O’Daly

has a similar side-effect profile could be used instead of a low-dose compound. In all cases, where acute adverse responses are not anticipated, a double-blind administration of the placebo/control is advised. However, in cases where the drug might have rare, but potentially serious acute effects, a single-blind approach may be more appropriate to facilitate a speedy response (i.e. without the need to break a randomisation code prior to administration of rescue medication or contacting emergency services). Finally, an individual’s response to a drug of interest may be influenced by day-to-day and diurnal variation in baseline physiology. One approach to deal with day-to-day changes in baseline involves collecting pre-dose scans that will be used to control scans for post-dose effects. In other words treatment effects will reflect not the difference between two sessions, but the betweensession difference in the relative (treatment-related) changes from baseline. Controlling for diurnal variation is more difficult, but one general approach is to ensure that each subject completes all their study sessions at the same time of day, thus the withinsubject effect of interest will not be confounded by diurnal effects. Additionally, there may be some value in collecting control data (e.g. placebo treatment) in more than one study arm. Multiple placebo study arms allow for the assessment of the test–retest reliability of post-dose scans which may be sensitive to habituation and expectancy effects. Furthermore, when balanced placebo sessions (to avoid confounding order effects) are used a more stable estimate of the placebo effect may be obtained. 3.3. Pre-processing

The general pre-processing procedures for fMRI are discussed in detail in many textbooks (13, 14). These steps typically include 1. correction for the effects of head movement (realignment of a volume time-series) 2. normalisation to a standard space 3. spatial smoothing of the data with a Gaussian kernel Additionally, it is possible to correct for slice-to-slice differences in the timing of BOLD sampling, although some packages make these adjustments to the statistical model rather than the data. Here we discuss some of the primary issues regarding preprocessing of fMRI data which are relevant to pharmacological studies.

3.3.1. Movement Correction

Commonly the first step of pre-processing is to realign all volumes within a time-series so that they are co-registered. This typically involves the calculation of a set of rigid-body spatial transformations (a subset of general affine transformations), including translations and rotations around the X-, Y- and Z-axes. Additional zoom and shear parameters can also be estimated and

Pharmacological Application of fMRI

559

applied to the data. While these approaches are relatively successful, there is a growing appreciation of potential for residual movement-related artefact due to a failure of the algorithm or the apparently successful correction of large movements. In this latter case, the data may no longer be reliable in some regions (due to distortions related to field inhomogeneties) and thus statistical analyses maybe adversely affected. Approaches to this issue include examining the contribution and correction of movement artefacts (TSdiffana: http://imaging. mrc-cbu.cam.ac.uk/imaging/DataDiagnostics; and ArtRepair: http://spnl.stanford.edu/tools/ArtRepair/ArtRepair.htm), isolating movement-related artefacts and creating new motioncleaned time-series (e.g. the FSL ICA toolbox – MELODIC – can be used to remove “noise” components from the data: http://www.fmrib.ox.ac.uk/analysis/research/melodic/), controlling for the influence of movement-corrupted volumes on the estimation of beta coefficients (model fit parameter estimates) using robust weighted least squares estimation (http://www.icn. ucl.ac.uk/motorcontrol/imaging/robustWLS.html). 3.3.2. Spatial Normalisation

Normalisation is the process whereby all the subject data is brought into a standard reference space to permit grouping and comparison. When normalising data to standard space one must decide whether or not to permit nonlinear (warping) parameters to be estimated and applied to the data or to only use affine transformations (similar to those used in movement correction). All things being equal, use of nonlinear normalisation based on the use of spatial priors of tissue distributions (e.g. grey matter masks) should provide better normalisation. However, the nonlinear parameters may warp data inappropriately to correct EPI distortion artefacts. This issue of distortion also impacts on the use of an intermediate structural image for normalisation. This approach assumes that (i) a high-resolution structural image will give more accurate normalisation; (ii) a good linear coregistration of the functional data to the structural; and (iii) poor normalisation parameter estimation due to differential image distortion between the functional and structural data is tolerable or systematic. One important point for multi-session studies, commonly used in pharmacological application of fMRI is whether each time-series should be normalised independently, or whether the same parameters should be used for all sessions of any one subject. This avoids between-session differences in normalisation being mistaken for between-session treatment effects. To minimise this, an intermediate high-resolution gradient echo (GE) image (with similar distortion properties to the functional timeseries) can be employed to calculate the normalisation parameters which can then be applied to data from all co-registered

560

Mehta and O’Daly

functional imaging sessions. Thus, a normalisation procedure for a multi-session pharmacological MRI study would comprise the following steps: • Co-registration of the mean image derived from a movement-corrected functional time-series to a highresolution GE-echoplanar image (EPI). • Estimation of spatial normalisation parameters from the high-resolution scan to an EPI template in a standard space (available in most image analysis packages, but see http://www.loni.ucla.edu/ICBM/ for more details on efforts to create standards). • Application of spatial normalisation parameters to the coregistered function time-series. • Application of the same spatial normalisation parameters to further co-register time-series from the same subject. 3.3.3. Spatial Smoothing

Prior to statistical analysis, the data is typically smoothed using a Gaussian kernel. This smoothing serves three primary purposes. First, it minimises individual differences (increased overlap) in regional signal distribution due to imperfect normalisation. Second, statistical parametric mapping is predicated on the assumption that the data is normally distributed. Unfortunately, with fMRI data this assumption may be violated. Smoothing of the data renders the variance of the data more normally distributed. Finally, by averaging the signal over a number of voxels (bias against high-frequency and in favour of a low-frequency signal) increases the signal to noise ratio. According to the matched filter theorem, the size of the smoothing kernel used should match the size of the effect expected. For example, for cortical effects smoothing by a 10 mm full width half maximum (FWHM) kernel is probably appropriate, but would bias signal in smaller subcortical regions, making it less likely that effects are seen in these regions (where the use of a 6 mm FWHM kernel may be more appropriate). In practice, a smoothing kernel of 8 mm (FWHM) is commonly used in SPM to balance the sensitivity in both regions. Additionally, if comparing or contrasting different imaging modalities (i.e. BOLD versus arterial spin labelling), it is important to use the same smoothing kernel even though ASL is more sensitive to small capillary beds than large draining veins and thus the extent of any signal change will theoretically be smaller.

3.4. Statistical Inference

Currently, the majority of fMRI analysis packages take a mass univariate approach involving regressing of a design matrix of explanatory variables (EV) against the time-series of every voxel in isolation. Prior to making any inference, the resulting parameter estimates (i.e. an estimation of the degree of correspondence between each EV and the time-series) must undergo some

Pharmacological Application of fMRI

561

correction for the likely false positives arising from testing a very large number of voxels at an arbitrary p-value, typically 0.05. As described in the previous chapter, family-wise error (FWE) correction while correctly controlling the voxel-wise false-positive rate, may be considered stringent to distributed changes of small effect size and thus may be insensitive to subtle, distributed effects of drugs on brain responses. However, many imaging packages allow one to threshold on the basis of the number of voxels in a cluster (i.e. cluster size). That is, these packages calculate the minimum cluster extent which would be deemed unlikely by chance. This measure is more sensitive to effects arising from a more widespread modulation of activity, such as that predicted to occur when neuromodulatory systems are modulated. 3.4.1. Single Subject (first Level) Models

As described above, the mass-univariate analysis approach involves regressing EVs, which encode the timing of events of interest (e.g. stimulus presentation, time of response) and nuisance regressors (e.g. movement parameters, regressors encoding respiratory parameters) against voxel time-series. Currently, there are two major approaches to applying fMRI in a pharmacological context, each examining a different aspect of a drug’s effect on brain function. The first simply explores the degree to which a drug modulates the task-related neural recruitment, whereas the second, commonly termed phMRI (3), involves scanning in a single state (e.g. at rest) while a drug is administered intravenously. Whereas the former aims to identify drug-related modulation of cognitive systems, the latter aims to identify the response profile for drug-sensitive brain regions. Typically, if subjects are performing a task, the individual design matrix is the same as that when no drug is used. This is particularly common when drugs are administered orally; participants are usually scanned on two occasions (i.e. randomised placebo controlled crossover design). However, for the phMRI model there are a number of options: treat the preand post-dose scans as two separate epochs (i.e. look for changes in signal sustained over the epoch); break the long epochs into manageable periods (e.g. 1 min time bins); or alternatively regress an input function encoding, for example, the predicted or measured plasma concentrations. The resulting maps will represent (i) the brain regions where there is a significant between-epoch difference in sustained BOLD signal, (ii) brain regions displaying any changes in minute-by-minute BOLD signal compared to predose levels or (iii) the brain regions where the response profile significantly correlate with the drug plasma concentration.

3.4.2. Group Modelling

The output of first-level (individual subject) models will commonly be taken forward to a second (group) modelling phase. Various model scans be used to test for effects of interest: one sample t-tests for single session group maps or paired t-tests to

562

Mehta and O’Daly

compare placebo and drug session differences in signal. An ANalyses Of VAriance (ANOVA) approach is commonly employed for multi-period studies (>2 study periods). In our experience, the sensitivity of repeated-measure designs is significantly enhanced by the inclusion of EVs encoding individual-subject effects (between-subject variability) reducing the amount of unexplained signal in the data (that is including a subject factor in any model). In addition to whole-brain hypothesis testing, region of interest (ROI)-based analyses are commonly employed to examine drug action in regions hypothesised a priori to be sensitive to the intervention (i.e. regions heavily innervated by a target neurochemical system or known from previous investigations in humans or experimental animals as sensitive targets). This involves extracting mean signal over all the voxels within an ROI from each subject’s first level model. While there are many approaches to this, the MARSBAR MATLAB Toolbox (http://marsbar.sourceforge.net/) is a particularly popular package. The mean signal for each subject for the condition, or contrast of interest, can be taken to a separate statistical analysis package [e.g. SPSS (SPSS Inc.) for further exploration]. 3.4.3. Testing a Priori Hypotheses (Regions of Interest Analyses)

In most cases, researchers will have a priori hypotheses regarding which brain areas will be sensitive to the administration of a drug. Region of interest (ROI) analyses are a useful approach to testing detailed predictions about the data from a specific brain region. There is a growing appreciation that great care is required to ensure that the selection of the voxels which constitute an ROI is unbiased. The key point is that the specification of an ROI should be independent of the comparison it will be used to test (15). Typical errors include finding a non-significant effect (i.e. an effect that is no longer significant following correction for multiple comparisons), extracting the mean signal over all the voxels in that cluster (cluster-derived ROI) and then carrying out a separate test in another statistical software package. A second error is to define the ROI voxels from an effect in one session, or group, and then use it to test for session or group differences. Union maps (the sum of regions showing a task-related effect across sessions, or groups, forms the ROI) may be more suitable. However, while theoretically appropriate, it is possible to have a large effect in a brain area in one group/session that does not overlap with the effect in another group/session, and the ROI would be biased in terms of testing a difference in signal intensity or voxel count. ROIs defined from cluster overlaps would not suffer from this. Alternatively ROIs can be defined anatomically or from independent data (e.g. a separate group). A recent novel approach proposed by Poldrack (16) is to remove one subject from a group, run the analysis and define the ROI from clusters in that analysis,

Pharmacological Application of fMRI

563

this ROI will be used to extract data from the subject who was left out. Following this, the subject is returned to the group and the process is repeated with another subject removed. This yields an unbiased ROI for each individual as their data was not included in the maps used to define the ROI used to extract data. 3.5. Pharmacokinetic Data

Pharmacokinetic data refers to drug liberation (e.g. controlledrelease formulations), absorption, distribution, metabolism and excretion from the body. The variance in pharmacokinetics may be reflected in the variance in the so-called pharmacodynamic effect. In the case of MRI, this would represent the change in BOLD or perfusion signal in different brain areas. There is an assumption here that the blood pharmacokinetics reflect brain pharmacokinetics and this may not be true with some compounds. For example, opioidergic compound are typically highly lipophilic and are rapidly absorbed and taken up into tissue, including brain tissue. As such, there is a dissociation with brain and plasma pharmacokinetics (17). Dissociation between plasma and brain pharmacokinetics have also been demonstrated for oral (antipsychotic) drugs, where the reduction in receptor occupancy over time lags behind the reduction in plasma concentration (18). With this assumption of equivalence between plasma and brain pharmacokinetics, the variance in drug levels (as measured in blood) and distribution (as may be measured with positron emission tomography or ascertained from receptor distribution) may lead to a variance in the MRI signal (19). An important question here is, given the range of plasma drug levels what type of model describes the relationship between drug levels and the MRI signal? Linear models may be sufficient, but considering the constrained physiological systems being modulated other curvilinear models may also be appropriate (e.g. Emax models). Modelling of the relationship between pharmacokinetics data and MRI data thus requires analysis routines capable of describing these different possible relationships, such as available in SigmaPlot (Systat Software Inc.). A practical implication of drug pharmacokinetics is that this defines the window of opportunity to detect drug effects in the brain, so this must be considered when assembling the imaging protocol.

