ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI
Anisotropically Restricted Diffusion in MRI Michael E. Moseley and Alex de ...
52 downloads
2096 Views
95MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI
Anisotropically Restricted Diffusion in MRI Michael E. Moseley and Alex de Crespigny Stanford University, CA, USA
diffusion is rarely restricted but is rather slowed by the presence and extent of permeable or semipermeable membranes. The complete concept of proton diffusion then is that of a tensor, where the diffusion rate or ADC value is a function of the diffusion direction in three-dimensional space. As the tissue water protons constantly `experience' the three-dimensional microenvironment through the processes of permeability, exchange, ¯ow, and thermally-induced diffusion, their traveled path hLi along any one given direction (along the magnet bore direction, for example) can be estimated by:9 hL2 i 2T
ADC
1 INTRODUCTION A dominant source of proton MR signal loss following radiofrequency excitation is caused by dephasing. While coherent ¯ow will cause dephasing, other motions can contribute. One important mechanism of dephasing is that of a random diffusion of protons. This random phase shifting leads to signal loss because of phase cancelation. While T1 and T2 relaxation times re¯ect complicated frequency-dependent rotational and proton exchange processes, diffusion is caused solely by random proton displacements or translations through a tissue. Since the phenomenon of proton diffusion in vivo is complicated, the term `apparent diffusion' described by the apparent diffusion coef®cient, ADC, has been used to denote signal loss induced by random proton ¯uctuations. This applies to protons in water, phosphorus in high-energy metabolites, indeed to any MR-observable nucleus.1±3 Considering water proton displacements in pure isotropic solutions, where the probability that a water proton will diffuse in any given direction is equal, the rate of random proton phase cancellation due to diffusion will be same regardless of the direction observed. This is termed `isotropic' diffusion and is seen in homogeneous environments. In ordered environments, such as that found in liquid crystals for example, the resulting diffusion coef®cient will depend on the direction of the motion being considered and on any geometric anisotropy created by the nature of the microenvironment.4,5 For water protons diffusing or moving within a tissue matrix, the observed diffusion rate and direction will re¯ect the molecular and macromolecular barriers or hindrances that the proton experiences during the translational process. 2 APPARENT DIFFUSION COEFFICIENTS AND OBSERVATION DIRECTION Simply stated, proton apparent diffusion is slowed if the protons are hindered in their random motion.6±8 Hindrances can be impermeable barriers that completely restrict diffusion. In living in vivo tissues, however, water protons are probably slowed solely by the presence of membranes, cell walls, and macromolecules due to differences in water proton permeability. More speci®cally, myelin ®bers and neuro®brils in white matter tracts possess a special hindrance for water proton movement which is greater in one direction than another. This causes protons to diffuse faster along any direction of least resistance, in this case along the long axis of the tracts, causing the apparent diffusion coef®cients to be anisotropic or directionally-dependent. It is important to note that in vivo proton
1
1
where over an observation time of T (s) the mean square displacement is expressed by an apparent diffusion coef®cient, ADC (usually given in m2 sÿ1 or cm2 sÿ1). Typical proton diffusion coef®cients, ADC, of 110ÿ5 cm2 sÿ1 are observed in cat cerebral gray matter,10 the average mean square displacement along that one given observation direction hLi becomes 1 (2 40 ms 10ÿ5 cm2 sÿ1)2 or 9 m, observed over an effective observation time, (T), of 40 ms. To date, most diffusion-weighted MR imaging and spectroscopy sequences have been based upon a spin echo Stejskal±Tanner (ST) sequence,11±13 either as the sequence itself or as a magnetization preparation set of prepulses, prior to another MR sampling scheme. Since the two ST `diffusionsensitizing' gradient pulses are symmetrical, all proton spin dephasing caused by the ®rst diffusion-sensitizing gradient pulse will be refocused (spin-rephased) by the second diffusion-sensitizing gradient pulse for stationary spins. Randomlymoving spins (because of ¯ow, perfusion, diffusion, etc.) will not completely refocus and will attenuate the observed echo or echo train. Note that mild coherent motions will result in phase shifts and not in signal attenuation without phase cancellation. The observed echo or echo train signal intensity S(Gi) can be expressed as: S
Gi S
0eÿbD
2
where the b value is de®ned12,13 by equation (3): b 2 2 G2
ÿ =3
3
and S(0) is the signal obtained without applied diffusion-sensitizing gradient pulses, is the gyromagnetic ratio, and G are the duration and the amplitude of the diffusion-sensitizing ST gradient pulses along a given direction, is the time interval between the leading edges of the diffusion-sensitizing gradients pulses and D is the water apparent diffusion coef®cient (often expressed as the ADC value).12,13 Measurement of apparent diffusion from MR images would require plotting the natural logarithm of the image intensities against the b value. The slope of this observed monoexponential decay directly leads to a diffusion coef®cient.11 There exist three major advantages to using a `pulsed' gradient approach for measuring in vivo diffusion. One is that the stronger pulsed gradient diffusion-sensitizing pulses overcome poor magnet homogeneity (which upon re¯ection, is a sort of residual magnetic ®eld gradient itself).11 Also, because the diffusion-sensitizing gradients are relatively short pulses, the time during which the sequence is sensitive to diffusion, ( ÿ /3), can be accurately controlled. From a knowledge of this obser-
2 ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI vation time, ( ÿ /3), the mean path length can be estimated from the above Einstein displacement equation given the measured diffusion coef®cient [equation (1)]. A ®nal advantage is that the direction of the diffusion-sensitizing gradient pulses can be controlled accurately. All MRI scanners possess three orthogonal and linear magnetic ®eld gradient coils, any or all of which can be pulsed to produce the diffusion sensitization, the effective angle of which is then the geometric sum of the pulsed gradients. This opens the path to mapping of the apparent diffusion tensor.14,15 The complete concept of proton diffusion is then that of a tensor14,15 where the diffusion rate or ADC value is a function of the diffusion direction in three-dimensional space. Thus the tensor for free, isotropic, diffusion can be visualized as a sphere, with the same ADC in all directions, while the anisotropic diffusion of a white matter tract can be viewed as an ellipsoid tilted at some angle to the three principal axes, depending on the particular ®ber orientation. Diffusion is then fastest along the principal or long axis of the ellipsoid, and takes a lower (but equal) value along the two orthogonal axes. The `trace' of this ellipsoidal diffusion tensor is just the average of the ADC values measured along any three orthogonal directions, and is independent of ®ber orientation thus yielding a useful measure of the average rate of diffusion without the complications of ®ber directionality.
3 MAPPING AND THE ORIGIN OF THE APPARENT DIFFUSION ANISOTROPY The b value-dependent diffusion effects on image intensity can be displayed in different ways. A diffusion-weighted image contains typically T1-, T2-, and diffusion-weighting, the amount of the diffusion-weighting being dependent on the b value. Within diffusion-weighted images, relative faster intravoxel motion will be observed as increased attenuation of the signal from that voxel, and consequently an observed hypointensity from that voxel as the b value is increased. Conversely, slower motions or diffusion leads to smaller signal attenuation. The observed result is that of regional relative hyperintensity. This is best seen in lipid regions outside the brain where lipid protons experience little or no diffusion because of the molecular size and environment of the long chain hydrocarbons. A pure diffusion of ADC image is calculated from the acquisition of two or more images of varying b values. ADC images eliminate T1- and T2-weighting; voxel intensities re¯ect only the apparent diffusion.12,13 The ADC images are produced by ®tting equation (2) to a series of two or more diffusionweighted images of varying b value. The ADC images are then displayed with the voxel intensity related or scaled to the rate of diffusion. Visualization of cerebral details is sometimes easier from diffusion-weighted images, which are devoid of cerebrospinal ¯uid (CSF) hyperintensity because of the relatively faster apparent diffusion of water protons in CSF. In the corresponding ADC images, the pronounced CSF hyperintensity (re¯ecting faster apparent diffusion) can often limit the visualization of more subtle ADC differences in gray and white matter. An appropriate example of ordered or anisotropic diffusion can be seen in proton diffusion-weighted images of ®brous vegetables or fruits. Apparent diffusion along the direction of
ordered ®bers (measured by application of the diffusion-sensitizing gradient parallel to the long axis), renders the diffusion-weighted intravoxel intensity within the ®bers hypointense, implying fast diffusion of water protons along the ®bers (since rapid diffusion produces signal attenuation on diffusion-weighted images). Corresponding images acquired with the diffusion-sensitizing gradient perpendicular to the long axis of the ®bers result in very hyperintense image intensity within the ®bers, implying relatively slow proton diffusion (little echo loss) along this direction. The translational distances traveled by water protons perpendicular and parallel to ®brous matter during the MR observation time ( ÿ /3) of roughly 20±50 ms is estimated to be 5±13 m (respectively). If water proton displacements in any given direction are signi®cantly shorter that the statistical diameter of a structure of barriers along that direction, then little or no anisotropy will be observed. Anisotropy is apparent when the water protons within a voxel can sample the microenvironment adequately. The rationale behind diffusional anisotropy in an ordered structure is that water protons within a matrix statistically prefer to move along ordered structures rather than across them (by the principle of least resistance). In other words, average proton displacements along the ®bers are greater due to fewer barriers being present in this direction. Proton displacements are shorter across the ®bers due to barriers presented by the very orientation of the ®bers themselves. When the diffusionsensitizing gradient direction coincides with the direction of least hindrance, the apparent diffusion appears greater and the observed ADC is larger. Anisotropy of diffusion was ®rst suggested as a possible explanation for the large regional differences in in vivo human white matter apparent diffusion coef®cients by Thomsen et al.16 More recent descriptions of the existence of and characterization of in vivo anisotropy have subsequently appeared.10,17,18 In axial and coronal diffusion-weighted images in animal models,10 the directional dependence of the white matter apparent diffusion is not observed in regions of gray matter. Within the corpus callosum, however, apparent diffusion can be as slow as 0.4 0.1 10ÿ5 cm2 sÿ1 (across the short axis of the tract) and as high as 1.3 0.1 10ÿ5 cm2 sÿ1 (along the long axis of the corpus callosum). This suggests faster proton diffusion in white matter along the direction of the long axis of the tract. From the observed image intensities and knowledge of the diffusion-sensitizing gradient direction, one can then ascertain the orientation of individual white matter tracts. The translational distances traveled by water protons perpendicular and parallel to the white matter orientation during ( ÿ /3 =) 40 ms can be estimated to be 6 and 11 m respectively, suggesting that axonal diameters may be determinable from such measurements.
4
CLINICAL IMPLICATIONS
The in vivo directional diffusion-weighted images observed in human volunteers indicate that water proton diffusion anisotropy is best observed in the large diameter, fast-conducting motor and somatosensory nerve ®bers (Figures 1±5). Signi®cant anisotropy is observed in adult volunteers in all white matter tracts such as the corpus callosum, splenium, and in-
ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI
Figure 1 Axial, multislice diffusion-weighted images of a human volunteer acquired on a GE Signa at 1.5 T. Top row: minimal diffusion-weighting (b = 12 s mmÿ2). Second row: diffusionsensitizing gradient applied along AP direction (up±down in images), b = 512 s mmÿ2. Third row: diffusion-sensitizing gradient applied along LR direction (left±right in images), b = 512 s mmÿ2. Bottom row: diffusion-sensitizing gradient applied along IS direction (through-plane in images), b = 512 s mmÿ2. All images were acquired with TR = 2 cardiac cycle (~1500 ms), TE = 102 ms and 124 ms (second echo used for navigation), 1 NEX. Gradient durations = 45 ms, gradient separation = 62 ms, gradient strengths = 1 mT mÿ1 (1 G cmÿ1), FOV = 24 cm, 128 256 matrix size, and scan times = 4.3 min for eight slices per TR. Navigation achieved by correction of navigational echo magnitude. CSF suppression not applied. All images are scaled except top row. On these images, containing T1-, T2-, proton density-, and diffusion-weighting, relative regions of hypointensity denote rapid apparent diffusion of protons. Regions of relative hyperintensity demarcate slower diffusion. Note little or no anisotropic effects in gray matter or lipids (which display very slow apparent diffusion). TR, repetition time; TE, echo time; NEX, average; FOV ®eld of view
ternal capsules. In addition, myelinated tracts in spinal cord,10 optic (A. de Crespigny, unpublished data), and peripheral nerves20 appear to be associated with signi®cant anisotropic effects. Several pioneering studies verifying white matter diffusional anisotropy in normal and abnormal human and neonate subjects have been presented.18,21±25 It has been shown that the newborn brain demonstrates isotropic diffusion of water molecules.22,23 As the brain matures, diffusion becomes anisotropic, being greater along the longitudinal axes of the major axonal bundles of the cerebral white matter than perpendicular to them. This anisotropic ordering of water motion most likely occurs along the long axis of the individual neuro®brils as well as along or within the axons isolated by the myelin sheath. Studies indicate that the presence of myelin is not necessary for ordering of proton apparent diffusion since anisotropic proton mobility is seen in nerves lacking microtubules and fast axonal
3
Figure 2 Diffusion-weighted images chosen from one slice and taken from images in Figure 1. Top left: minimally diffusion-weighted; top right: diffusion-sensitizing gradient applied along AP direction (up± down in images), b = 512 s mmÿ2. Bottom left: diffusion-sensitizing gradient applied along LR direction (left±right in images), b = 512 s mmÿ2. Bottom right: diffusion-sensitizing gradient applied along IS direction (through-plane in images), b = 512 s mmÿ2. Note regional hypointensity in white matter tracts when applied diffusion-sensitizing gradient direction is parallel to white matter orientation. This is best seen in the image at top right, in the lower limbs of the splenum of the corpus callosum
transport.26 It is reasonable to assume that the development of anisotropy is related to the development of myelin sheaths; however, the temporal evolution of anisotropy seems to precede that of myelination as described by most authors.27 Additionally, diffusional anisotropy has been proved to occur in nonmyelinated nerve bundles.26 Another possibility is that diffusion anisotropy develops in relation to the maturation of the microtubules, intraaxonal organelles that facilitate axonal transport.28,29 Thus, diffusion-weighted imaging (DWI) seems to be useful in the real time assessment of cerebral maturation. The use of diffusion-weighted imaging for the determination of water proton apparent diffusion anisotropy may aid in the study of peripheral CNS tissues, which exhibit water proton diffusional anisotropy effects. One novel application, termed `MR neurography'20 can increase tissue contrast by using MR diffusion-weighting in a variety of directions to evaluate the diffusion tensor found in nerve tissue. An important adjunct to neurography is the use of fat saturation to reduce hyperintense artifacts due to the slow mobility of protons in lipid tissues and in the use of maximum intensity projection (MIP) methods to track nerve tissue in a MR angiography-like fashion. Most clinical usage of diffusion-weighted MRI has been hampered by two signi®cant problems, gradient strength and motion artifacts. Many clinical scanners possess gradients capable of strengths of no more than 10 mT mÿ1 (1 G cmÿ1), limiting b values to roughly 300 s mmÿ2 along any given direction. These limitations typically require pulsing two or all
4 ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI three gradients together for longer durations (with increases in TE to well over 100 ms) to push the b value to above roughly 500 s mmÿ2, above which diffusion effects can be clearly seen above the inherent T2-weighting. This in turn limits the signalto-noise ratio (SNR) and degrades image quality, an important factor in diffusion imaging since good diffusion-weighting will itself attenuate SNR exponentially. More importantly, longer TE values lead to signi®cant motional artifacts. Linear excursions of just a few millimeters can easily ruin images where diffusion-weighted tissue contrast arises from proton displacements on the order of microns. The solutions employed to date have included restraint of the head with rigid holders or vacuum cushions that seal the head in the coil. These are uncomfortable and are dif®cult to implement on
Figure 4 Enlargements from the ADC and anisotropy maps for one slice taken from the images shown in Figure 3. Top left: ADC image with diffusion-sensitizing gradient applied along AP direction (up± down in images). Top middle: ADC image with diffusion-sensitizing gradient applied along LR direction (left±right in images). Top right: ADC image with diffusion-sensitizing gradient applied along IS direction (through-plane in images). Bottom left: trace image. Bottom middle: standard deviation image
uncooperative patients, particularly when several images are required to map the apparent diffusion tensor or to synthesize ADC images. Fast-scan and high-speed MRI capable of acquiring diffusion-weighted images in seconds or less requires typically large gradient strengths (not currently present on most clinical scanners) and has been limited to relatively low-resolution images which make proper visualization of the individual white matter tracts very dif®cult. The use of multiple shots or echo trains in
Figure 3 Series of three diffusion-weighted images were acquired at b values of 12, 350, and 512 s mmÿ2 to produce the pure diffusion (ADC) images shown here for the three gradient axes and for the same four slices shown in Figure 1. Top row: diffusion-sensitizing gradient applied along AP direction (up±down in images). Second row: diffusion-sensitizing gradient applied along LR direction (left±right in images). Third row: diffusion-sensitizing gradient applied along IS direction (through-plane in images). Note that regional hyperintensity on ADC images indicates rapid apparent diffusion. CSF hyperintensity is an example of fast apparent motion or diffusion, so much so that smaller variations in gray and white matter are masked out and dif®cult to appreciate. For this reason, diffusion-weighted or other image synthesis is needed. Fourth row: images of the `trace' of the diffusion tensor determined for each of the slices shown in the top three rows. The trace of the tensor is essentially an average of the diffusion images from the three principal axes and is useful for removing orientation effects from diffusion-weighted or ADC images. The trace image is useful for discrimination of ADC alterations due to stroke or other cerebrovascular diseases,19 which is often dif®cult because of regional ADC differences due to white matter anisotropy. Bottom row: the synthesized `standard deviation' images19 taken from the ADC images and the trace image. The SD image is essentially a pixel-by-pixel map of the deviations from the trace and can be very useful for the identi®cation of white matter tracts and discrimination from stroke or regional gray matter