3.6. Conclusions

The requirements for pharmacological studies combined with MRI stretch beyond the needs described for structural MRI. Additional equipment to record physiological variables (heartrate and respiration rate) within the MR environment are widely available and may even be supplied by scanner manufacturers. Once this data is synchronised with the MRI data, it can be used in regression analyses to control for their effects on the MRI signal.

564

Mehta and O’Daly

The subjects themselves also contribute to additional variance in the data and this can be reduced through careful screening and definition of inclusion/exclusion criteria to minimise differences between subjects in drug absorption, metabolism and clearance and avoid unknown or known drug–drug interactions. The number of sessions and scans each subject experiences in pharmacological MRI studies requires careful consideration. To summarise, the main issues are whether (i) to control for day-to-day variability and (ii) to incorporate multiple drug or placebo arms to assess the stability of effects. These decisions have significant implications for the study timelines and cost. While some procedures may provide a stable output over multiple sessions (e.g. sensorimotor tasks), others may not (e.g. rule learning, stress induction) and thus the precise research questions must be incorporated in the decision-making process of the study design. The choice of drug will often be determined by the research questions and the dose must be carefully chosen. One issue is that of side effects, such as nausea, stomach pain, increased heart rate, and drowsiness. These can have two important consequences. First, during cognitive activation studies, these side effects may act as non-specific “distractors” to subjects and thus changes in brain activation may not be representative of the effects of the drug on cognitive networks. Indeed, the second implication is that the results of the study may be exquisite maps of the brain networks underpinning the experience of the side effects. This is problematic unless your research question concerns the side effects per se.

4. Notes 1. In some cases previous experience with a drug may be beneficial, for example, when the drug has psychotomimetic properties (e.g. cannabis) some brief prior experience may limit the likelihood of individuals reacting aversively to the experience, although the presence of the drug during the study may want to be controlled. References 1. Honey, G. D., Bullmore, E. T., Soni, W., Varatheesan, M., Williams, S. C., Sharma, T. Differences in frontal cortical activation by a working memory task after substitution of risperidone for typical antipsychotic drugs in patients with schizophrenia. Proc Natl Acad Sci USA 1999;96:13432–13437.

2. Logothetis, N. K. What we can do and what we cannot do with fMRI. Nature 2008;453:869–878. 3. Leslie, R. A., James, M. F. Pharmacological magnetic resonance imaging: A new application for functional MRI. Trends Pharmacol Sci 2000;21:314–318.

Pharmacological Application of fMRI 4. Steward, C. A., Marsden, C. A., Prior, M. J., Morris, P. G., Shah, Y. B. Methodological considerations in rat brain BOLD contrast pharmacological MRI. Psychopharmacology 2005;180:687–704. 5. Breiter, H. C., Gollub, R. L., Weisskoff, R. M. et al. Acute effects of cocaine on human brain activity and emotion. Neuron 1997;19:591–611. 6. Stein, E. A., Pankiewicz, J., Harsch, H. H. et al. Nicotine-induced limbic cortical activation in the human brain: A functional MRI study. Am J Psychiatry 1998;155: 1009–1015. 7. Deakin, J. F., Lees, J., McKie, S., Hallak, J. E., Williams, S. R., Dursun, S. M. Glutamate and the neural basis of the subjective effects of ketamine: A pharmaco-magnetic resonance imaging study. Arch Gen Psychiatry 2008;65:154–164. 8. Olson, I. R., Rao, H., Moore, K. S., Wang, J., Detre, J. A., Aguirre, G. K. Using perfusion fMRI to measure continuous changes in neural activity with learning. Brain Cogn 2006;60:262–271. 9. Sperling, R., Greve, D., Dale, A. et al. Functional MRI detection of pharmacologically induced memory impairment. Proc Natl Acad Sci USA 2002;99:455–460. 10. Iannetti, G. D., Wise, R. G. BOLD functional MRI in disease and pharmacological studies: Room for improvement? Magn Reson Imaging 2007;25: 978–988. 11. Glover, G. H., Li, T. Q., Ress, D. Image-based method for retrospective correction of physiological motion effects in

12.

13.

14.

15.

16. 17.

18.

19.

565

fMRI: RETROICOR. Magn Reson Med 2000;44:162–167. Field, A. S., Laurienti, P. J., Yen, Y. F., Burdette, J. H., Moody, D. M. Dietary caffeine consumption and withdrawal: Confounding variables in quantitative cerebral perfusion studies? Radiology 2003;227:129–135. Huettel, S. A., Song, A. W., McCarthy, G. Functional Magnetic Resonance Imaging, 2nd ed. Sunderland, MA: Sinauer Associates; 2009. Jezzard P., Matthews P. M., Smith S. M., eds. Functional Magnetic Resonance Imaging: An Introduction to Methods. New York, NY: Oxford University Press; 2002. Kriegeskorte, N., Simmons, W. K., Bellgowan, P. S., Baker, C. I. Circular analysis in systems neuroscience: The dangers of double dipping. Nat Neurosci 2009;12: 535–540. Poldrack, R. A. Region of interest analysis for fMRI. Soc Cogn Affect Neurosci 2007;2: 67–70. Melichar, J. K., Nutt, D. J., Malizia, A. L. Naloxone displacement at opioid receptor sites measured in vivo in the human brain. Eur J Pharmacol 2003;459:217–219. Tauscher, J., Jones, C., Remington, G., Zipursky, R. B., Kapur, S. Significant dissociation of brain and plasma kinetics with antipsychotics. Mol Psychiatry 2002;7: 317–321. Muller, U., Suckling, J., Zelaya, F. et al. Plasma level-dependent effects of methylphenidate on task-related functional magnetic resonance imaging signal changes. Psychopharmacology 2005;180:624–633.

wwwwwww

Chapter 29 MRI of Neuronal Plasticity in Rodent Models Galit Pelled Abstract Modifications in the behavior and architecture of neuronal networks are well documented to occur in association with learning and memory, as well as following injury. These plasticity mechanisms are crucial to ensure adequate processing of stimuli, and they also dictate the degree of recovery following peripheral or central nervous system injury. Nevertheless, the underlying neuronal mechanisms that determine the degree of plasticity of neuronal pathways are not fully understood. Recent developments in animaldedicated magnetic resonance imaging (MRI) scanners and related hardware afford a high spatial and temporal resolution, making functional MRI and manganese-enhanced MRI emerging tools for studying reorganization of neuronal pathways in rodent models. Many of the observed changes in neuronal functions in rodent’s brains following injury discussed here agree with clinical human fMRI findings. This demonstrates that animal model imaging can have a significant clinical impact in the neuronal plasticity and rehabilitation arenas. Key words: Neuronal plasticity, animal models, functional MRI, manganese-enhanced MRI, learning and memory, nerve injury.

1. Introduction The brain has a tremendous capability to adapt itself to internal and external events. The ability of neurons to change their internal properties and of neuronal networks to reshape their connections is referred to as plasticity. The outcome of these plasticity changes can affect the time it takes the brain to process a specific stimulus and generate a suitable response. Therefore, appropriate rewiring of neuronal connections during development and in adulthood is crucial to ensure proper and adequate propagation and processing of stimuli. One of the fundamental goals in M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_29, © Springer Science+Business Media, LLC 2011

567

568

Pelled

neuroscience is to determine the genetic, molecular, systemic, and environmental basis of plasticity changes. The blood oxygenation level-dependent (BOLD) functional MRI (fMRI) method enables the detection of hemodynamic changes due to changes in neural activity throughout the brain. Human fMRI has had a major impact in cognitive neuroscience and neurosurgical planning, where an emphasis is on the role of plasticity in recovery and maintenance of brain functions in a wide range of diseases. Indeed, a revolution has occurred indicating that extensive and widespread plasticity occurs in the adult brain. For example, in people that are blind from an early age, Braille “reading” leads to activation of the visual cortex (1, 2)! In stroke victims, stimulation of the affected hand can lead to fMRI responses in ipsilateral brain regions that are normally not activated (3). It is now apparent that plasticity in the adult brain can involve longrange modifications in the function and architecture of neuronal networks. Numerous human fMRI studies demonstrated altered neuronal processing and activation in learning and memory paradigms, as well as in pathological conditions. However, the vast knowledge of the different neuronal mechanisms that play a role in plasticity originates in animal models. The ability to manipulate normal neuronal functions in rodents using genetic, molecular biology, and neurosurgical tools and to monitor the consequent changes in neuronal behavior makes rodent models the primary preference when investigating the underlying mechanisms of neuronal plasticity. Multi-unit electrophysiological techniques have been the most popular methodology to investigate modification of neuronal networks in animal models. This technique provides unmatched spatial and temporal resolution. However, a major drawback is that information can only be obtained from one or two localized areas and from a limited number of neurons at a time. Therefore, many aspects of cortical reorganization, such as global network rearrangements, altered performance of a complex neuronal network, and changes in region to region functional connectivity, have been difficult to assess. Over the past decade, new developments in MRI of rodents afford a spatial resolution of approximately 100 μm and a temporal resolution for functional changes in the order of 500 ms. MRI is hence an emerging tool for studying plasticity in animal models. Rodent models are most common for non-human imaging associated with plasticity. Changes in the spatial localization and the magnitude of fMRI responses were observed in rodent models following lesions in the central nervous system (4–6) and injuries of the peripheral nervous system (7–10). Similar to human fMRI studies, it is apparent that reorganization of neuronal pathways

MRI of Neuronal Plasticity in Rodent Models

569

following injury in the rodent brain is reflected by the fMRI responses. In addition to fMRI, there is growing interest in developing new molecular imaging probes to study brain function and connectivity, which are predicted to greatly expand the range of information available from MRI. Manganese-enhanced MRI (MEMRI) contrast is based on the ability of manganese ions to trace neuronal pathways and have been used in several animal models to detect long-term modifications of neuronal connections in response to learning (11–13) or following an injury (14–17).

2. Materials 2.1. Animal Model 2.1.1. Neuronal Plasticity in Learning and Memory

Learning and memory processes in the developing and adult brain are often accompanied with architectural and functional modifications of neuronal networks. These plasticity changes occur within seconds (short-term) and could last for days, weeks, and a lifetime (long-term). Human functional imaging studies have demonstrated robust differences in fMRI responses of corticaland subcortical-related regions associated with learning and memory tasks. During these functional imaging studies, the subjects are alert and can perform a variety of complex tasks that involve the activation of multisensory modalities, such as cognition, visual, motor, and tactile (18–20). Thus, the effect that learning and memory processes have on changes of functional cortical responses associated with short- and long-term neuronal plasticity can be evaluated. In contrast, the majority of functional imaging studies of nonhuman subjects are performed under general anesthesia. This hampers the ability of the animal to execute behavioral tasks related to learning or memory. Therefore, neuronal plasticity changes associated with learning and memory have been so far demonstrated on animal models that exhibit long-term and massive morphological modifications of neuronal architecture and functions (11–13) (Note 1).

2.1.2. Neuronal Plasticity in Pathology

Injury to the central or peripheral nervous system leads to modifications in the behavior of cortical and subcortical neuronal networks that may compensate for loss of function. Depending on the time (during development or in adulthood) and the location (i.e., cortical, subcortical, spinal cord, or peripheral nerves) that

570

Pelled

the injury has occurred, different neuronal mechanisms may be involved in reshaping the behavior of the relevant neuronal pathways. Altered functional responses of cortical neurons associated with cortical reorganization following injury have been demonstrated in stroke (Note 2), Parkinson’s disease, and peripheral nerve injuries in rodent models using fMRI. Changes in the functional architecture of neuronal pathways associated with reorganization of cortical and/or subcortical pathways have been demonstrated in stroke (4, 5, 17, 21), peripheral nerve injury (7, 8, 10, 22), Parkinson’s disease (6, 16), traumatic brain injury (14), lesions to olfactory pathway (23), whisker removal (9), and epilepsy (15) using MEMRI-based track-tracing methods. 2.2. Anesthesia

In order to prevent head movement that will interfere with MRI signals and to reduce stress, animals are usually anesthetized during scanning. For fMRI studies, only a small number of drugs have been shown to minimize the neurovasculature coupling interference caused by the anesthetic. 2.2.1 For fMRI studies: A bolus (80 mg/kg) and a continuous (27 mg/kg/h) I.V. injection of alpha-chloralose (in PBS) (Sigma-Aldrich). To prevent motion pancuronium bromide (4 mg/kg, AmerisourceBergen) is administrated once per hour (Note 3). 2.2.2 For MEMRI studies: 2% isoflurane administrated through a nose cone (24).