Figure 5 Enlargements of another slice from the image set in Figure 3; image layout is as for Figure 4. This slice clearly shows the genu of the corpus callosum in the center of the ADC images (top row), which is not seen in the trace image (bottom left) but stands out in the anisotropy image (bottom middle) due to the strong orientation effects of the corpus callosum
ANISOTROPICALLY RESTRICTED DIFFUSION IN MRI
high-speed imaging to improve resolution or SNR will in turn render the images again sensitive to motional artifacts. Phase navigation30,31 combined with a conventional single or multi echo imaging sequence has great potential for clinical application. It involves simple modi®cations of standard clinical pulse sequences using additional spin- or gradient-echoes acquired without phase encoding to map and correct each phase-encoded line or train of k-space. This can yield relatively motion-free images of excellent resolution with limited gradient strengths at the expense of conventional scan times. Uncooperative patients can create motion artifacts not easily correctable with linear phase navigation methods (A. de Crespigny, unpublished data). However, with motion correction or navigation, a multishot approach (using EPI, RARE, or spiral k-space coverage) could become the optimal apparent diffusion imaging technique, using conventional gradients to build resolution in much shorter scan times. This would be fast enough to acquire a series of images with increasing b value or varying gradient direction in order to generate an ADC or an orientation map. Utilization of this unique MRI technique for determining the orientation of white matter through the measurement of water proton apparent diffusion has the potential greatly to improve our understanding and assessment of demyelination disorders, white matter infarcts with associated edema development,32 neoplasms involving white matter tracts, and of neonatal brain and spinal cord development. Signi®cant proton anisotropic effects have also been observed in human arm muscle, 33 opening up the potential of using this technique to map performance or pathology. Finally, images devoid of white matter orientation effects can be synthesized by mapping the trace of the tensor enabling better visualization and earlier detection of ischemia in the CNS.19 5 RELATED ARTICLES Diffusion: Clinical Utility of MRI Studies; Image Formation Methods; Intracranial Infections; Magnetic Resonance Imaging of White Matter Disease; Spin Warp Data Acquisition. 6 REFERENCES 1. G. E. Wesbey, M. E. Moseley, and R. L. Ehman, Invest. Radiol., 1985, 19, 484. 2. C. T. W. Moonen, P. C. M. van Zijl, D. LeBihan, and D. Despres, Magn. Reson. Med., 1990, 13, 467. 3. M. E. Moseley and P. Stilbs, Chem. Scr., 1980, 15, 215. 4. M. E. Moseley and A. Loewenstein, Mol. Cryst. Liq. Cryst., 1982, 90, 117. 5. M. E. Moseley, J. Phys. Chem., 1983, 87, 18. 6. E. O. Stejskal and J. E. Tanner, J. Chem. Phys., 1965, 43, 3579. 7. J. E. Tanner and E. O. Stejskal, J. Chem. Phys., 1968, 49, 1768. 8. R. L. Cooper, D. B. Chang, A. C. Young, C. J. Martin, and D. Ancker-Johnson, Biophys. J., 1974, 14, 161. 9. A. Einstein, `Investigations on the Theory of the Brownian Movement', Dover Publications, New York, 1956. 10. M. E. Moseley, Y. Cohen, J. Kucharczyk, J. Mintorovitch, H. S. Agari, N. F. Kenland, J. Tsurula, and D. Norman, Radiology, 1990, 176, 439.
5
11. E. O. Stejskal and J. E. Tanner, J. Chem. Phys., 1965, 42, 288. 12. D. LeBihan, E. Breton, D. Lallemand, P. Grenier, E. Cabanis, and M. Laval-Jeantet, Radiology, 1986, 161, 401. 13. D. LeBihan, E. Breton, D. Lallemand, M. Aubin, J. Vignaud, and M. Laval-Jeantet, Radiology, 1988, 168, 497. 14. P. J. Basser, J. Mattiello, and D. LeBihan, J. Magn. Reson. Ser. B, 1994, 103(3), 247. 15. J. Mattiello, P. J. Basser, and D. LeBihan, Magn. Reson., Med. B, 1997, 37(2), 292. 16. C. Thomsen, O. Henriksen, and P. Ring, Acta Radiol., 1987, 28, 353. 17. D. Chien, R. B. Buxton, K. K. Kwong, and B. R. Rosen, J. Comput. Assist. Tomogr., 1990, 14, 514. 18. W. M. Chew, J. Tsuruda, and M. E. Moseley, Radiology, 1990, 177, S121. 19. P. van Gelderen, M. M. de Vleeschouwer, D. DesPres, J. Pekar, P. C. M. van Zijl, and C. T. W. Moonen, Magn. Reson. Med., 1994, 31, 154. 20. F. A. Howe, A. G. Filler, B. A. Bell, and J. R. Grif®ths, Magn. Reson. Med., 1992, 28, 328. 21. T. L. Chenevert, J. A. Brunberg, and J. Pipe, Radiology, 1990, 177, 401. 22. H. Sakuma, Y. Nomura, K. Takeda, T. Tagami, T. Nakagawa, Y. Tamagawa, Y. Ishi, and T. Tsukomoto, Radiology, 1991, 180, 229. 23. M. A. Rutherford, F. M. Cowan, A. Y. Manzur, L. M. Dubowitz, M. Pennock, J. V. Majnal, I. R. Young, and G. M. Bydder, J. Comput. Assist. Tomogr., 1991, 15, 188. 24. M. Doran, J. V. Hajnal, N. Van Bruggen, M. D. King, I. R. Young, and G. M. Bydder, J. Comput. Assist. Tomogr., 1990, 14, 865. 25. J. V. Hajnal, M. Doran, and G. M. Bydder, in `Magnetic Resonance Imaging', eds. D. D. Stark and W. G. Bradley, Mosby, St. Louis, MO, 1991, Vol. 2, p. 1081. 26. C. Beaulieu and P. S. Allen, Proc. 11th Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 1728. 27. A. J. Barkovich and D. M. Wimberger, in `Magnetic Resonance Neuroimaging', eds. J. Kucharczyk, M. Mosley, and A. J. Barkovich, CRC Press, Boca Raton, FL, 1993. 28. F. Espejo and J. Alvarez, J. Comp. Neurol., 1986, 250, 65. 29. V. Faundez and J. Alvarez, J. Comp. Neurol., 1986, 250, 73. 30. R. A. Ordidge, J. A. Helpern, Z. Qing, and R. A. Knight, Proc. 11th Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 1902. 31. R. Asato, T. Tsukamoto, R. Okumura, Y. Miki, E. Yoshitome, and J. Konishi, Proc. 11th Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 1226. 32. T. Ebisu, S. Naruse, Y. Horikawa, S. Ueda, C. Tanaka, M. Uto, M. Umeda, and T. Higuchi, J. Magn. Reson. Imaging, 1993, 3, 863. 33. M. E. Moseley, and M. F. Wendland, Proc. 10th Ann Mtg. Soc. Magn. Reson. Med., San Francisco, 1991, p. 108.
Biographical Sketches M. E. Moseley. b 1951. B.S., 1973, North Carolina State University, Ph.D., 1980, Uppsala University. Postdoctoral work at Weizmann Institute of Science, Rehovot (with Z. Luz). Faculty at UCSF, 1985± 93; Stanford, 1993±present. Approx. 150 publications. Research interests include MR methods of mapping diffusion, perfusion and functional physiology. A. de Crespigny. b 1965. Ph.D., 1991, Cambridge University. Postdoctoral work at UCSF. Approx. 22 publications. Research interests include MR methods of mapping anisotropy in neonates, high-speed MRI, and image display and fusion.
BRAIN PARENCHYMA MOTION OBSERVED BY MRI
Brain Parenchyma Motion Observed by MRI Van J. Wedeen Harvard Medical School, USA
and Brigitte Ponceleti Massachusetts General Hospital, MA, USA
which descend at velocities as great as 5±10 mm sÿ1. By contrast, the cerebral mantle is almost immobile with respect to the calvarium. Velocities of intermediate size and complex 3D form are present in central deep gray and midbrain structures, among them the compressive motion of the thalami postulated by DuBoulay. Superimposed on the cardiac pulsation is a respiratory pulsation of a much smaller size.2 Observations of brain parenchymal strain patterns (see Section 5.1) suggest that variation in vascular turgor, particularly of the large arteries, may in¯uence the observed parenchymal motion. MRI measurement of CNS motion may be confounded by head movements of musculoskeletal or cardiac-ballistic origins.3
3 1 INTRODUCTION MRI studies of brain motion afford a unique window onto the mechanical homeostasis of the central nervous system (CNS). Velocity-sensitive MRI techniques may be based upon modulation of the magnitude or phase of the NMR image data, and have the capacity to detect the distributions of velocities, the displacements, or the material strainrates within the brain parenchyma. Having the capacity to resolve tissue displacements on the scale of micrometers, limited in practice by proton diffusion, NMR movies have mapped the physiological pulsation of the brain and spinal cord with the heartbeat and have measured brain parenchymal deformations due to everyday cranial accelerations. Clinical applications are projected to include the prediction of response to surgical treatment for a tethered spinal cord, and, potentially, the diagnosis of normalpressure hydrocephalus. Brain motion studies will be of value in elucidating the mechanisms of CNS trauma of both exogenous and endogenous forms.
1
OVERVIEW OF MRI OF CNS MOTION
CNS motion has been assessed by a variety of NMR imaging techniques. The earliest MRI demonstration of CNS motion was offered by Feinberg and Mark, who in 1987 presented echocardiogram (ECG)-gated movies of Fourier velocity spectra along selected 1D lines within the head, as shown in Figure 1.6,11 (For pioneering work on Fourier transform imaging spectra, antecedent to the present techniques and to phase contrast MRA, see Moran.12) Later investigators have estimated mean velocity using reduced numbers of velocity phase encoding steps and increased spatial or temporal resolution in conventional 2D multishot MRI,13±15 echo planar MRI,10 and realtime 1D-selective line MRI (`M-mode' NMR).16 The technique of NMR `interferography' of Hennig is a variant type of velocity encoding, based on a spatial modulation of local signal magnitude rather than phase.17 A quite different approach to CNS pulsation was taken by Lee, Wang, and Mezrich, who measured pulsatile changes in the volume of the lateral ventricles using conventional MRI.18
2 BRAIN MOTION Physiological brain motion re¯ects the reaction of the brain parenchyma, spinal cord, and cerebrospinal ¯uid (CSF) to changes in arterial and venous pressure and volume with the cardiac pulse and respiration.1 According to the Monro±Kellie doctrine, cranial entry of an arterial blood volume bolus with each systole must be matched by an equal egress of blood, CSF, and brain parenchyma from this incompliant compartment.2±4 Ultimately, parenchymal egress is effected by the movement of the cervical cord caudally through the foramen magnum. A detailed picture of the physiological brain motion has gradually emerged. DuBoulay postulated that CSF pulsation is pumped by the brain parenchyma, speci®cally the compression of the third ventricle by a medial motion of the thalami.5 MRI has since con®rmed this and has been instrumental in the emergence of a more complete view. Anatomically, brain parenchymal pulsation may now be seen chie¯y as a radial in¯ow of brain tissue inward and downward toward the foramen magnum: a herniation in miniature.6±10 For a superb treatment of multiaxial cine velocity data, see Greitz et al.8 Most rapid motion is seen in the cervical cord and brain stem,
Figure 1 Fourier-encoded cine velocity study for a selected line of tissue at the level of the foramen of Monroe. Cardiac gate delays from the R-wave are denoted. Cranio caudal velocities are depicted on the vertical axis, with single-pixel resolution of 0.4 mm sÿ1. Line selection was performed using orthogonal slice selects. The basic downward ``piston'' motion of the midbrain is demonstrated. (Reproduced by permission from Feinberg and Mark6)
2 BRAIN PARENCHYMA MOTION OBSERVED BY MRI 4 MOTION-SENSITIVE ECHO PLANAR MRI OF BRAIN MOTION The displacements associated with brain motion range from a typical maximum of 1±2 mm in the cervical cord down to unmeasurably small values in the cerebral mantle. Consequently, brain motion justi®es the highest feasible levels of motion sensitivity. MRI can achieve high velocity sensitivities by the method of magnetic gradient phase encoding, familiar in magnetic resonance angiography (MRA) and diffusion imaging and more recently in the imaging of myocardial motion.19,20 A pulsed gradient G is applied such that $0TEG(t) dt = 0 to induce a phase shift for velocity V of = kv . V, where kv = $0TE G(t) t dt. The velocity resolution of a phase-sensitive MRI experiment is approximately V = N/kvS, where N is the desired number of standard deviations of discrimination and S is the signal-tonoise ratio (SNR) of the MRI phase over the region in question, typically equal to times the SNR of the MRI magnitude.21 If the velocity changes over time, the phase shift represents the time-average velocity weighted by the integral of the gradient: E(T) = $0TG(t) dt.10 4.1 Subtraction of Head Motion Rigid motion of the brain and skull can be an important confounding factor in high-sensitivity studies of brain motion.10 Head motion is caused by cardiac pulsation, respiration, and involuntary activity of skeletal muscles seen even in cooperative subjects with good head ®xation. Echo planar MRI has demonstrated that head motions of 1 mm sÿ1 may occur even under relatively optimal conditions. Using ordinary means of head ®xation, extrinsic skull motion and intrinsic CNS motion in the supratentorial region appear to be of similar magnitudes, but may be out of phase. In the cervical region, the craniocaudal motion of the cord is several-fold greater than extrinsic motion. The effects of rigid head motion in studies of brain parenchyma motion can be conveniently removed by mathematical means.10 Using the fact that a rigid motion de®nes a linear function, velocities of the skull are measured, and brain motion due to this rigid motion estimated by linear interpolation (brain parenchymal strain rate may be corrected for rigid motion by subtracting from each component qVi/qXj times its mean value over the skull).22 When this rigid component of motion is subtracted out, the speci®c motion of the brain relative to the skull is revealed. 4.2 Limitations MRI velocity sensitivity is ultimately limited by molecular diffusion and by the spatiotemporal characteristics of the motion pattern under study. Signal attenuation due to tissue water selfdiffusion limits the intensities and durations of the gradient pulses that may be usefully applied. A gradient pulse of intensity G and duration establishes in the tissue parallel isophase planes of spatial period G diffusional attenuation becomes severe as the Einstein diffusion length (2D)1/2, where D is the diffusivity, approaches the isophase lamellar spacing.23±25 A second limit to velocity sensitivity is imposed by signal attenuation due to velocity gradients within each voxel. Velocity gradients may result from rigid rotation or material shear. In the simplest case, velocity and phase are linear functions of position.
In this case, image magnitude will be attenuated by a Fourier coef®cient of the spatial pro®le of the voxel.22 Typically, the voxel pro®le has the form of a rectangular function in the slice direction (z) and a sincfunction in each in-plane coordinate (x and y). Then the attenuation due to shear, ASH, becomes simply ASH rect
@=@x rect
@=@y sinc
@=@z
1
when the spatial coordinates (x, y, z) are expressed in units of pixel diameters, is in cycles and rect (u) = 1 if | u | < 12, 0 otherwise. Equation (1) implies that attenuation due to in-plane velocity shear is all-or-none: 0% attenuation when | q/qxi | < 1 1 2 cycle/pixel and 100% attenuation when | q/qxi | > 2 cycle/ pixel. No de®nite evidence of shear attenuation has to date been demonstrated in brain parenchyma, although such attenuation has been seen in the CSF in zones of high-shear such as the aqueduct.26 4.3
Normal Pulsatile Brain Motion
Qualitatively, the intrinsic brain motion consists of a primary systolic downstroke of the brainstem, which moves like a plunger within the relatively ®xed cerebral mantle, followed by a slower diastolic recovery toward the initial con®guration. In synchrony with this axial motion, the thalami move inward bilaterally to compress the third ventricle. Anteroposterior velocities are typically less than 25% of maximum craniocaudal or mediolateral velocities. Figure 2 shows the three components of intrinsic brain parenchyma velocity in axial slices in a normal volunteer. The slices are at the midventricular level, and cine data were acquired with 50 ms temporal resolution. Gray scale images show the magnitude and mediolateral, craniocaudal, and anteroposterior velocity component images 100 ms after the ECG Rwave. The curves plot the velocity components as functions of time for the elliptical regions of interest indicated. The net craniocaudal displacement 100 ms after the R-wave is shown as a 3D surface in Figure 3. Craniocaudal velocity in a midline sagittal plane is shown in Figure 4. The velocity curves [Figure 2(b)] show a monophasic or biphasic motion burst 50±200 ms after the ECG R-wave, synchronous with the systolic arterial pulse wave arrival, and subsequently in diastole much smaller velocities of recovery (diastolic RMS velocity < 20% systolic). The velocity images demonstrate the rapid caudal descent and of central structures, and the medial compressive motion of the thalami; anteroposterior velocities are relatively small. In normal volunteers, absolute CNS motion proves to be reproducible between heartbeats 15% (SD). It follows that conventional multishot MRI with an ECG trigger would produce reasonably accurate maps of CNS motions. 4.4
Tethered Spinal Cord