2.3. Physiological Monitoring Equipment

Anesthesia, especially for prolonged times, can affect normal physiological functions. Appropriate monitoring of animal physiology is required. 2.3.1 Temperature, heart rate, and breathing rate should be continuously monitored with specialized MRI dedicated monitoring systems. 2.3.2 The functional MRI signal is sensitive to changes in blood pH, partial O2 and CO2 , and blood pressure. The following monitoring instruments could be used to determine these physiological levels – Oxygen saturation probes to continuously monitor oxygen saturation, heart rate, respiration rate. – A blood gas analyzer machine to determine pH, partial O2 , and CO2 levels of <100 μl arterial blood samples. – A blood pressure monitor for arterial blood pressure.

2.4. Manganese Chloride Concentrations for MEMRI

Changes in neuronal functions and reorganization of neuronal pathways associated with plasticity can be visualized with MEMRI (Note 4). 2.4.1 Direct stereotactic administration of isotonic manganese chloride (MnCl2 ) to the brain using animal-dedicated

MRI of Neuronal Plasticity in Rodent Models

571

stereotactic frames is used to trace a specific neuronal pathway. 2.4.2 Continuous systemic (I.V. or I.P.) administration of isotonic MnCl2 over the course of an hour is used to detect morphological changes associated with large-scale neuronal pathway reorganization. 2.5. Stimulation for fMRI

Depending on the lesion location in the central or peripheral nervous system, neuronal plasticity of the somatosensory cortex may be evaluated using fMRI responses resulting from peripheral stimulation. 1. Tactile stimulation: Two thin needle electrodes are inserted between the digits of forelimbs and/or hindlimbs. Stimulation of 3 Hz, 300 μs, 2–3 mA repeated for 10–30 s is ideal for alpha-chloralose anesthesia (25). 2. Whisker stimulation: A piezoelectric device was used to drive a small comb to generate movement of whiskers in one side of rats’ face (26). A broader stimulation of the rat’s whisker pads was obtained with small needle electrodes placed directly in that region (9). 3. Visual stimulation: MRI compatible light-emitting diodes were placed proximate to the rat’s eyes to deliver light stimulation (27).

3. Methods 3.1. Induction of Neuronal Plasticity Associated with Learning and Memory

3.1.1 Plasticity in the auditory system: Sound stimulation was delivered in an acoustic isolation chamber. Adult or neonatal mice were exposed to a single (16, 32, and 40 kHz) or 2-tone (16 + 40 kHz) stimuli for 24 h. Imaging of increases in activity in auditory-related neuronal pathways was preformed immediately (if stimulation was over 24 h) or in the days following stimulation (if stimulation was repeated daily) (11, 13). Figure 29.1 shows T1 signal enhancements corresponding to neuronal architectural modifications in the inferior colliculus in response to pure-tone stimulation. 3.1.2 Plasticity in the visual system: A continuous visual stimulation was composed of moving square waves on several computer monitors inside the rodents’ cages and delivered for 8 h. Modifications of neuronal activity in the rats’ visual cortex were assessed using MEMRI immediately following termination of stimulation (12).

572

Pelled

Fig. 29.1. Pure-tone stimulation induced tonotopic signal enhancement in the inferior colliculus (IC). Averaged coronal IC images (n ≥ 8 in each group) demonstrated spatial differences in MEMRI enhancement between mice kept in a quiet environment (a) and mice exposed to either 16 (b) or 40 kHz (c). Pseudocolor-coded images (d–f) made it easier to appreciate the enhancement patterns (color scale included). t-Test analysis, comparing either the 16-kHz-exposed group to the quiet controls (green) (g) or the 40-kHz-exposed group to the quiet controls (red) (h), shows the regions of significant differences in MEMRI enhancement (p < 0.05) in general agreement with the established electrophysiological tonotopic map. Figure from (11) copyright (2009) National Academy of Sciences, USA.

3.2. Inducing Neuronal Plasticity Associated with Pathology

Stroke: Two hours of occlusion of the middle cerebral artery was performed on adult rodents. Imaging of cortical reorganization took place from the day following stroke to a few weeks later (4). Parkinson’s disease: Stereotactic brain injection of 6-OHDA into the substantia nigra pars compacta was performed in adult rats. Modifications in neuronal track tracing and fMRI responses were assessed 2 weeks following stereotactic procedures, when significant loss of dopaminergic neurons was established (6). Traumatic brain injury (TBI): A fluid percussion device was used to deliver pressure pulses into exposed cortical regions. Imaging took place from the day following TBI to a few weeks later (14). Epilepsy: Kainic acid was injected I.P. to induce seizures. Imaging of hippocampal activity associated with epilepsy took place 2 weeks following seizure induction (15). Peripheral nerve deafferentation: Complete denervation of the hindlimb or forelimb by dissecting the limb’s main nerves was performed in juvenile (7) or adult (10) rats. Modification

MRI of Neuronal Plasticity in Rodent Models

573

in functional neuronal architecture and responses were measured in the weeks following nerve deafferentation. Peripheral nerve crush: Crushing the nerves located in rodent’s forelimb or hindlimb will result in regeneration of the nerve. Imaging the return of cortical activity associated with regeneration took place in the days and weeks following nerve injury (8). Whisker removal: Follicle ablation of all rat’s whiskers was performed at postnatal day 10. Imaging reorganization of sensory representations was preformed 8 weeks following whisker removal (9). 3.3. Imaging Neuronal Plasticity Methods

3.3.1 fMRI: Functional modification of neuronal behavior in the somatosensory cortex can be evaluated with fMRI with the following parameters (Note 5): Anesthetized rodents receive stimulation as described above. Gradient-echo or spin-echo T2∗ -weighted EPI images are acquired with long echo time (TE) (15–30 ms) and short repetition time (TR) (100– 1,000 ms) depending on field strength. Number of image acquisition is dependent on the stimulation paradigm. 3.3.2 Resting-state fMRI: Functional modifications of neuronal networks that result in changes in the functional connectivity between different brain regions were observed in rat models for peripheral nerve injury (10) and stroke. Similar sequence parameters to fMRI protocols are generally used. A temporal correlation of neuronal oscillation during rest can be calculated using various approaches, such as pre-selected seed voxel analysis and power spectral analysis of different frequency peaks (28). 3.3.3 MEMRI: Anesthetized animals are imaged with the following parameters: Gradient-echo or spin-echo T1weighted sequences with short TR (<300 ms) and short TE (<15 ms) (29–31) depending on field strength. Multiple averaging is performed to increase signal-to-noise ratio.

3.4. Evaluation of Plasticity 3.4.1. Differences in fMRI Responses

In healthy rodents, stimulation of a limb leads to fMRI activation of the contralateral somatosensory cortex with minimal activity in the ipsilateral somatosensory cortex. Cross-correlation of BOLD signal with stimulation paradigm and t-test analysis are most common to detect and calculate fMRI characteristics. Following central or peripheral nervous system injury, the magnitude and/or the location of fMRI responses may be altered.

574

Pelled

Fig. 29.2. a Group-averaged functional MRI t-test maps overlaid on echo-planer MRI of primary somatosensory cortex (SI) activation following forepaw stimulation in sham-operated and denervated rats. In sham-operated rats (n=7), stimulation of a forepaw resulted in significant contralateral SI activation with minimal ipsilateral SI activation. In contrast, when the healthy forepaw was stimulated in denervated rats (n=7), both contralateral and ipsilateral SI exhibited a significantly larger fMRI activation area. The number of pixels above the cross-correlation threshold was calculated for each individual rat and averaged across the group. Since the forepaw representation is slightly different for each individual rat, these averaged t-test maps may not precisely define SI representation. b Stimulus-induced local field potential (LFP) responses in individual rats, beginning 50 μm below the SI cortical surface with LFP recordings performed at 150 μm increments. Marked changes in LFP activity were observed when the forepaw (FP) contralateral to the recorded SI was stimulated in sham-operated and denervated rats. However, stimulation of the forepaw ipsilateral to the recorded SI resulted in a minimal change in LFP activity in either group. Green arrows indicate stimulus onset. c Group averages of stimulus-induced LFP-negative deflection amplitude. The amplitude of the initial negative defection of the LFP response in the denervated rats was greater in the healthy cortex, in particular in lamina IV, when the intact forepaw was stimulated. Significant differences were also observed in lamina II + III. However, the amplitudes of the LFP deflection in both control and denervated rats were similar in the SI cortex ipsilateral to the forepaw stimulation. Note the tenfold scale difference between responses to contralateral (top) and ipsilateral (bottom) stimulation (∗ , paired t-test, p<0.05). From (32) copyright (2009) National Academy of Sciences, USA.

Figure 29.2 demonstrates that increases in the magnitude of fMRI response in the contralateral somatosensory cortex following peripheral nerve injury concurred with increases in neuronal activity as was measured with electrophysiology (32). Thus, differences in the magnitude of fMRI responses following injury may indicate the level of neuronal engagement in processing a particular stimulus. Following stroke in rodent models, stimulation of the affected limb resulted in fMRI responses in both affected and non-affected cortices (4, 5). fMRI responses in brain regions not normally activated by the specific stimuli applied may indicate activation and involvement of additional neuronal pathways to compensate for loss of function. 3.4.2. Differences in Functional Connectivity Assessed with Resting-State fMRI

Generally for fMRI studies, only task-activated networks are visualized. On the other hand, modification in the behavior of large neuronal networks due to plasticity can be evaluated in cortical, as well as subcortical, brain regions. Figure 29.3 demonstrates

MRI of Neuronal Plasticity in Rodent Models

575

Fig. 29.3. Color-coded maps of mean functional brain connectivity of right primary motor cortex (M1) (white arrows) in rats before and at 3, 7, and 70 days after 90-min right middle cerebral artery occlusion. Data are derived from serial resting-state blood oxygenation level-dependent (BOLD) functional MRI experiments in 14 animals. Functional connectivity was calculated from the Fisher-transformed correlation coefficient (z ) between low-frequency BOLD signals in a seed region (right, i.e., ipsilesional, M1) and all other brain voxels. Maps are overlaid on a T2 -weighted multislice anatomical rat brain template. Functional connectivity with the left, contralesional sensorimotor cortex was reduced at subacute stages after right-sided stroke, but partly recovered thereafter. Image courtesy of Maurits P.A. van Meer, Kajo van der Marel, and Rick M. Dijkhuizen. Image Sciences Institute, University Medical Center Utrecht, The Netherlands.

connectivity maps calculated from seed voxels located in the rats’ primary motor cortex. Functional recovery of transcallosal networks can be observed in the days following stroke (van Meer, M et al., unpublished data). 3.4.3. Differences in T1-Enhanced Regions

Increases or decreases in T1-weighted signals within specific regions of interests (ROI) were demonstrated in a variety of rodent models for plasticity. The percentile or absolute increases in T1-weighted signal can be derived from serial T1-weighted images. Increases in T1 values in the ROI following systemic MnCl2 administration may indicate increases in neuronal processing or neuronal efficacy related to the task or stimulation paradigm. Direct stereotactic MnCl2 administration to track trace specific neuronal pathways has been demonstrated to be valuable to detect new and/or a modification of neuronal pathways involved with plasticity.

4. Notes 1. Changes in the function of neuronal networks may be accompanied with minimal morphological modifications, especially in cases of short-term plasticity associated with

576

Pelled

learning and memory. In these cases, MEMRI may not be the ideal approach to image plasticity. 2. A few reports demonstrated that occlusion of the cerebral artery do not lead to a massive cortical reorganization in rats, for example, Weber et al. (33). 3. Rodents cannot recover from this anesthesia; hence these are non-survival procedures. Recently, several studies have demonstrated the use of a new anesthetic that permits recovery following fMRI measurements. This now offers the ability to pursue longitudinal fMRI studies of rodents (34, 35). 4. The concentration and the amount of MnCl2 depend on the administration route (24). 5. Functional MRI responses are reproducible and robust in cortical regions. Thus, neuronal plasticity that involves mostly subcortical structures will be challenging to assess with fMRI techniques.

Acknowledgments The author would like to thank Dr. Alan Koretsky for many years of insightful discussions and critical reading of this chapter and Dr. Rick Dijkhuizen and his colleagues for their contribution to this chapter. References 1. Sadato, N., Pascual-Leone, A., Grafman, J. et al. Activation of the primary visual cortex by Braille reading in blind subjects. Nature 1996;380:526–528. 2. Wittenberg, G. F., Werhahn, K. J., Wassermann, E. M., Herscovitch, P., Cohen, L. G. Functional connectivity between somatosensory and visual cortex in early blind humans. Eur J Neurosci 2004;20:1923–1927. 3. Feydy, A., Carlier, R., Roby-Brami, A. et al. Longitudinal study of motor recovery after stroke: Recruitment and focusing of brain activation. Stroke 2002;33: 1610–1617. 4. Dijkhuizen, R. M., Ren, J., Mandeville, J. B. et al. Functional magnetic resonance imaging of reorganization in rat brain after stroke. Proc Natl Acad Sci USA 2001;98: 12766–12771.