Under normal conditions, the spinal cord participates in the pulsatile movement of the CNS, with signi®cant vertical motion seen throughout cervical and upper thoracic levels. Physiological cord motion is normally rapid, with velocities of up to 10 mm sÿ1, and may exhibit several cycles of damped oscillation, possibly related to elastic recoil of the dentate ligaments.27
BRAIN PARENCHYMA MOTION OBSERVED BY MRI
Figure 2(a)
3
4 BRAIN PARENCHYMA MOTION OBSERVED BY MRI (b) Cephalic anterior right
1.5
Velocity (mm s–1)
1 0.5 0 –0.5
–1 Caudal posterior –1.5 0 left
100
200
300 400 500 600 Delay from R-wave (ms)
700
800
Figure 2 (a) Images of intrinsic brain velocity components for axial slices at the midventricular level in a normal subject 100 ms after the ECG Rwave, the approximate moment of peak parenchymal pulsatile velocities. To the left are the magnitude images; to the right are the mediolateral, craniocaudal and anteroposterior velocity component images with scales. The craniocaudal velocity map shows the rapid central descent (dark gray to black), and the mediolateral image shows the rapid bilateral medial velocities of the thalami (white±black dipole). The bottom image shows the anteroposterior velocity component. (b) The velocity±time curves for the left thalamic ROI designated by the circle. Parenchymal velocities are highest between 50 and 200 ms after the ECG R-wave during cardiac systole; much slower velocities of parenchymal recovery are seen beyond about 200 ms after the R-wave, during diastole. ~, cephalocaudal; *, anteroposterior; &, left±right
During normal development, the vertebral column undergoes greater craniocaudal growth than the spinal cord, and so the distal cord appears to ascend within the spinal canal. If the ascent of the cord is hindered by a mass, bone spur, or ®brous band, the cord may become stretched, which may lead by ischemia to symptoms of paresis, paresthesias, and loss of sphincter control. This syndrome is called the `tethered cord',
and is treated surgically. Given the normal physiological motion of the cord, it is reasonable to inquire whether changes in the normal pattern of motion could be a diagnostic marker of tethering. Investigators at The Johns Hopkins University School of Medicine and Georgetown have shown that motion of the cervical cord is a good predictor of clinical response to surgery
Figure 3 Net craniocaudal displacement of an axial slice 300 ms after the R-wave. Displacement is represented by the surface height, with a full scale of 3 mm. Image magnitude has been texture mapped onto the 3D surface. The predominant displacement is a monophasic downward movement of central structures; this ®gure corresponds to maximal displacement. Bidirectional CSF motion is seen in the horns of the lateral ventricles
BRAIN PARENCHYMA MOTION OBSERVED BY MRI
5
Figure 4 Craniocaudal velocity in a midline sagittal slice. A magnitude image is at the top, and gray-coded images of craniocaudal velocity triggered at various times after the R-wave are shown at the bottom. Velocities in general increase with proximity to the foramen magnum, this downstroke systolic and is synchronous with the known systolic egress of CSF from the calvarium. Maximal velocities in the upper cervical cord are about 3 mm sÿ1 caudally
for a tethered spinal cord.7,13 If cord motion is normal, the patient is unlikely to bene®t from surgical untethering, and an alternative etiology for neuropathic symptoms should be sought. Based on an experience of over 400 patients, MRI CNS motion studies are now standard at the Massachusetts General Hospital in cases of suspected tethering.
4.5 Normal-Pressure Hydrocephalus (NPH) NPH is a syndrome of uncertain etiology whose classic form includes dementia, ataxic gait, and urinary incontinence.28 Accurate diagnosis is imperative, as treatment can fully restore neurological function, whereas failure to treat may lead to progressive neurological dysfunction and death. At present, NPH diagnosis remains problematic, with no set of clinical or laboratory criteria clearly able to predict the individual response to CSF shunting.29,30 Prompted by recognition of the abnormal CSF dynamics in NPH,29,31 MRI studies of brain motion in NPH have been undertaken. Wahkloo et al. found brain motion in NPH patients to have greater amplitude and complexity than in age-matched controls.32 This ®nding is particularly interesting because of
the support it lends to hypotheses linking the neurophysiology of the NPH syndrome with endogenous mechanical trauma.33
5
NMR IMAGING OF PARENCHYMAL STRAIN RATES
Strain de®nes the degree of contraction and elongation at each point in a material during deformation. Velocity-sensitive NMR imaging can be used to reconstruct images of the distribution of physiological strain rates within tissues, including the myocardium and the brain parenchyma, as described here. By de®ning the pattern of tissue deformation at each location, strain imaging may shed light on both the mechanical causes and the speci®c neural effects of disorders of the mechanical homeostasis in the CNS. In an elastic material, strain is proportional to stress, which is the internal force per unit area, the constant of proportionality being the elasticity (in general, a fourth-order tensor). Strain rates are described by a strain rate tensor S whose components Sij are de®ned by the directional derivatives q/qxj of velocity components Vi according to
6 BRAIN PARENCHYMA MOTION OBSERVED BY MRI Sij 1=2
@Vi @Vj @xj @xi
2
The rate of tissue elongation or contraction in the direction of a unit vector v is simply: vT . S . v. Measurement of the complete 3D strain rate tensor S by equation (2) requires acquisition of complete 3D spatial and velocity data. As such complete acquisitions have not yet been performed in the brain, images have been produced that reconstruct tensor components based upon one or two components of velocity.20 To portray strain rate tensor ®elds usefully, novel display methods are needed. Let i and Ei denote the eigenvalues and eigenvectors of the strain rate tensor, respectively, and let f be a positive constant such that | fi | << 1 for all observed S. Then the value of S at each location can be depicted by means of a small rhomboid whose axes are given by the four vectors
1 f i Ei
3
Intuitively, this rhomboid shows the con®guration of a unit square after S has acted upon it for a time equal to f. These rhomboids show the magnitude and direction of strain rate at each location; their long axes indicate the directions of most rapid stretch. Alternatively, a color code may be used to display the values of tensor invariants of the strain rate.
5.1
Normal Brain Parenchymal Strain Rates Due to Physiological Pulsation
Normal systolic parenchymal in-plane strain rate patterns are shown in Figure 5. In-plane (x and y) components of velocity were encoded in a gated cine echo planar acquisition, and strain rates were computed according to equation (2). Data were acquired using a velocity sensitivity | kv | = 3.7 rad s mmÿ1 (gradient pulse intensity G = 0.085 mT mÿ1, duration = 35 ms, separation = 40 ms). Given an MRI magnitude SNR per pixel of 25: 1, this experiment should have the capacity to resolve velocity differences of 0.02 mm sÿ1 per pixel (P < 0.05). De®ning the `strain rate' sensitivity ks so that the phase shift per pixel equals ks . S; then ks = Rkv where R refers to the dimensions of the voxel. In the experiments described, the strain sensitivity for in-plane strains is | ks | = 3.7 rad sÿ1 mmÿ1 (for 3 mm = 11 rad s). MRI motion sensitization in effect magni®es strain rates by the ratio of the strain rate sensitivity | ks | to the motion observation period ; here, | ks | / = 275: 1. The display shows a strain rate rhomboid, constructed according to equation (3), and uses color to depict the trace, Tr (S) = 1 + 2. The trace represents the rate of compression or dilation of area within the 2D image plane. Because tissue is incompressible, in-plane area change implies compression or extension in the through-plane direction (i.e. Tr (S) + qVz/qz =
Figure 5 Images of in-plane 2D strain rates of the brain parenchyma in a normal subject in (a) coronal±oblique and (b) sagittal planes 150 ms after the ECG R-wave, corresponding to the rapid parenchymal motion of cardiac systole. A rhomboid graphic at each pixel shows the magnitude and orientation of the 2D strain rate tensor S at that location [according to equation (3) with f = 4 s]. Color indicates the rate of change of in-plane area, the trace of S, with red denoting expansion [Tr(S) > 0] and blue shrinkage [Tr(S) < 0]. The predominant in¯ow of brain parenchyma toward the foramen magnum is striking in both image planes. The coronal slice shows prominent downward tissue elongation in and near the internal capsule; the sagittal slice shows downward and forward deformation of the cerebellum and midbrain, and vertical stretching of the upper cervical cord, also seen in the coronal image. Compression of the thalami upon the region of the third ventricle is evident in the strain rhomboids in the coronal image and is also evident in the sagittal image, where red hues of in-plane expansion near the third ventricle (at center) suggest compression in the through-plane direction
BRAIN PARENCHYMA MOTION OBSERVED BY MRI
0). Color hue indicates the sign of Tr (S) at each location, with red corresponding to positive and blue to negative. Color saturation is proportional to | Tr (S) | , with full saturation at | Tr (S) | 5 0.05 rad sÿ1. Figure 5 shows coronal and sagittal images of strain rate 150 ms after the ECG R-wave. Strain rate rhomboids demonstrate the predominant elongation of parenchyma toward the foramen magnum. In the coronal image, there is seen a dramatic elongation in the region of the upper cervical cord, internal capsule, and corona radiata; in the sagittal image such elongation is seen in the cerebellum and midbrain. Maximal strain rates in these zones are on the order of 0.1 rad sÿ1. The red color in the region of the third ventricle in the sagittal image, denoting Tr (S) > 0, demonstrates the through-plane compressive motion of the thalami in this zone. Small strain rates are seen in the cerebral mantle. Cine velocity studies have shown, and strain rate cine studies have con®rmed, that motion in the territory of the basilar circulation precedes that in the middle cerebral territory cerebellar motion.9 Of interest, we see in the coronal image dilational strain patterns deep in the Sylvian ®ssures bilaterally. We can speculate that this re¯ects pulsatile motion of the middle cerebral vessels due to pulsatile turgor, the stiffening of the `vascular endoskeleton'.34 5.2 Brain Parenchymal Shear Due to Small Cranial Accelerations The human brain is an organ uniquely vulnerable to acceleration injury. Acceleration injury is associated with a variety of clinical syndromes, including acute concussion syndromes of mild to profound severity and delayed effects, as in the dementia pugilistica or `punch drunk' syndrome. An attractive unifying hypothesis is that these syndromes may be caused by acceleration-induced shear injury to deep cerebral white matter. This idea was advanced on theoretical grounds by Holburn, and soon received strong support in the experiments of Pudenz and Shelden, who, placing transparent windows into the skulls of monkeys, directly observed the swirling motion of the cerebral cortex due to cranial acceleration.35,36 Clinicopathological correlations in patients with severe concussion have demonstrated prominent damage to deep white matter tracts including the corpus callosum, cerebellar peduncles, and corticospinal tracts.37 Particularly intriguing were the ®ndings of hemorrhage at gray±white matter boundaries, suggesting elastic impedance mismatch as a cause of trauma. Similar lesion patterns have been seen in animal models of concussive nonpenetrating head injury.38 These experiments, however, have become increasingly dif®cult to justify due to concerns for animal welfare. MRI may afford the opportunity to characterize the response of the human brain to mild acceleration noninvasively, and, from this, perhaps to extrapolate the phenomenology of trauma. To investigate this, coronal images sensitive to anteroposterior velocity were acquired while a normal subject shook the head `No' twice per second for 10 s. Contributions of rigid head motion were subtracted as described above, and data were processed to demonstrate the qVz/qx component of parenchymal shear, where z is the through-plane coordinate and x is the horizontal coordinate. Figure 6 shows intrinsic parenchymal strain rates at three points in the head shake cycle. Accelera-
7
Figure 6 Brain parenchymal strain rates due to voluntary head shake. Images acquired during (a) clockwise cranial acceleration, (b) zero acceleration, and (c) counter-clockwise acceleration. The magnitude images are shown to the left and the strain rate images are to the right. The strain rate gray scale is proportional to qVz/qx, where x is the horizontal and z is the through-plane direction. Locally increased qVz/ qx, indicative of localized rotation or shearing strain, are seen adjacent to the falx, in bilateral vertical bands deep to the temporal lobes, and at the spinomedullary junction where the brainstem seems to twist slowly upon the upper cervical cord
tion-induced shear rates in this experiment are at most 0.1 rad sÿ1. This study demonstrates concentrations of strain adjacent to the falx, consistent with transmission of angular acceleration from the falx to the hemispheres. Shearing strain is seen to concentrate in parasagittal bands deep to the temporal lobes, and torsional strain to concentrate at the spinomedullary junction.
6
EFFECTS OF PARENCHYMA MOTION ON MRI OF PERFUSION AND DIFFUSION
Motion of the head and brain can interfere with studies of cerebral diffusion and cerebral function. Functional NMR studies of cerebral hemodynamic activation based on T1 in¯ow or T2* deoxyhemoglobin effects must detect changes in signal magnitude of 1%.39 These measurements, whether made with conventional or echo planar MRI, can be confounded by parenchymal in¯ow40±42 related to cardiac and respiratory motions.43,44 Cerebral diffusion measurements are exquisitely motion-sensitive; when acquired using conventional multishot MRI, these studies are easily degraded by uncontrolled cerebral motion leading to phase inconsistency among Fourier data lines. This has been addressed by cardiac gating,45 data postprocessing,46 compensation for measured motion using `navigator pulses',47 and by the adoption of single-shot acquisition techniques. Interestingly, whole-head motion has
8 BRAIN PARENCHYMA MOTION OBSERVED BY MRI overshadowed the effect of intrinsic brain motion in studies to date, suggesting the relative ineffectiveness of head ®xation. 7 PROSPECTS Motion-sensitive MRI opens a novel window onto the complex phenomena of brain parenchyma motion. This emerging picture of brain motion, however, represents only a ®rst step of a program to understand the underlying biophysics of brain motion. At issue is the desire to know intracerebral distributions of mechanical stress. At a basic level, the distributions of stress may reveal the sites of action of the cardiovascular impulse and of resistance to it. Clinically, stress concentrations may account for the speci®city of the neurological syndromes related to derangements of CNS dynamics (e.g. the relation of NPH to gait disturbance), and help predict the outcomes of intervention (e.g. how and when does fenestration of the optic nerve sheath alleviate the visual de®cit of pseudotumor cerebrii). One promising avenue of study is exempli®ed by the work of Pelc and Enzmann, who have measured concurrent intracerebral volume changes of brain parenchyma, blood, and CSF,15 an important step toward the uni®cation of brain motion with CSF dynamics in the spine and with the physiology of arterial pulse wave propagation.48±50 8 RELATED ARTICLES Assessment of Regional Blood Flow and Volume by Kinetic Analysis of Contrast-Dilution Curves; Methods and Applications of Diffusion MRI; Marker Grids for Observing Motion in MRI; Phase Contrast MRA; Time-of-Flight Method of MRA. 9 REFERENCES 1. J. E. A. O'Connell, Brain, 1943, 66, 204. 2. A. Monro, `Observations on the Structure and Function of the Nervous System', Creech and Johnson, Edinburgh, 1783. 3. G. Kellie, Trans. Med. Chir. Soc. Edinburgh, 1824, 1, 84. 4. L. B. Flexner, J. H. Clark, and L. H. Weed, Am. J. Phys., 1933, 101, 292. 5. G. H. DuBoulay, Br. J. Radiol., 1966, 39, 255. 6. D. A. Feinberg and A. S. Mark, Radiology, 1987, 163, 793. 7. D. C. McCullough, L. M. Levy, G. DiChiro, and D. L. Johnson, Pediatr. Neurosurg., 1990, 16, 3. 8. D. Greitz, R. Wirestam, A. Franck, B. Nordell, C. Thomsen, and F. Stahlberg, Neuroradiology, 1992, 34, 370. 9. D. R. Enzmann and N. J. Pelc, Radiology, 1992, 185, 653. 10. B. P. Poncelet, V. J. Wedeen, R. M. Weisskoff, and M. S. Cohen, Radiology, 1992, 185, 11. D. A. Feinberg, J. C. Hoenninger, L. E. Crooks, L. Kaufman, J. C. Watts, and M. Arakawa, Radiology, 1985, 156, 734. 12. P. Moran, JMIR, 1982, 1, 197. 13. L. M. Levy, G. Di Chiro, D. C. McCullough, A. J. Dwyer, D. L. Johnson, and S. S. L. Yang, Radiology, 1988, 169, 773. 14. R. Wirestam, D. Greitz, C. Thomsen, F. StaÊhlberg, F. Nordell, and M. Stubgaard, Proc. IXth Ann Mtg. Soc. Magn. Reson. Med., New York, 1990, p. 1107. 15. D. E. Enzmann, M. R. Ross, and N. J. Pelc, Proc. Xth Ann Mtg. Soc. Magn. Reson. Med., San Francisco, 1991, p. 155.