5. Dijkhuizen, R. M., Singhal, A. B., Mandeville, J. B. et al. Correlation between brain reorganization, ischemic damage, and neurologic status after transient focal cerebral ischemia in rats: A functional magnetic resonance imaging study. J Neurosci 2003;23:510–517. 6. Pelled, G., Bergman, H., Goelman, G. Bilateral overactivation of the sensorimotor cortex in the unilateral rodent model of Parkinson’s disease – A functional magnetic resonance imaging study. Eur J Neurosci 2002;15: 389–394. 7. Pelled, G., Chuang, K. H., Dodd, S. J., Koretsky, A. P. Functional MRI detection of bilateral cortical reorganization in the rodent brain following peripheral nerve deafferentation. Neuroimage 2007;37: 262–273.

MRI of Neuronal Plasticity in Rodent Models 8. Pelled, G., Dodd, S. J., Koretsky, A. P. Catheter confocal fluorescence imaging and functional magnetic resonance imaging of local and systems level recovery in the regenerating rodent sciatic nerve. Neuroimage 2006;30:847–856. 9. Yu, X., Wang, S., Chen, D. Y., Dodd, S., Goloshevsky, A., Koretsky, A. P. 3D mapping of somatotopic reorganization with small animal functional MRI. Neuroimage 2009;49(2):1667–1676. 10. Pawela, C. P., Biswal, B. B., Hudetz, A. G. et al. Interhemispheric neuroplasticity following limb deafferentation detected by restingstate functional connectivity magnetic resonance imaging (fcMRI) and functional magnetic resonance imaging (fMRI). Neuroimage 2010;49(3):2467–2478. 11. Yu, X., Sanes, D. H., Aristizabal, O., Wadghiri, Y. Z., Turnbull, D. H. Largescale reorganization of the tonotopic map in mouse auditory midbrain revealed by MRI. Proc Natl Acad Sci USA 2007;104: 12193–12198. 12. Bissig, D., Berkowitz, B. A. Manganeseenhanced MRI of layer-specific activity in the visual cortex from awake and free-moving rats. Neuroimage 2009;44:627–635. 13. Yu, X., Wadghiri, Y. Z., Sanes, D. H., Turnbull, D. H. In vivo auditory brain mapping in mice with Mn-enhanced MRI. Nat Neurosci 2005;8:961–968. 14. Bouilleret, V., Cardamone, L., Liu, Y. R., Fang, K., Myers, D. E., O’Brien, T. J. Progressive brain changes on serial manganese-enhanced MRI following traumatic brain injury in the rat. J Neurotrauma 2009;26(11):1999–2013. 15. Nairismagi, J., Pitkanen, A., Narkilahti, S., Huttunen, J., Kauppinen, R. A., Grohn, O. H. Manganese-enhanced magnetic resonance imaging of mossy fiber plasticity in vivo. Neuroimage 2006;30:130–135. 16. Pelled, G., Bergman, H., Ben-Hur, T., Goelman, G. Manganese-enhanced MRI in a rat model of Parkinson’s disease. J Magn Reson Imaging 2007;26(4):863–870. 17. van der Zijden, J. P., Bouts, M. J., Wu, O. et al. Manganese-enhanced MRI of brain plasticity in relation to functional recovery after experimental stroke. J Cereb Blood Flow Metab 2008;28:832–840. 18. Karni, A., Meyer, G., Jezzard, P., Adams, M. M., Turner, R., Ungerleider, L. G. Functional MRI evidence for adult motor cortex plasticity during motor skill learning. Nature 1995;377:155–158. 19. Little, D. M., Klein, R., Shobat, D. M., McClure, E. D., Thulborn, K. R. Changing

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

577

patterns of brain activation during category learning revealed by functional MRI. Brain Res Cogn Brain Res 2004;22:84–93. Olesen, P. J., Westerberg, H., Klingberg, T. Increased prefrontal and parietal activity after training of working memory. Nat Neurosci 2004;7:75–79. Soria, G., Wiedermann, D., Justicia, C., Ramos-Cabrer, P., Hoehn, M. Reproducible imaging of rat corticothalamic pathway by longitudinal manganese-enhanced MRI (LMEMRI). Neuroimage 2008;41:668–674. Tucciarone, J., Pelled, G., Koretsky, A. P. Changes in callosal and thalamic connectivity following peripheral nerve damage to the rodent forepaw detected with Manganese Enhanced MRI. Proc Intl Soc Mag Reson Med 2009;17. Cross, D. J., Flexman, J. A., Anzai, Y., Morrow, T. J., Maravilla, K. R., Minoshima, S. In vivo imaging of functional disruption, recovery and alteration in rat olfactory circuitry after lesion. Neuroimage 2006;32: 1265–1272. Silva, A. C., Lee, J. H., Wu, C. W. et al. Detection of cortical laminar architecture using manganese-enhanced MRI. J Neurosci Methods 2008;167:246–257. Goloshevsky, A. G., Silva, A. C., Dodd, S. J., Koretsky, A. P. BOLD fMRI and somatosensory evoked potentials are well correlated over a broad range of frequency content of somatosensory stimulation of the rat forepaw. Brain Res 2008;1195:67–76. Lu, H., Patel, S., Luo, F. et al. Spatial correlations of laminar BOLD and CBV responses to rat whisker stimulation with neuronal activity localized by Fos expression. Magn Reson Med 2004;52:1060–1068. Pawela, C. P., Hudetz, A. G., Ward, B. D. et al. Modeling of region-specific fMRI BOLD neurovascular response functions in rat brain reveals residual differences that correlate with the differences in regional evoked potentials. Neuroimage 2008;41: 525–534. Majeed, W., Magnuson, M., Keilholz, S. D. Spatiotemporal dynamics of low frequency fluctuations in BOLD fMRI of the rat. J Magn Reson Imaging 2009;30: 384–393. Tucciarone, J., Chuang, K. H., Dodd, S. J., Silva, A., Pelled, G., Koretsky, A. P. Layer specific tracing of corticocortical and thalamocortical connectivity in the rodent using manganese enhanced MRI. Neuroimage 2009;44:923–931. Chuang, K. H., Koretsky, A. P. Accounting for nonspecific enhancement in neuronal

578

Pelled

tract tracing using manganese enhanced magnetic resonance imaging. Magn Reson Imaging 2009;27:594–600. 31. Chuang, K. H., Koretsky, A. Improved neuronal tract tracing using manganese enhanced magnetic resonance imaging with fast T(1) mapping. Magn Reson Med 2006;55:604–611. 32. Pelled, G., Bergstrom, D. A., Tierney, P. L. et al. Ipsilateral cortical fMRI responses after peripheral nerve damage in rats reflect increased interneuron activity. Proc Natl Acad Sci USA 2009;106:14114–14119. 33. Weber, R., Ramos-Cabrer, P., Justicia, C. et al. Early prediction of functional recovery after experimental stroke: Functional mag-

netic resonance imaging, electrophysiology, and behavioral testing in rats. J Neurosci 2008;28:1022–1029. 34. Pawela, C. P., Biswal, B. B., Hudetz, A. G. et al. A protocol for use of medetomidine anesthesia in rats for extended studies using task-induced BOLD contrast and resting-state functional connectivity. Neuroimage 2009;46: 1137–1147. 35. Weber, R., Ramos-Cabrer, P., Wiedermann, D., van Camp, N., Hoehn, M. A fully noninvasive and robust experimental protocol for longitudinal fMRI studies in the rat. Neuroimage 2006;29: 1303–1310.

Chapter 30 MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption Yuexi Huang and Kullervo Hynynen Abstract MR-guided transcranial focused ultrasound (FUS) has been demonstrated as a non-invasive tool for treating various brain diseases. First, FUS can thermally ablate brain tissues under real-time MR thermometry monitoring. The MRI guidance significantly improves the precision of the thermal dose deposition. Second, in conjunction with microbubble contrast agents, FUS can reversibly disrupt the blood–brain barrier for delivery of macromolecular drugs to the brain parenchyma. This offers huge potential for treating brain diseases with a much higher local drug concentration than other drug delivery methods. In this chapter, a detailed protocol of MR-guided focused ultrasound for brain thermal ablation and BBB disruption in an animal research setting is presented. Key words: Focused ultrasound, magnetic resonance imaging, thermal ablation, blood–brain barrier, drug delivery systems, ultrasound contrast agents, MRgFUS.

1. Introduction Treatment of brain diseases using focused ultrasound (FUS) was first reported in the 1940s (1, 2) and was investigated by many groups over the next few decades (3–6). Focusing devices were developed, acoustic parameters for various applications were established, and clinical trials were performed. However, the need for a craniotomy to provide an acoustic window for the ultrasound beam significantly limited the attractiveness of FUS as a minimally invasive technique. Although transcranial lesion production was attempted (7) the distortion of the foci and the accompanied skull heating prevented it from being adopted in M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, DOI 10.1007/978-1-61737-992-5_30, © Springer Science+Business Media, LLC 2011

579

580

Huang and Hynynen

clinical use. In addition, the lack of a reliable imaging tool for real-time guidance and follow-up evaluation further impeded the broad acceptance of FUS as a neurosurgical tool in clinical practice. The revolution started in the 1980s, as MRI emerged as an excellent tool for visualizing the brain anatomy. More importantly, MRI is capable of measuring temperature in vivo with accuracy sufficient for monitoring FUS ablation (8–10). On the ultrasound side, a hemispherical phased-array design minimized skull heating, and the foci distortion was corrected based on CT guidance (11). A prototype brain system has been developed by Insightec (Haifa, Israel), which is under clinical evaluation at multiple hospitals. Similarly, Supersonic Imagine (France) is developing a clinical prototype device. In addition to thermal ablation, FUS may also be applied for the targeted opening of the blood–brain barrier (BBB). The BBB prevents the passage of most macromolecules to the brain parenchyma and thus seriously limits the usefulness of macromolecular drugs in the treatment of brain disease (12–14). A reversible disruption of the BBB at targeted locations has been achieved by using focused ultrasound bursts in conjunction with microbubble contrast agent (15). Since ultrasound can be focused noninvasively through the human skull under MRI guidance (11) this method offers huge potential for non-invasive drug delivery to the brain. In this chapter, a detailed description of MR-guided focused ultrasound for brain thermal ablation and BBB disruption in an animal research setting is presented. Similar methods can be extended for human subjects once human brain FUS systems become more established and widely available.

2. Materials 2.1. MR-Compatible Focused Ultrasound System

1. The ultrasound is generated by a focused single-element, MR-compatible piezoelectric transducer. The size of the focal spot is determined by the physical characteristics of the transducer. For example, a transducer with a 100 mm diameter and 80 mm radius of curvature at a center frequency of 1.0 MHz produces a focus approximately 1.7 mm in diameter in the lateral dimension and 7.5 mm in the axial dimension (full width half maximum of pressure) (see Note 1). 2. A frequency generator (Tektronix Model AFG 3102) is used to generate the RF signal for the transducer.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

581

3. The RF signal is amplified by an RF amplifier (E & I Model A150) for driving the transducer. 4. The forward and reflected electric power is measured in continuous wave (CW) mode using a digital power meter (HP Model 438A) and a dual directional coupler (Werlatone Model C1373) (see Note 2). 5. An MR-compatible positioning system is used for positioning the transducer relative to the animal brain (16). The system includes a water tank and a three-axis transducer positioning system (discussed in more detail in Section 3.2). 2.2. Degassed Water

1. A coupling medium is required between the transducer and the brain for the propagation of the ultrasound beam. Water is the most convenient medium since its acoustic attenuation is small and the speed of sound and density is close to those in soft tissues. Thus, the ultrasound beam propagates almost completely undisturbed through water and couples well to the skin with minimal reflections. Since gas bubbles significantly disturb ultrasound propagation, it is important to minimize the chance of bubble formation in the water. This can be done by degassing the water prior to use. 2. Degassing can be done by boiling the water for 20–30 min and then cooling down to room temperature. An alternative method is to use a degassing system that removes the gas with a vacuum pump.

2.3. Ultrasound and MRI Contrast Agents

1. For BBB disruption, both ultrasound and MRI contrast agents are used. Two commercial ultrasound contrast R (GE Healthcare, Milwaukee, WI, USA) agents, Optison R  and Definity (Lantheus Medical Imaging, N. Billerica, MA, USA), were widely used and produced almost identical results (15, 17–27). Both are stored at +4◦ C. Similar results with other microbubble contrast agents were also published (28–30) (see Note 3). 2. The disruption of the BBB via focused ultrasound can be visualized using standard MRI contrast agents, such as gadopentetate dimeglumine (MAGNEVIST, Berlex Laboratories, Inc., Wayne, NJ) (molecular weight of 928), that do not penetrate through an intact BBB. 3. Thermally coagulated tissue can be visualized using contrastenhanced MRI as a non-perfused (dark) core area surrounded by a contrast-enhanced ring (11).

2.4. MRI Scanner

1. Standard clinical 1.5 and 3 T scanners or standard animal scanners at high field strength can be used for these studies.