16. S. E. Maier, C. J. Hardy, and F. A. Jolesz, Proc. XIIth Ann Mtg. Soc. Magn. Reson. Med., New York, 1993, p. 151. 17. J. Henning, D. Ott, T. Adam, and H. Friedburg, JMIR, 1990, 8, 543. 18. E. Lee, J.-Z. Wang, and R. Mezrich, Am. J. Neuroradiol, 1989, 10, 1145. 19. N. J. Pelc, R. J. Herfkins, and L. R. Pelc, JMIR, 1991, 1, 181. 20. V. J. Wedeen, Magn. Reson. Med., 1992, 27, 52. 21. N. J. Pelc and M. A. Bernstein, Proc. IXth Ann Mtg. Soc. Magn. Reson. Med., New York, 1990, p. 475. 22. V. J. Wedeen, R. M. Weisskoff, and B. P. Poncelet, Magn. Reson. Med., 1994, July. 23. V. J. Wedeen, B. R. Rosen, and T. J. Brady, `The Magnetic Resonance Annual', Raven Press, New York, 1987, p. 174. 24. P. T. Callaghan and C. D. Eccles, JMIR, 1988, 71, 1. 25. D. G. Cory and A. N. Garroway, Magn. Reson. Med., 1990, 14, 435. 26. W. G. Bradley, Jr, A. R. Whittermore, K. E. Kortman, A. S. Watanabe, M. Homyak, L. Teresi, and S. J. Davis, Radiology, 1991, 178, 459. 27. D. J. Mikulis, M. L. Wood, B. Poncelet, O. A. M. Zerdoner, and D. E. Bohning, Proc. XIth Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 626. 28. R. D. Adams, C. M. Fisher, S. Hakim, R. G. Ojemann, and W. H. Sweet, N. Engl. J. Med., 1965, 273, 117. 29. R. P. Friedland, J. Am. Med. Assoc., 1989, 262, 2577. 30. N. R. Graff-Radford, J. C. Godersky, and M. P. Jones, Neurology, 1989, 39, 1601. 31. H. D. Portnoy, M. Chopp, C. Branch, and M. B. Shannon, J. Neurosurg., 1982, 56, 666. 32. A. K. Wahkloo, F. Jungling, and J. Hennig, Proc. XIth Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 627. 33. C. M. Fisher, Neurology, 1982, 32, 1358. 34. K. R. Davis, Personal communication, 1992. 35. A. H. S. Holburn, Lancet, 1943, ii, 438. 36. R. H. Pudenz and C. H. Shelden, J. Neurosurg., 1946, 3, 487. 37. S. J. Strich, Lancet, 1961, 443. 38. T. A. Genarelli et al., Acta Neuropath. Suppl., 1981, 7, 23. 39. K. K. Kwong, J. W. Belliveau, D. A. Chesler, I. E. Goldberg, R. M. Weisskoff, B. P. Poncelet, D. N. Kennedy, B. E. Hoppel, M. S. Cohen, and R. Turner, Proc. Natl. Natl. Acad. Sci., USA, 1992, 89, 5675. 40. J. R. Singer, Science, 1959, 130, 1652. 41. L. Axel, Am. J. Roentgenol., 1984, 143, 1157. 42. F. W. Wehrli, J. R. MacFall, L. Axel, D. Shutts, G. H. Glover, and R. J. Herfkins, Noninvasive Med. Imaging, 1984, 1, 127. 43. P. Jezzard, D. Le Bihan, C. Cuenod, L. Pannier, A. Prinster, and R. Turner, Proc. XIIth Ann Mtg. Soc. Magn. Reson. Med., New York, 1993, p. 1392. 44. R. M. Weisskoff, J. Baker, J. Beleiveau, T. L. Davis, K. K. Kwong, M. S. Cohen, and B. R. Rosen, Proc. XIIth Ann Mtg. Soc. Magn. Reson. Med., New York, 1993, p. 7. 45. D. Chien, R. B. Buxton, K. K. Kwong, T. J. Brady, and B. R. Rosen, J. Comput. Assist. Tomogr., 1990, 14, 514. 46. J. J. M. Cuppen, J. P. Groen, J. J. E. In den Klef, and H. A. Tuithof, Proc. IVth Ann Mtg. Soc. Magn. Reson. Med., London, 1985, p. 962. 47. W. S. Kim, C. W. Mun, D. J. Kim, and Z. H. Cho, Magn. Reson. Med., 1990, 13, 25. 48. M. Chopp and H. D. Portonoy, J. Neurosurg., 1980, 53, 516. 49. B. Berman and G. Agarwal, Surg. Neurol., 1984, 22, 83. 50. M. Kosteljanetz, Acta Neurolog. Scand., 1987, 75, 5.
Biographical Sketches Van J. Wedeen. b B.A. (Mathematics), Harvard College, USA, 1977. M.D., Albert Einstein College of Medicine, USA, 1981. Certi®cation
BRAIN PARENCHYMA MOTION OBSERVED BY MRI in Nuclear Medicine, 1993. Introduced to NMR by T. J. Brady, B. R. Rosen, and I. Pykett. Department of Radiology, Massachusetts General Hospital, 1983±present. Approx. 30 publications. Research specialties: the theory and application of MRI to the study of motion and ¯ow, cardiovascular MRI, mathematical issues in neurocognitive functional MRI.
9
Brigitte P. Poncelet. b B.S. (Physics), Free University of Brussels, Belgium. Introduced to NMR by C. Segebarth and P. van Dijk. Ph.D. candidate, Free University of Brussels, 1989±present. Research fellow in Radiology, Massachusetts General Hospital, USA, 1990±present. Approx. 10 publications. Research specialties: the MRI study of motion and ¯ow, functional MRI.
Diffusion and Perfusion in MRI
resulted in the recent and fascinating incursion of MRI into the field of functional brain imaging. With MRI it is now possible to monitor completely noninvasively and with unmatched combined spatial and temporal resolution the activity of the brain during mental processes and the execution of cognitive tasks.
Denis Le Bihan Warren G. Magnuson Clinical Center, Bethesda, MD, USA
2 DIFFUSION MRI
1 2 3 4 5
Introduction Diffusion MRI Perfusion MRI Related Articles References
1
INTRODUCTION
1 1 6 11 11
The effects of diffusion on the NMR signal have been deeply analyzed and powerful diffusion measurement techniques using NMR have been extensively used in physics and chemistry. The recent application of such principles, coupled to NMR imaging (MRI) in the field of medical sciences, represents a spectacular and somewhat unpredicted development. The measurement of molecular displacements in biological tissues in vivo is of enormous interest in terms of its potential applications—from the determination of the molecular organization in tissues, to the emergency management of stroke patients, to the monitoring of laser surgery. MRI is the only means available today of approaching the molecular diffusion process in vivo noninvasively. In this article the basic principles of diffusion measurements made using NMR are presented, together with a description of the various methods that have been proposed for producing images of diffusion. Some technical considerations are discussed, followed by a description of the effects on diffusion measurements of the microdynamics and microstructure of biological tissues. Emphasis will be given to the difficulties encountered when implementing diffusion imaging, and the significance of diffusion measurements in biological tissues. It has been proposed that, because of its relative safety, its high spatial and temporal resolution, its ability to acquire images in any orientation, and its greater availability compared to other modalities such as positron emission tomography (PET), MRI can be used to image perfusion. Two approaches have been developed that allow MRI to be used to image quantities related to perfusion. One approach mimics conventional nuclear medicine methods, but uses radioactively inert external tracers. The other approaches are more original and MRI specific, using blood directly as an endogenous natural tracer. Such methods offer new insights, for instance valuable information can be obtained by monitoring variations in blood oxygenation in human brain cortex during activation tasks. This article covers the physical principles and fields of applications of these techniques, both tracer and blood pool agent based approaches and methods specific to NMR where flowing blood is magnetically labeled. Finally, a large section of this article is devoted to the ‘blood oxygen level dependent’ (BOLD) contrast approach which has
2.1 Diffusion and NMR
The effect of diffusion on the NMR signal can be understood from the bipolar pulsed gradient spin echo experiment (echo time TE ) suggested by Stejskal and Tanner1 (Figure 1). The first gradient pulse induces a phase shift ϕ 1 of the spin transverse magnetization, which depends on the spin position z 1 : ϕ1 = γ Gδz1
(1)
where γ is the gyromagnetic ratio, G is the gradient strength along the z axis, δ is the gradient duration. The time interval between the pulse onsets is denoted by . After the 180◦ rf pulse, ϕ 1 is inverted to −ϕ 1 . The second pulse will produce a phase shift ϕ 2 : ϕ2 = γ Gδz2
(2)
where z 2 is the spin position during the second pulse. The net transverse magnetization is: Mxy /M0 = exp[i(ϕ2 − ϕ1 )] = exp[iγ Gδ(z2 − z1 )]
(3)
where all relaxation effects have been included in M 0 . Obviously, for ‘static’ spins z 1 = z 2 , so the bipolar gradient pair is easily evaluated. For ‘moving’ spins, however, there is a net dephasing which will depend on the spin history during the time interval between the pulses. Indeed, we must now consider a population of spins, which may have different motion histories. If P (z 2 , z 1 , ) dz 2 is the conditional 90˚
180˚
Echo
G
G
d
d ∆
Figure 1 Stejskal and Tanner diffusion spin echo sequence. Gradients pulses G are separated by a time interval ; their duration is δ
2 DIFFUSION AND PERFUSION IN MRI probability of finding a spin initially at z 1 between positions z 2 and z 2 + dz 2 after a time interval , the amplitude attenuation is: Mxy /M0 = exp[iγ Gδ(z2 − z1 )]P (z2 , z1 , ) dz2 (4)
most widely used sequence in NMR diffusion measurements. Taking into account spin diffusion during δ, equation (6) becomes:
In the case of the diffusion process, one has:
The diffusion time T d can be formally defined as ( − δ/3), but its physical significance is clear only when the gradient pulse duration δ is sufficiently short compared with . Tissues with short T 2 values require short TE s which may not allow sufficiently long diffusion times to produce a large enough diffusion effect.