582

Huang and Hynynen

Small surface coils (receive only or transmit-receive) are preferred for high-resolution images of the animal brains.

3. Methods 3.1. Ultrasound Transducer Calibration

1. Prior to the experiments, focal acoustic intensity and pressure amplitude as a function of applied driving voltage and RF power need to be calibrated. 2. The transducer should be placed in degassed water in a tank with walls lined with sound absorbing rubber mats. 3. Position a calibrated hydrophone (spot diameter 0.5 mm, GEC-Marconi Research Center, Chelmsford, England) at the focal distance from the transducer. The hydrophone should be connected to an oscilloscope to measure its voltage (see Note 4). 4. Move the hydrophone across the focal plane until the maximum voltage generated by the hydrophone is found. 5. Record the amplitude of the voltage as a function of the driving RF voltage from the frequency generator. 6. Using the calibration coefficient of the hydrophone, calculate the acoustic intensity or pressure amplitude values. 7. High-power values which are beyond the range of the hydrophone measurements can be linearly extrapolated based on measured RF power or can be directly measured using a fiber optic hydrophone or using a laser vibrometer and a reflecting membrane (31).

3.2. Transducer Positioning and Focus Registration

1. Figure 30.1 shows a diagram of the transducer positioning system. 2. The ultrasound transducer should be mounted on a mechanical arm holder within a plastic tank filled with degassed water. 3. Place the MRI coil around the opening of the plate covering the tank. The head of the animal will be positioned at the opening when it is lying supine on the plate. 4. A plastic membrane must be placed through the hole of the MRI coil and filled with degassed water. This will allow the water level to be raised to provide water coupling for the ultrasound beam to the animal brain without risking overflow of the tank.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

583

Fig. 30.1. A diagram of the sonication system and the equipment.

5. Move the positioning arm and transducer, aiming the focal spot through the opening. The movement of the transducer arm can be accomplished with a manual lead-screw-based positioning system or a computer-controlled motor system (16) (see Note 5). 6. Before starting the experiments, the position of the ultrasound focal spot needs to be registered to the MRI coordinate system. Aim the ultrasound beam on the water surface at a low-power level (~1 W), i.e., the distance between the transducer and the water surface would be the focal length. The focal spot appears as a fountain in the water. 7. Place a marker at the focal spot (see Note 6). 8. Perform a three-plane localizer MRI sequence to image the marker. 9. Measure the location of the marker in the MRI coordinate system. 10. Use this location as the reference location when targeting structures in the brain. 3.3. Animal Preparation

1. BBB disruption in rats and mice can be performed transcranially without creating a bone window. When using a lowfrequency transducer (<1 MHz), rabbits can also be sonicated transcranially for BBB disruption (see Note 7). 2. The animal should be anesthetized using a mix of 40 mg/kg ketamine (Aveco Co, Inc., Fort Dodge, IA) and 10 mg/kg of xylazine (Lloyd Laboratories, Shenandoah, IA) (see Note 8). 3. Remove the hair over the top of the head by shaving and applying hair removal lotion.

584

Huang and Hynynen

4. Insert an intravenous catheter in the ear (rabbit) or tail vein (mice, rats). 5. Place the animal on a water blanket through which temperature-controlled water is circulated to maintain the body temperature of the animal. 6. Place a rectal thermometer to monitor the temperature of the animal. 7. In cases of thermal ablation, a skull window for the ultrasound beam needs to be opened to avoid excessive skull heating. It may be possible to cool the surface of the skull in rats and mice and to be able to ablate the brain through the skull. 8. After removing the hair over the head, apply antiseptic lotion to sterilize the skin. 9. Cut the skin over the central line of the top of the skull. 10. Push the skin sideways to expose the skull surface. 11. Remove a piece of skull (approximately 20×20 mm for rabbits and 10×10 mm for rats). 12. Replace the skin over the bone window. 13. Suture the skin back together. 14. Allow the animal to recover from the anesthesia. 15. Provide pain medication for 2 days post-surgery. 16. Allow the wound to heal and then remove the sutures. 17. Allow the suture holes to heal. 18. The experiment must be performed a minimum of 10 days post-surgery. 3.4. Animal Positioning

1. Place the animal supine on the plate such that the top of the head is in the middle of the imaging coil over the opening for the ultrasound beam. 2. Inject degassed water into the plastic membrane under the head. 3. Secure the body and the head with tape and straps or a stereotactic fixture. 4. Perform a three-plane localizer to assure appropriate head position and good water coupling. 5. Re-position the head if necessary and repeat imaging. 6. Perform high-resolution T2-weighted FSE sequence for detailed anatomical mapping. A typical set of MR parameters is listed in Table 30.1. 7. Select the target locations from the images. 8. Aim the ultrasound beam focus to the first target location.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

585

Table 30.1 Parameters used in the MR imaging Frequency (MHz)

50% focus width (mm)

50% focus length (mm)

0.26

8

40

0.69

2.3

14

1.1

1.7

7.5

1.5

1

4.8

3.3

0.6

3.2

3.5. Thermal Ablation 3.5.1. MR Thermometry Sequence

1. A water proton resonance frequency (PRF)-based MR thermometry sequence is generally used to monitor the temperature during thermal ablation, because it is largely tissue independent and has a linear performance over a wide temperature range (32) (see Note 9). 2. Prescribe a multi-phase 2D gradient-echo sequence across the targeted location. It is preferable that the slice be perpendicular to the ultrasound beam to minimize averaging errors on the slice-selection dimension. Slice thickness is generally 1.5–3 mm. If a slice parallel to the ultrasound beam is prescribed, slice thickness should be reduced to 1–2 mm to minimize partial volume errors (33). 3. In-plane resolution is generally sufficient to avoid significant partial volume errors if a 128 by 128 matrix is chosen for a field-of-view (FOV) of 4–0 cm, depending on the size of the animal brain. 4. Repetition time (TR) should be prescribed according to the temporal resolution, and the temporal resolution should be chosen based on the rate of heating. For focused thermal ablation, a temporal resolution of 3–5 s should be sufficient to monitor the course of the temperature change for exposures 20 s and longer. If a 5 s temporal resolution is used, TR should be set as 39 ms (5 s – 128 phase-encoding steps). For shorter sonications, higher temporal resolution should be selected. 5. Echo time (TE) should be chosen based on the tissue T2, T2∗ , temperature sensitivity, and dynamic range. Typically a TE of 10–20 ms is a good trade-off. 6. The real and imaginary part of the MR data should be saved for processing the temperature map. The MR data can be pulled from the MR scanner in real time to a computer for real-time processing and displaying of the temperature maps.

586

Huang and Hynynen

7. The MR thermometry sequence should be started before the ultrasound exposure. Two thermometry images are acquired as the baselines and the ultrasound exposure starts simultaneously with the acquisition of the third thermometry image. 8. Phase subtraction of the baseline image from the subsequent images is performed to generate the temperature map. A four-quadrant arctangent operation is used for the phase φ subtraction (34). The temperature change T = TE·B·γ ·α , where φ is the phase difference in radiant, TE is the echo time, B is the strength of the magnetic field, γ is the gyromagnetic ratio, and α is the temperature sensitivity coefficient at –0.01 ppm/◦ C (see Note 10). 3.5.2. Ultrasound Exposure

1. Ultrasound exposure should begin simultaneously with the third phase of the MR thermometry sequence. For thermal ablation, continuous wave exposures or high duty cycle bursts (e.g., 50% duty cycle) are generally applied to raise the tissue temperature to around 60◦ C for almost instant tissue necrosis (<1 s). Typically an acoustic intensity greater than 1,000 W/cm2 is needed. Figure 30.2 shows an example of MR thermometry measurements of FUS ablation with and without microbubbles (35). 2. The acoustic power and the sonication duration should be adjusted based on the measured temperature change.

3.5.3. Verifying the Thermal Lesion

1. T2w FSE sequence should be used for visualizing the thermal lesion. Typically the lesion appears as a hypointense center zone surrounded by a hyperintense outer zone. The hyperintense zone represents the perifocal edema. The lesion grows over time, therefore the size of the edema area in T2w images will be larger, e.g., at 48 h, than the images immediately acquired after the ablation. 2. Gd-enhanced T1w images can also be used for verifying lesions by the loss of enhancement at the treated site. The lack of contrast enhancement indicates lack of perfusion due to the blockage of the capillary vessels by the temperature elevation. For this purpose, inject Gd contrast agent intravenously at 0.125 mmol/kg following the ablation and scan with a T1w FSE sequence.

3.6. BBB Disruption 3.6.1. Microbubble Injection

1. Inject the ultrasound contrast agent containing microbubbles intravenously through the ear vein (rabbits) or tail vein (rats, mice) simultaneously with the start of the sonication. R dosage is 10 μl/kg of body weight, which The Definity is that recommended by the manufacturer for clinical use.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

587

Fig. 30.2. Sagittal MR thermometry images without (a) and with (b) microbubbles. 10-s continuous wave sonications at 3 W (4.4 MPa) were applied without microbubbles and 10-s pulsed sonication (50% duty cycle) at 1.2 W (2.8 MPa) was applied with microbubbles. The temperature elevation at the focal spot was higher with microbubbles despite the lower power (c) (Data are from Ref. (36). Reprinted with permission of RSNA).

R The Optison dosage is 50 μl/kg of body weight, which is within the range of 0.5–5.0 ml recommended for human use (see Note 11). For rats and mice, the microbubbles may need to be diluted with saline before injection so that a reasonable volume can be injected accurately.

2. When working with the ultrasound contrast agent, use an 18-gauge needle to avoid breaking the microbubbles. Inject slowly. 3. Following the microbubble injection, approximately 0.5 ml/kg of saline is injected to flush the agent from the IV line. 3.6.2. Ultrasound Exposure

1. Start the sonication simultaneously with the microbubble injection. The sonication has a burst length of 1–100 ms and a repetition frequency of 0.5–2 Hz (see Note 12). Typically, the sonication duration will be between 30 s and 5 min. The

588

Huang and Hynynen

driving voltage and acoustic power should be adjusted based on the estimated in situ acoustic pressure at the particular ultrasound frequency (e.g., ~200 kPa at 500 kHz). 2. The sonication can be repeated in another location after the previously injected microbubbles cleared from the circulation system. This requires a minimum 5 min interval between two microbubble injections (see Note 13). 3.6.3. Delivery of Macromolecular Agent

3.6.4. Verifying the BBB Disruption

The most effective delivery of the macromolecular agent through the BBB is achieved when the agent is injected simultaneously with the microbubbles, at the beginning of the sonication. The macromolecular agent can also be injected after the sonications, as a certain level of BBB penetration will remain up to approximately 6 h after the exposure and may last longer if higher power is used (15). 1. A T1-weighted FSE image across the brain at the focal depth should be acquired before the sonication as a baseline image. 2. Inject Gd-based MRI contrast agent intravenously at 0.125 mmol/kg. 3. Repeat the T1-weighted image multiple times. 4. Subtract the signal intensity of the baseline image from the subsequent images. 5. The increase in signal intensity at the sonicated locations, relative to that in the same anatomical structure on the contralateral side, is proportional to the size of the BBB

Fig. 30.3. T1-weighted contrast-enhanced MR images (a coronal, b sagittal) of a rat brain demonstrating four spots of BBB opening at a spacing of 2 mm in the right hemisphere. The opening of the four spots was achieved after a single microbubble injection by moving the transducer along the line with an automated positioning system during the sonication.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

589

opening and the amount of macromolecular agent delivered into the brain (Fig. 30.3).

4. Notes 1. Several different frequencies between 0.25 and 5 MHz have been reported (17–28). The use of lower frequencies not only increases the focal spot size but also reduces the number of erythrocyte extravasations per unit area. The reduction in the diameter of the transducer, or increase in the radius of curvature, also increases the focal spot size. Examples of the 50% focal spot dimensions for different transducers are given in Table 30.2. The electrical impedance of the transducer is typically matched to the output impedance of the amplifier by an external LC-matching network to allow optimum power transfer. Although we manufacture our own transducers in-house, custom MRI compatible transducers can be purchased from several manufacturers, such as Imasonic Inc. (Besancon, France).

Table 30.2 Examples of measured focal spot dimensions (50% of the peak pressure amplitude) for f-number 0.8 focused transducers Sequence

TR/TE

Echo train length

Field of view

Matrix size

Slice thickness

bandwidth

T2-weighted FSE

2,000/75 ms

8

4–10 cm

128×128

1 mm

7 kHz

T1-weighted FSE (Gd-enhanced)

500/17 ms

4

4–10 cm

128×128

1 mm

16 kHz

2. Other frequency generators, RF-amplifiers, and power meters can be used. Equivalent devices having all of the components integrated (the frequency generation, amplification and power measurement) are manufactured by Advanced Surgical Systems Inc. (Tucson, AZ, USA). 3. We also performed experiments with lipid/ perfluorocarbon-based microbubbles made in house with an Artenga microbubble generator (model MGD4, Artenga Inc, Ottawa, Canada) and observed a similar level of BBB disruption. 4. Hydrophones can also be purchased from a commercial vendor, such as Precision Acoustics (Dorchester, UK). The

590

Huang and Hynynen

reported values are often estimates for the pressure amplitude in the brain obtained by decreasing the measured water values based on ultrasound attenuation through the brain with an average amplitude attenuation coefficient of 5 Np/m/MHz. If through-skull exposures are performed then the insertion loss induced by the skull bone needs to be taken into account. There is significant variation in the through-skull propagation from location to location and from animal to animal, and this may increase as the animal gets older. Typically, the lower the frequency, the lower the losses in the skull. Therefore it is recommenced that the loss measurements be performed with skull samples from the animals used in the experiments. Examples of measured losses are given in Table 30.3.

Table 30.3 Measured pressure amplitude transmission through animal skulls. The values are average values for limited measurements and indicate the pressure amplitude value that is measured when the beam propagates through the skull in water (water measurement = 100%). There are large variations from location to location and skull to skull and these are only to be used as guidelines Species

0.25 MHz (%)

0.69 MHz (%)

1.1 MHz (%)

Rabbit

72

34

8.6

Rat

85

53

Mouse

87

5. The transducer is mounted on a mechanical positioning system that can be moved to aim the focus at the desired location in the brain based on the MR images. The positioner needs to be MRI compatible. This means that all parts should be non-magnetic. In general, plastic and other non-metallic materials are fine. Non-magnetic metals (such as brass, copper, aluminum) can also be used, but their amount should be kept small and removed from the imaging space. A computer-controlled experimental animal positioning system developed for our experiments has been described in Ref. (31). 6. The marker can be a plastic plate with a hole about the diameter of the focus. The position of the hole can be visualized in MR images if filled with water.

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

591

7. When a frequency higher than 1 MHz is used for the rabbit study, a craniotomy needs to be performed to provide an acoustic window. 8. Isoflurane gas can be used for animals. However, our experience found that isoflurane reduced the amount of BBB disruption when compared with ketamine and xylazine. 9. Other than PRF-based method, T1 relaxation time, proton density, and diffusion-weighted methods were also investigated for MR temperature measurement. However, they are tissue dependent and were only used in cases where the PRF effect is not significant. For example, T1w measurement was used in fatty tissues due to the lack of PRF change in fat (36). 10. If phase-wrap occurs, i.e., the temperature change is over the dynamic range of the phase difference from the baseline image, subtraction of phase images of consecutive time points then summation of the stepwise temperature changes together is preferred. 11. These agents consist of pre-formed bubbles that are either R R ) or human serum albumin (Optison ) lipid (Definity shells filled with the perfluorocarbon gas Perflutren. 12. The sonication parameters were adjusted empirically. Most of our early experiences were with a 20s sonication. With this short sonication, a 10 s delay between the bubble injection and the start of the sonications was used on rabbits to allow the microbubbles to enter into the brain. On the other hand, a sonication as long as 300 s allowed simultaneous injection of the microbubble and provided the maximum BBB disruption, but came with a higher chance of hemorrhage and/or tissue damage. Two minute sonications have been used lately as a compromise between the simplicity of the injection and a good BBB disruption. The acoustic power and peak pressure amplitude for the threshold of the BBB disruption arefrequency dependent and are roughly proportional to 1/ frequency. The BBB disruption is not enhanced by increasing the burst length to 100 ms, but the threshold is increased when 1 ms or shorter exposures are used (15). Burst repetition frequencies between 0.5 and 2 Hz provided equal results. The BBB disruption increases with increasing concentration of microbubbles. 13. There are two options when multiple locations are targeted. When a manual or slow positioning device is used, each of the locations is targeted serially with a minimum time delay of 5 min between two injections of microbubbles. When fast positioning is possible, either with an

592

Huang and Hynynen

automatic motor system or with a phased-array transducer, all of the locations can be sonicated during a single injection of microbubble by interleaving sonications at the targeted locations at an overall repetition frequency of 1 Hz.

Acknowledgments Sources of Support: NIH (grant numbers EB00705 and EB003268) and the Terry Fox Foundation. References 1. Lynn, J. G., Zwemer, R. L., Chick, A. J., Miller, A. G. A new method for the generation and use of focused ultrasound in experimental biology. J Gen Physiol 1942;26: 179–193. 2. Lynn, J. G., Putnam, T. J. Histological and cerebral lesions produced by focused ultrasound. Am J Pathol 1944;20:637–649. 3. Fry, W. J., Mosberg, W. H., Jr, Barnard, J. W., Fry, F. J. Production of focal destructive lesions in the central nervous system with ultrasound. J Neurosurg 1954;11:471–478. 4. Fry, F. J. Precision high intensity focusing ultrasonic machines for surgery. Am J Phys Med 1958;37:152–156. 5. Cosman, B. J., Hueter, T. F. Instrumentation for ultrasonic neurosurgery. Electronics 1959;5:53–57. 6. Fry, W. J., Meyers, R. Ultrasonic method of modifying brain structures. Confin Neurol 1962;22:315–327. 7. Fry, F. J., Goss, S. A. Further studies of the transkull transmission of an intense focused ultrasonic beam: Lesion production at 500 kHz. Ultrasound Med Biol 1980;6:33–38. 8. Cline, H. E., Schenck, J. F., Watkins, R. D., Hynynen, K., Jolesz, F. A. Magnetic resonance-guided thermal surgery. Magn Reson Med 1993;31:628–636. 9. De Poorter, J., De Wagter, C., De Deene, Y., Thomsen, C., Stahlberg, F., Achten, E. Noninvasive MRI thermometry with the proton resonance frequency (PRF) method: In vivo results in human muscle. Magn Reson Med 1995;33:74–81. 10. Ishihara, Y., Calderon, A., Watanabe, H., Okamoto, K., Suzuki, Y., Kuroda, K., Suzuki,

11.

12. 13.

14. 15.

16.

17.

18.

Y. A precise and fast temperature mapping using water proton chemical shift. Magn Reson Med 1995;34:814–823. Hynynen, K., Clement, G. T., McDannold, N., Vykhodtseva, N., King, R., White, P. J., Vitek, S., Jolesz, F. A. 500-element ultrasound phased array system for noninvasive focal surgery of the brain: A preliminary rabbit study with ex vivo human skulls. Magn Reson Med 2004;52:100–107. Abbott, N. J., Romero, I. A. Transporting therapeutics across the blood-brain barrier. Mol Med Today 1996;2:106–113. Kroll, R. A., Neuwelt, E. A. Outwitting the blood-brain barrier for therapeutic purposes: Osmotic opening and other means. Neurosurgery 1998;42:1083–1099. Pardridge, W. M. Drug and gene delivery to the brain: The vascular route. Neuron 2002;36:555–558. Hynynen, K., McDannold, N., Vykhodtseva, N., Jolesz, F. A. Noninvasive MR imagingguided focal opening of the blood-brain barrier in rabbits. Radiology 2001;220:640–646. Chopra, R., Curiel, L., Staruch, R., Morrison, L., Hynynen, K. An MRI-compatible system for focused ultrasound experiments in small animal models. Med Phys 2009;36:1867–1874. Sheikov, N., McDannold, N., Vykhodtseva, N., Jolesz, F., Hynynen, K. Cellular mechanisms of the blood-brain barrier opening induced by ultrasound in presence of microbubbles. Ultrasound Med Biol 2004;30:979–989. Hynynen, K., McDannold, N., Sheikov, N. A., Jolesz, F. A., Vykhodtseva, N. Local and reversible blood-brain barrier disruption by

MR-Guided Focused Ultrasound for Brain Ablation and Blood–Brain Barrier Disruption

19.

20.

21.

22.

23.

24.

25.

26.

noninvasive focused ultrasound at frequencies suitable for trans-skull sonications. Neuroimage 2005;24:12–20. Hynynen, K., McDannold, N., Vykhodtseva, N., Raymond, S., Weissleder, R., Jolesz, F. A., Sheikov, N. Focal disruption of the blood-brain barrier due to 260-kHz ultrasound bursts: A method for molecular imaging and targeted drug delivery. J Neurosurg 2006;105:445–454. Sheikov, N., McDannold, N., Jolesz, F., Zhang, Y. Z., Tam, K., Hynynen, K. Brain arterioles show more active vesicular transport of blood-borne tracer molecules than capillaries and venules after focused ultrasound-evoked opening of the blood-brain barrier. Ultrasound Med Biol 2006;32:1399–1409. Kinoshita, M., McDannold, N., Jolesz, F. A., Hynynen, K. Noninvasive localized delivery of Herceptin to the mouse brain by MRIguided focused ultrasound-induced bloodbrain barrier disruption. Proc Natl Acad Sci USA 2006;103:11719–11723. Treat, L. H., McDannold, N., Vykhodtseva, N., Zhang, Y., Tam, K., Hynynen, K. Targeted delivery of doxorubicin to the rat brain at therapeutic levels using MRIguided focused ultrasound. Int J Cancer 2007;121:901–907. Choi, J. J., Pernot, M., Small, S. A., Konofagou, E. E. Noninvasive, transcranial and localized opening of the blood-brain barrier using focused ultrasound in mice. Ultrasound Med Biol 2007;33:95–104. McDannold, N., Vykhodtseva, N., Hynynen, K. Use of ultrasound pulses combined with definity for targeted blood-brain barrier disruption: A feasibility study. Ultrasound Med Biol 2007;33:584–590. McDannold, N., Vykhodtseva, N., Hynynen, K. Blood-brain barrier disruption induced by focused ultrasound and circulating preformed microbubbles appears to be characterized by the mechanical index. Ultrasound Med Biol 2008;34:834–840. McDannold, N., Vykhodtseva, N., Hynynen, K. Effects of acoustic parameters and ultrasound contrast agent dose on focused ultrasound induced blood-brain barrier disruption. Ultrasound Med Biol 2008;34: 930–937.

593

27. Sheikov, N., McDannold, N., Sharma, S., Hynynen, K. Effect of focused ultrasound applied with an ultrasound contrast agent on the tight junctional integrity of the brain microvascular endothelium. Ultrasound Med Biol 2008;34:1093–1104. 28. Yang, F. Y., Fu, W. M., Yang, R. S., Liou, H. C., Kang, K. H., Lin, W. L. Quantitative evaluation of the use of microbubbles with transcranial focused ultrasound on blood-brain-barrier disruption. Ultrasound Med Biol 2007;33:1421–1427. 29. Liu, H. L., Wai, Y. Y., Chen, W. S., Chen, J. C., Hsu, P. H., Wu, X. Y., Huang, W. C., Yen, T. C., Wang, J. J. Hemorrhage detection during focused-ultrasound induced blood-brain barrier opening by using susceptibility-weighted magnetic resonance imaging. Ultrasound Med Biol 2008;34: 598–606. 30. Xie, F., Boska, M. D., Lof, J., Uberti, M. G., Tsutsui, J. M., Porter, T. R. Effects of transcranial ultrasound and intravenous microbubbles on blood brain barrier permeability in a large animal model. Ultrasound Med Biol 2008;34:2028–2034. 31. Hynynen, K. The threshold for thermally significant cavitation in dog’s thigh muscle in vivo. Ultrasound Med Biol 1991;17: 157–169. 32. Graham, S. J., Chen, L., Leitch, M., Peters, R. D., Bronskill, M. J., Foster, F. S., Henkelman, R. M., Plewes, D. B. Magn Reson Med 1999;41:321–328. 33. Chung, A. H., Jolesz, F. A., Hynynen, K. Thermal dosimetry of a focused ultrasound beam in vivo by magnetic resonance imaging. Med Phys 1999;26:2017–2026. 34. Bernstein, M. A., King, K. F., Zhou, X. J. Handbook of MRI Pulse Sequence. Burlington, VT: Elsevier Academic Press; 2004, 560–562. 35. McDannold, N., Vykhodtseva, N., Hynynen, K. Microbubble contrast agent with focused ultrasound to create brain lesions at low power levels: MR imaging and histologic study in rabbits. Radiology 2006;241: 95–106. 36. Hynynen, K., McDannold, N., Mulkern, R. V., Jolesz, F. A. Temperature monitoring in fat with MRI. Magn Reson Med 2000;43:901–904.

wwwwwww

SUBJECT INDEX

A

C

Activity induced manganese MRI (AIM-MRI). . . . . . .147, 153–161 Adaptive noise cancellation (ANC) . . . . . . . . . . . . . . 311, 314 Adiabatic half passage (AHP) . . . . 183–184, 193–195, 197, 208–209 Alzheimer’s disease . . . . 9, 34, 101–102, 193, 364, 380, 401, 441, 511–531, 535–549 Amide proton transfer (APT). . . . . . . . . . . . . . . . . . .227–236 Amnestic mild cognitive impairment (aMCI) . . . . . . . . . 536 Amyloid plaque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101, 512, 528–529 Amyloid precursor protein (APP) . . . . . . . . . . 513–516, 519, 521, 523 Anaesthesia . . . . . . . . . . . . . . . . 337, 341, 351–352, 392, 455, 457–462, 467, 494, 497–498, 501 Anaphylactic shock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Angiogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190, 452 Anisotropy . . . 127–128, 131–132, 135, 137, 140–142, 180, 253, 255–256, 258, 264, 266, 351, 357 Apolipoprotein E (ApoE) . . . . . . . . . . . . . . . . . . . . . . . . . . . 514 Apparent diffusion coefficient (ADC) . . . . . . 134, 474, 478, 480–483 Arterial spin labelling (ASL) . . . . . . . . . . 282, 305, 327–344, 357, 474, 480, 560 Artificial ventilation . . . . . . . . . . . . . . . . . . . . . . 155, 170, 290, 521–522 Atherosclerosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380 Average artefact subtraction (AAS) . . . . . . . . . . . . . 310–313, 318–319

Calcium (Ca2+ ) . . . . . . . . . . . . 146–147, 153–154, 156, 169, 176, 187, 277 Calibration scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Carr Purcell Meiboom Gill (CPMG) . . . . . . . . . . 60, 73, 75, 77, 80, 91 Cell labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436, 447 Cellular imaging . . . . . . . . . . . . . . . . . . . . . . . . . 9–11, 443–444 Cerebral blood flow (CB) . . . . . . . . . . 7, 282, 285, 289, 328, 331, 333, 337–344, 474, 477–482, 552 Cerebral blood volume (CBV) . . . . . . . . . . . . . . . . 5, 282, 481 Cerebrospinal fluid (CSF) . . . . . . . . . . . . . . . . . 33, 35, 46, 50, 58–60, 114–115, 120–122, 125, 142, 153, 180–181, 189, 192, 196–197, 217, 232, 236, 247, 259, 261–262, 343, 359, 367, 372, 427, 432, 500, 539, 541–542, 548 Chemical exchange saturation transfer (CEST) . . . . . . . . . 6, 10, 227, 230, 235, 271–279 Chemical shift effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Chemical shift imaging (CSI) . . . . . . . . . . . . . . . . . . . 10, 183, 185–188, 193, 195 Chemical shift selective (CHESS) . . . . . 210, 214, 217, 427 Choline (Cho) . . . . . . . . . . . . . 204–206, 209, 213, 216, 220, 453–454, 465, 468 Computer tomography (CT) . . . . . . . . . . . . . . . 5–8, 11, 103, 253, 264–268, 436, 580 Continuous ASL (CASL) . . . . . . . . 331–332, 337, 340–341, 478–479 Continuous wave (CW) modality . . . . . 229–230, 278, 581, 586–587 Contrast agent . . . . . . . 5, 10, 18, 26, 58, 146, 169, 186, 228, 233–234, 253, 255, 272–273, 305, 327–330, 332, 337, 364–365, 367, 380–381, 397–404, 406–407, 410–411, 413, 436, 453, 455, 459–461, 464, 469, 474, 479, 481, 512–513, 517, 519–521, 581, 586–588 Cortical thickness . . . . . . . . . . . . . . . . . . . . . . . . . . 93, 536–537 Cranioscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Creatine (Cr) . . . . . . . . . . . . . . . . . 6, 204–206, 213, 216, 272, 428, 453, 465, 468 Cryogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18, 23, 518, 530 CxP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398, 404–405, 411–414 Cytoarchitecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146, 151

B Ballistocardiogram (BCG) . . . . . . . . . . . . . . . . . . . . . . . . . . 308 Bandwidth . . . . . . . . . . . . . . . 75, 78, 181, 209, 214, 242–244, 278, 427, 431, 589 Beamformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Bioeffects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Bioluminescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 Biomarker . . . . . . . . . . . . . . . . . . . . . . 100, 190, 364, 422, 431, 536 Blood brain barrier . . . . . . . . . . 146, 148, 153, 207, 372, 380, 398, 453, 492, 519, 579–592 Blood oxygen level dependent (BOLD) . . . . . . . . . . . . 7, 305 Bloomberg Purcell Pound (BPP) theory . . . . . . . . . . . . 70–72 Brain ablation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579–592 Brain atlas . . . . . . . . . . . . . 150, 165, 251–269, 442, 444, 481, 499, 501, 526 Brain boundary shift integral algorithm . . . . . . . . . . . . . . 110 Brain extraction . . . . . . . . . . . . . . . . . . . . . . 111, 113–115, 119 Brain tumour . . . . . . . . . . . . . . . . . . . . . 30, 451–453, 465, 467 Bright blood sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328

D Deformation-based morphometry (DBM) . . . . . . . . . . . . . . 9 Deoxyhemoglobin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 Dielectric resonance artifact . . . . . . . . . . . . . . . . . . . . . . . . . 540 Diffusion tensor (DT) . . . . . . . . . . . 6, 8–9, 66, 95, 103–104, 127–142, 253, 255–257, 261, 357, 544 Diffusion-weighted imaging (DWI) . . . . . . . . 100, 474, 478

M. Modo, J.W.M. Bulte (eds.), Magnetic Resonance Neuroimaging, Methods in Molecular Biology 711, c Springer Science+Business Media, LLC 2011 DOI 10.1007/978-1-61737-992-5, 

595

MAGNETIC RESONANCE NEUROIMAGING

596 Subject Index

Directionally-encoded colour (DEC) maps . . . . . . . . . . . 132, 142, 258, 261 Dirty-appearing white matter (DAWM) . . . . . . . . . . . . . 100 Driven equilibrium single pulse observation of T1 (DESPOT1) . . . . . . . . . . . . . . . . . . . 34, 80, 82–84 Dysprosium (Dy) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187–188

E Echo planar imaging (EPI) . . . . . . . . . . . . . 5, 32, 37–38, 43, 47–49, 58, 80, 82, 129, 132–134, 275, 289, 310, 316, 332, 337–338, 341, 414, 477–479, 559–560, 573 Edema . . . . . . . . . . . . 30, 55–56, 72, 93, 178, 380, 453, 477, 540, 586 Electrocardiogram (ECG) . . . . 26, 138, 259, 312–313, 319, 336–337, 352, 384, 392, 516, 521 Electroencephalogram (EEG) . . . . . . . . . . . . . . 101, 303–322 Electromotive force (emf ) . . . . . . . . . . . . . . . . . . . . . . 308–309 Electron paramagnetic resonance imaging (EPRI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401–405 Endothelial adhesion molecules . . . . . . . . . . . . . . . . . . . . . 380 Epilepsy . . . . . . . . . . . . . . . . . . . . 101, 307, 321, 538, 570, 572 Equation model fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . 79–80 Explanatory variables (EV) . . . . . . . . . . . . . . . . . . . . . 560–562

F False discovery rate (FDR) . . . . . . . . . . . . 358, 360, 545–546 Family wise error rate (FWER) . . . . . . . . . . . . . . . . . . . . . . 360 Farraday’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24, 308 Fast fourier transform (FFT) . . . . . . . . . . . 49, 185–186, 196, 431–432 Fast low angle single shot (FLASH) . . . . . . . . . . 3, 7, 32, 48, 81–82, 85, 88, 163, 179, 277, 332, 376, 463, 478 Ferrozine assay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441 Field inhomogeneities . . . . . . . . . . . . . . . . . 35, 40, 47, 56, 75 Field strength . . . . 27, 31, 36, 47–48, 57, 66, 68, 71, 87, 97, 102–103, 132, 150, 152, 157–159, 161, 163, 167, 178, 203–204, 207, 209–210, 240–241, 309, 313, 316–317, 333, 341, 350, 374–375, 453, 513–514, 517, 519, 537, 539–540, 548, 573, 581 FLAIR . . . . . . . . . . . . . . . . . . . . . . 97, 100, 191, 216, 233–234 18-Fluorodeoxyglucose (FDG) . . . . . . . . . . . . . . . . . . . . . . . . 7 19F-MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Focussed ultrasound (FUS) . . . . . . . . . . . . . . . . 579–580, 586 Frequency maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245, 247 Full width half maximum (FWHM) . . . . . . . 298, 355, 359, 466–467, 560, 580 Functional MRI (fMRI) . . . . . . . . . . . . . . . . . . 6–9, 111, 153, 281–298, 303–322, 342, 374, 474, 478–481, 526, 551–564, 568–576

G Gadolinium (Gd) . . . . . . . . . . . . . 18, 43, 191, 228, 234, 406, 479, 519–520, 523, 525, 527, 586, 588–589 Gaussian blur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354–356 Generalized autocalibrating partially parallel acquisition (GRAPPA) . . . . . . . . . . . . . . . . . . . . . . . . . 134, 215 Ghosting artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 Glioma . . . . . . . . . . . . 190–191, 216, 228, 272–274, 451–469 Glomerular filtration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Glutamate . . . . 176–177, 204–205, 207, 219, 453, 465, 468 Glutamine . . . 145–146, 204–205, 207, 219, 382–383, 437, 439, 453–454, 465, 468

Gradient artefact . . . . . . . . . . . . . . . . 306–314, 318–319, 322 Gradient recalled echo (GRE) . . . . . . 78–79, 179, 181–183, 186, 190–191, 193–194, 242–244, 246–247, 275, 523 Gradient and spin echo (GRASE) . . . . . . . . . . . . . . . . . . . 332 Group modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561–562 Gyromagnetic ratio . . . . . . . . . . . 66, 133, 179, 519–520, 586

H Hemodynamic response . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282 Hepatic encephalopathy (HE) . . . . . . . . 30–31, 43, 206–207 Highly focussed ultrasound (HIFUS) . . . . . . . . . . . 192–193 HIV-related dementia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380 Hodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Hotelling’s T2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357 Huntington’s disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 262–263 Hyperbolic secant (HS) . . . . . . . . . . 101, 184, 208–209, 427 Hypercapnia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288

I Image processing . . . . . . . . . . . 109–126, 350–351, 354–355, 374, 438, 537, 544 Image registration . . 110–111, 113, 116–118, 124, 262, 351 Immunodeficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452, 454 Independent component analysis (ICA) . . . . . 313, 476, 559 Inflammation . . . . . . . . . . . . . . . . . . 55, 72, 93, 206, 379–395 Interclass-correlation coefficient (ICC) . . . . . . . . . . . . . . . 499 Inversion recovery (IR) . . . . 32, 59, 73–74, 80, 82, 153, 189, 191, 217, 230, 233, 339 Iron . . . . 10, 68, 72, 101, 239–247, 272, 364–365, 367, 369, 371, 375, 379–382, 384, 395, 436–437, 439–441, 444, 446, 488, 503–504, 509, 512–513, 522, 529–530

J Jacobian determinant . . . . . . . . . . . . 356–357, 360, 543–544

L Lactate (lac) . . . . . . . . . . . . . . . . 204–206, 209, 219, 426, 432, 453–454, 465, 468 Larmor frequency . . . . . . . . . . . . . . . . . . . . . . . . 66, 70–71, 317 Legislation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17, 19 Levenberg-Marquard algorithm . . . . . . . . . . . . . . . . . . . . . 132 Lewy bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101, 487, 538 Lipid suppression . . . . . . . . . . . 210, 212, 214, 216–218, 230 Liposome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403 Localization by adiabatic selective refocusing (LASER) . . . . . . . . . . . . . . . . . . . . . . . . . . . 208–209 Look-locker method . . . . . . . . . . . . . . . . . . . . . . . . . . 32, 80–82 Luxol fast blue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Lysine-rich protein (LRP) . . . . . . . . . . . . . . . . . . . . . . 272–275

M Magnetic particle imaging (MPI) . . . . . . . . . . . . . . . . . . . 6, 10 Magnetic resonance (MR) compatibility . . . . . . . . . 28, 317–318, 477, 521, 580–581 controlled area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22, 129 thermometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585–587 Magnetic resonance angiography (MRA) . . . . . . . . . . . . . . . 8 Magnetic resonance safety . . . . . . . . . . . . . . . . . . . . . . . . 17–28 Magnetic resonance spectroscopy (MRS) . . . . . . . . . . 5, 8–9, 103, 203, 422, 453–454

MAGNETIC RESONANCE NEUROIMAGING 597 Subject Index Magnetization recovery . . . . . . . . . . . . . . . . . . . . . . . 74, 81–82 Magnetization transfer ratio (MTR) . . . . . . . . . . . . . 235–236 Magnetophosphenes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23–24 Magnet quench . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18, 23–24 Mahalanobis metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482 Malaria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380 Manganese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145–170 Manganese enhance MRI (MEMRI) . . . . . . . 145–170, 569 Mannitol. . . . . . . . . . . . . . . . . . . . . . . .148, 155–156, 170, 205 Marmoset . . . . . . . . . . . . . . . . . . . . . . . 284, 286–288, 290–298 McDESPOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Metabolic concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453 Metabolomic profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422 Microbubble . . . . . . . . . . . . . . . . 580–581, 586–589, 591–592 Micron-sized particles of iron oxide (MPIO) . . . . . . . . . . 10, 380–395 conjugation . . . . . . . . . . . . . 381–382, 384, 386, 388, 394 Misclassification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548 Misregistration . . . . . . . . . . . . . . . . . . . . . . . 118, 137, 142, 358 Molecular imaging . . . . . . . . . . . . . . . . . . . . . . . . 366, 380, 569 Morphometric imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539 Motion correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 Mouse . . . . . . . 150, 160, 251–269, 350–352, 360, 368, 374, 383–384, 390–391, 475, 514–516, 526–527, 529, 590 mRNA . . . . . . . . . . . . . . . . 364–366, 368–369, 371, 376–377 Multi-component relaxometry (MCR) . . . . . . . . . . . . . . . 103 Multiple asymmetric spin echo (mASE) . . . . . . . . . . . . . . 522 Multiple quantum filters (MQF) . . . . . . . . . . . . . . . . 187, 189 Multiple sclerosis (MS) . . . . . . . . . 30, 34, 49, 55–56, 91, 96, 99–100, 215, 253, 380 Multi-spectral MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 Myelin water fraction (MWF) . . . . 34, 53–55, 59, 100, 102 Myo-inositol (mI) . . . . . . . . . . . . . . . 204–206, 453, 465, 468

N N3 algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354, 498, 540 23NA-MR. . . . 175, 177–181, 183–184, 186–188, 190–197 N-acetyl aspartate (NAA) . . . . . . . . 190, 204–207, 209, 216, 219–220, 431, 453, 465, 468 Nephrogenic systemic fibrosis (NSF) . . . . . . . . . . . . . . . . . . 26 Neurodegeneration . . . . . . . . . . . . . . . . . . . . . . . . . 68, 487–509 Neurofibrillary tangles (NT) . . . . . . . . . . . . . . . . . . . . . . . . 511 Neuronal plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567–576 Neuropsychology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Neurovascular coupling . . . . . . . . . . . 281–283, 285, 291, 298 Nitroxyl contrast agent . . 397–401, 403–404, 407, 411, 415 Noise reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 110–113, 123 Non-human primates . . . . . . . . . . . . . . . . . 169, 188, 283, 286 Non-negative least squares (NNLS) . . . . . . . . 35, 53–54, 60 Normal-appearing white matter (NAWM) . . . . 55–56, 100

Pareto scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429–430, 433 Parkinson’s disease . . . . . . 146, 401, 487–509, 552, 570, 572 Partial least squares-discriminant analysis (PLS-D) . . . . . . . . . . . . . . . . . . . . . . . . . . . 429–430 Perfusion . . . . . . . . . . . . . . . 26, 176, 190, 255–256, 259–260, 327–328, 330–333, 338–339, 343, 357, 377, 380, 411, 444, 474, 477, 490–502, 520, 526–528, 553, 563, 586 Peripheral nerve stimulation (PNS) . . . . . . . . . . . . . . . . . . . 24 Pharmacokinetics . . . . . . . . . . . . . . . . . . . . 272, 402, 555, 563 Pharmacological MRI (phMRI) . . . 552–553, 560–561, 564 Phase contrast. . . . . . . . . . . . . . . . . . . . . . . .329–330, 335–336 Phenotyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349–360 Phrenology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4, 7 Physiological monitoring . . 27–28, 152, 274, 282, 287–289, 298, 319, 350, 384, 427, 490, 496–497, 522, 570 Pittsburgh imaging compound B (PIB). . . . . . . . . . . . . . . 512 Pixel-by-pixel analysis . . . . . . . . . . . . . . . . . . . . . . . . . 482–483 Placebo . . . . . . . . . . . . . . . . . 26, 553, 557–558, 561–562, 564 Pneumoencephalography (PEG). . . . . . . . . . . . . . . . . . . . .4–5 Point-resolved spectroscopy (PRESS) . . 204, 208–211, 214, 218–220, 275, 427–428, 464–467 Poly-L-lysine . . . . . . . . . . . . . . . . . . . . . . . . 235, 272, 437, 439 Polymorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451, 556 Poon-Henkelman sequence . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Positron emission tomography (PET) . . . . . . . . . 5–8, 10–11, 103, 364, 436, 487, 512, 552, 563 Post-labelling delay (PLD) . . . . . . . . . . . . . . . . . . . . . 338–343 Precessional frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66, 70 Presenilin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513–514 Principal component analysis (PCA) . . . 312, 429–430, 546 Proteasome inhibition . . . . . . . . . . . . . . . . 488, 491–495, 508 Protein extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 Proton density . . . 30, 35–36, 39, 69, 98, 123, 246, 522, 591 Proton electron double resonance imaging (PEDRI) . . . 403 Proton EPSI (PEPSI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 Proton resonance frequency (PRF) . . . . . . . . . . . . . . 272, 517, 585, 591 Prussian blue . . . . . . . . . . 375, 440–441, 446–447, 491, 503, 509, 512, 520, 528–529 Pseudo CASL (pCASL) . . . . . . . . . . . . . . . . . . . . . . . 331–332 Psychopharmacology . . . . . . . . . . . . . . . . . . . . . . . . . . . 551–552 Pulse artefact . . . . . . . . . . . . . . . 307–310, 312–314, 319, 322 Pulsed ASL (PASL) . . . . . . . . . . . . . . . . . . . . . . 331–333, 344

Q Quality control (QC) . . . . . . . . . . . . 352, 498–499, 508, 537, 540, 542 Quantitative MRI (qMRI). . . . . . . . . . . . . . . . . . . . . . . .29–61 Quantitative T2∗ image (QUTE) . . . . . . 32, 35, 37, 39–43, 47–49, 57–58, 244, 246

O

R

Oligodeoxynucleotides (ODN) . . . . . . . . 364–366, 368, 376 Oligonucleotides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Optical projection tomography (OPT) . . . . . . . . . . . . . . . 253 Optimal basis sets (OBS) . . . . . . . . . . . . . . . . . . . . . . . 312–313 Overhauser MRI (OMRI) . . . . . . . . . . . . . . . . . 401, 403–404 Oxygen consumption (CMRO2 ) . . . . . . . . . . . . . . . . 282, 305

Radiofrequency (RF) . . . . . . . . . . . . . 10, 18, 25, 35, 39, 218, 229, 272–273, 350, 392, 479, 513, 517–518 burns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25–26 coil . . . . . . . . . . . . . . . 27, 86–87, 128, 178, 209, 274, 277, 286–287, 291, 315, 352, 427, 475, 478, 490, 495–498, 517, 519, 524–526, 528, 530 effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25–26 heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317–319 inhomogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 pulse . . . 25, 36, 42, 66–68, 76, 80–82, 84–89, 183–186, 193–194, 208–209, 211, 243, 272, 315, 329, 331–332, 334, 337, 478, 519, 522

P Parallel imaging . . . . . . . . 6, 57, 59, 129, 133–135, 215, 240, 243–245 Parametric maps . . . . . . . . . . . . . . . . . . . . . 128, 132, 142, 356

MAGNETIC RESONANCE NEUROIMAGING

598 Subject Index

Radioprotection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400 Rapid acquisition with relaxation enhancement (RARE) . . . . 160–161, 275, 277–278, 374, 376, 443–444, 463–465, 478 Rat . . . . . . . . . . . . . . . . 59, 150–152, 154–155, 157, 159–160, 164, 169, 191, 272–274, 286, 289, 333, 337, 340, 343, 381–384, 387, 389–392, 404–405, 411–413, 437, 452, 454, 456–458, 460–464, 466–469, 477, 481, 491, 502, 520, 526, 573–575, 588, 590 Reactive oxygen species (ROS) . . . . . . . . 398–401, 406, 411 Reconstruction methods . . . . . . . . . . . . . . . . . . . . . . . 185, 245 Redox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397–415 Relaxation time mapping . . . . . . . . . . . . . . . . . . . . . . 37–38, 69 Reporter gene . . . . . . . . . . . . . . . . . . . . . . . . . . 6, 271–279, 436 Respiratory compensation . . . . . . . . . . . . . . . . . . . . . . 240–242 Rotational asymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . 495–496

S Safety policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19, 555 Safety screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20–21 Saturation binding assay . . . . . . . . . . . . . . . . . . . . . . . . 370–371 Schizophrenia . . . . . . . . . . . . . . . . . . . . . . . . . . . 34, 61, 91, 322 Selective serotonin re-uptake inhibitors (SSRI). . . . . . . . 552 Sensitive encoding (SENSE) . . . . . . . . . . . . . . 134, 215–216, 230–231, 233, 244–245, 364–365, 368–369, 376–377, 427, 476 Sensitivity improvement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240 Simultaneous acquisition of spatial harmonics (SMASH) . . . . . . . . . . . . . . . . . . . . . . . . . . 215, 244 Single photon emission computed tomography (SPECT) . . . . . . . . . . . . . . . . . . . . . . . . . 5, 10, 436 Single point ramped imaging with T1 enhancement (SPRITE) . . . . . . . . . . . . . . . . . . . . . 179, 183–185 Singular value decomposition (SVD) . . . . . . . . 430–433, 481 Smoothing . . . . . . . . . . . . . . . . . . . . . . . 95, 467, 541, 558, 560 Somatosensory stimulation . . . . . . . . . . . . . . . . . . . . . . . . . . 480 Song-birds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Spatial normalization . . . . . . . . . . . . . . . . . . 94, 259–263, 541 Specific absorption rate (SAR) . . . . . . 25–26, 195, 230, 236, 317–318, 331 Spin-echo full intensity acquired localized spectroscopy (SPECIAL) . . . . . . . . . . . . . . . . . . . . . . . . 208, 210 Spiral MRSI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213–215 Spoiled gradient recalled echo (SPGR) . . . . . . . . . 68–69, 82 Statistical parametric mapping (SPM) . . . . 6, 110, 526, 537, 542–544, 554, 560 Steady state free precession (SSFP) . . . . . 34–35, 80, 84–85, 88–89 Stejskal-Tanner equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Stem cell . . . . . . . . . . . . . . . . . . . . 10, 421–433, 435–448, 452 Stereology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491, 502 Stimulated echo acquisition mode (STEAM) . . . . 208–210, 217, 464–465 Stroke . . 9, 11, 100–101, 175, 178, 191–192, 228, 236, 253, 343, 380, 401, 441, 473–483, 521, 538, 568, 570, 572–575

Superparamagnetic iron oxide particles (SPIO) . . 382, 436, 443–444, 447 Susceptibility maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245–246

T T1 mapping with partial inversion recovery (TAPIR). . . 32, 37–43, 56–58 TACQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Tempol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400, 411–412 Thermal ablation . . . . . . . . . . . . . . . . . . . . . . . . . 580, 584–586 Thioflavine S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512–513 Time-of-flight (TOF) . . . . . . . . . . . . . . . . 329–330, 333–336, 340–341 Tissue classification . . . . . . . . . . . . . 110–111, 114–116, 119, 121, 124 Tissue sodium concentration (TSC) . . . . . . . . 175, 177–180, 186–187, 190–192, 196–197 Tomography . . . . . . . . . . . 4–6, 103, 253, 364, 436, 474, 487, 512, 563 T One by Multiple Read Out Pulses (TOMROP). . . . . . 80 Tractography . . . . . . . . . . . . . . . . . . . . . . . . . . 95, 104, 128, 134 Tract tracing . . . . . . . . . . . . . . . . . . . . 147, 149, 155, 161–162, 168 Transfection agent . . . . . . . . . . . . . . . . . . . . 436–437, 439, 447 Transplantation . . . . . . . . . . . . . 271, 277, 435–436, 438–442, 444–446

U Ultra-short TE (UTE) . . . . . . . . . . . 183, 185–187, 193, 195 Ultrasound (US) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579–592 Unit variance (UV) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429–430

V Vasculature . . . . . . . . . . . . . . . . . 305, 343–344, 386, 531, 570 Vasculature cell adhesion molecule (VCAM). . . . . 382–386, 389–391, 393 Vasogenic oedema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453 Velocity encoding (VENC) . . . . . . . . . . . . . . . . 335–336, 341 Voxel-based morphometry (VBM) . . . . . . . 6, 9, 59, 93, 488 Voxel lesion-symptom mapping (VLSM) . . . . . . . . . . . . . 6, 9

W Warping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342, 540–541, 559 Wash-out periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557 Water mapping . . . . . . . . . . . . . . . . . . . . 32, 35–36, 39–48, 59 Water suppression . . . . . . . . . . . . . . . 203, 207, 210–211, 214, 217, 427–428, 432 White matter tracts . . . . . . . . . . . . 93, 95, 102, 255–256, 535

X X-ray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–6, 253, 400, 436

Z Z-spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229, 232–235, 278