P (z2 , z1 , ) = (4πD)−1/2 exp[−(z1 − z2 )2 /4D]
(5)
where D is the diffusion coefficient. Combining equations (4) and (5), the signal attenuation is: Mxy /M0 = exp[−(γ Gδ)2 D]
(6)
In practice, as δ is usually not negligible compared with , movement during the application of the gradient pulses can no longer be neglected (see below). Also, the contribution of other gradient pulses, such as the imaging pulses, must be taken into account.2 Solving the Bloch–Torrey equation,3 one obtains: Mxy /M0 = exp
− γ2
TE 0
t
G(t ) dt
2
dt D
0
≡ exp(−b · D)
(7)
This relationship is valid for any gradient pulse combination, provided the sign of G is inverted for all gradient pulses following the 180◦ pulse. This expression may become analytically complex when many gradient pulses are involved, as in MRI. Therefore, it is useful to characterize the sensitivity of MRI sequences to diffusion using the ‘gradient factor’ b.2,4 One must remember, however, that this equation is valid only for diffusion in an infinite and homogeneous medium. If diffusion is restricted or anisotropic, the signal attenuation must be calculated using a more general formalism based on diffusion tensors.5 2.2
M/M0 = exp[−γ 2 G2 Dδ 2 ( − δ/3)]
(9)
2.2.2 Stimulated Echo Technique
A stimulated echo is generated from a sequence consisting of three rf pulses separated by time intervals τ 1 and τ 2 . Gradient pulses must be inserted within the first and the third periods of the stimulated echo sequence (Figure 2). The diffusion time now includes τ 2 and can be much longer than with the spin echo sequence, because only longitudinal relaxation occurs during this time interval. The Stejskal–Tanner relationship [equation (9)] still applies, provided that the period τ 2 is included in .8 The longer diffusion time is useful for studying very slow diffusion rates, or in compensating for the unavailability of large gradients.9 Unfortunately the signal is reduced by one-half compared with the spin echo signal. 2.2.3 Gradient Echo Technique
The effect of diffusion on the amplitude of a gradient echo formed by a bipolar gradient pulse pair of reversed polarity does not differ from that of a spin echo sequence.2 However, if the gradient echo is part of a steady state free precession (SSFP) sequence, where some degree of phase coherence is propagated throughout successive cycles, multiple echo paths with different diffusion times and different diffusion weightings must be considered.10 In MRI, the
The Different Approaches
Almost any NMR sequence can be designed to measure diffusion.
90˚
90˚
Stimulated echo
90˚
t1
t2
t1
2.2.1 Constant Field Gradient Spin Echo Method
G
G
In the presence of a simple constant linear gradient G 0 , the echo attenuation is:6,7
d
d
M(nT E)/M0 = exp(−γ 2 G20 DT E 3 /12n)
(8)
where n is the number of echoes in a multiple echo experiment. This simple approach is difficult to combine with NMR imaging because the presence of a constant gradient during the rf pulses of the imaging sequence severely impairs the selected slice profile. In addition, the relaxation time T 2 of the medium should not be too short, so that enough diffusion attenuation occurs during TE . 2.2.1.1 Bipolar Gradient Pulse Spin Echo Technique. This technique, suggested by Stejskal and Tanner,1 is by far the
∆
Figure 2 Diffusion-weighted stimulated echo sequence. The stimulated echo results from the spin excitation by three rf pulses separated by time intervals τ 1 and τ 2 . During τ 2 , relaxation is driven by T 1 and not T 2 . Diffusion effects can be enhanced by placing gradient pulses within the τ 1 periods, where transverse magnetization is sensitive to field inhomogeneities
DIFFUSION AND PERFUSION IN MRI
contrast-enhanced (CE-fast) scheme has been proposed.11 – 13 Unfortunately, this sequence remains very sensitive to motion artifacts.14 Moreover, the effects of diffusion and relaxation are intermingled so that diffusion measurements are always contaminated to some degree by relaxation effects.15
Diffusion measurements can also be achieved by means of the rf field (B1 ) produced by an NMR rf coil oriented perpendicularly to the main transmit/receive NMR coil.16,17 With rf gradients, extremely short switching times can be achieved, since there are no eddy currents. Furthermore, substantial gradient strength may be produced from the rf transmitter, allowing the measurement of very low diffusion coefficients.
2.3
0 –0.1 –0.2 log (S/S0)
2.2.4 Diffusion Measurements with B1 Field Gradients
3
–0.3 –0.4 –0.5 –0.6 –0.7 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 b (×103 s mm–2)
Figure 3 Plot of the logarithm of the signal intensity in brain white matter against the gradient factor b. The plot is linear, and the slope is equal to the diffusion coefficient (D = 0.60 ± 0.01 × 10−3 mm2 s−1 )
Combining Diffusion NMR with MRI 90˚
‘Diffusion-weighted’ images are obtained by inserting gradient pulses into any NMR imaging sequence. For quantification it is necessary to determine the degree of diffusion weighting, which is achieved by calculating the factor b [equation (7)] associated with the acquisition parameters. Without additional gradients, typical spin echo imaging sequences have intrinsically very low b values, typically less than 1 s mm −2 , so that diffusion effects are completely negligible (for water, D = 2 × 10−3 mm2 s−1 at room temperature and the attenuation is less than 1%).2,18 When inserting gradient pulses for diffusion, however, the combination of the imaging and diffusion gradients produce cross terms in the b factors.2,19 These cross terms will depend on the time course of these gradient pulses and may be significant, so that the Stejskal–Tanner relation [equation (9)] is incorrect in most cases. Using the b factor values associated with each image, it is possible to compute diffusion images, i.e. images where the diffusion coefficient is determined for each pixel according to equation (7).2,4,9,18 The computation of such diffusion images is obtained by fitting the signal intensity of each pixel of a series of images differently sensitized to diffusion (i.e. with different b values) with equation (7) (Figure 3). Using the spin echo sequence, it is possible to vary the strength or the duration of the diffusion-sensitizing gradients, or their direction, in order to enhance anisotropic diffusion effects.2 The spin echo two-dimensional FT method is certainly the simplest to implement.4,18,20 Other sequences2 that have been investigated are stimulated echo sequences; line-integral projection reconstruction; variants of the SSFP technique; ‘turbo’ sequences; and echo planar imaging.21,22 These schemes have been suggested to overcome, at least partly, some of the problems encountered with the spin echo two-dimensional FT method. With echo planar imaging (EPI), the entire set of echoes needed to form an image is collected within a single acquisition period (single shot) of 25–100 ms. This is obtained by switching the echo signal formation in a train of gradient echoes, by means of a large gradient the polarity of which is very rapidly inverted as many times as is required to achieve the desired image resolution.23 EPI may easily be sensitized to diffusion21,22 (Figure 4).
180˚
Echoes rf Slice select Readout Phase encode
Figure 4 Diffusion echo planar imaging sequence. The spin echo is split into a series of gradient echoes by quickly switching the readout gradient. Sensitization to diffusion of an echo planar imaging sequence can be easily achieved by additional gradient pulses (hatched boxes)
Sensitization consists of providing a pair of large compensated gradients for a period of time before rapid gradient switching and data acquisition. The refocusing may be achieved either by simply reversing the polarity of the gradient halfway through the period over which it is applied, or by inserting a 180◦ rf refocusing pulse at the midpoint without reversing the gradient polarity. With EPI, motion artifacts are virtually eliminated. The accuracy on diffusion achieved with EPI is generally extremely good, as many images differently sensitized to diffusion can be generated due to the very short acquisition time (typically less than 100 ms). The EPI technique is the method of choice for in vivo diffusion imaging, although it is very vulnerable to susceptibility artifacts, which are responsible for image distortion or signal dropout, and to chemical shift artifacts which require efficient fat suppression. Diffusion-EPI has been successfully used to measure the diffusion of water in the human brain in volunteers and patients21 (Figure 5).
2.4 Experimental Considerations in Diffusion MRI 2.4.1 Gradient Requirements
Any mismatch between the diffusion sensitizing gradient pulses may cause artifactual signal losses, due to improper spin
4 DIFFUSION AND PERFUSION IN MRI
Figure 5 Diffusion images calculated from a set of 16 diffusion-weighted coronal echo planar images of the brain of a normal volunteer. These are 64 × 64 pixel images with a field of view of 16 cm and thus an in-plane resolution of 2.5 × 2.5 mm. The section thickness was 10 mm. The diffusion gradients varied from 0 to 38 mT m−1 in the z direction, the duration of each lobe was 20 ms, and their separation was 6 ms (b = 0 to 872 s mm−2 ). The acquisition time for each image was about 50 ms. Note that diffusion is highest in ventricular cavities containing cerebrospinal fluid (free water). Because of diffusion anisotropy, diffusion in corpus callosum is very low, while vertical frontal white matter tracts have higher diffusion (see Table 1)
rephasing. Two major sources of problems are commonly seen. One is gradient instability, which may arise when gradient amplifiers are driven hard for fast switching of large gradient intensities. Variations from shot to shot of the bipolar gradient balance results in widely distributed ghost artifacts. While such an artifact does not occur when a single-shot technique such as EPI is used, it is still essential to have high quality gradient amplifiers with some reserve capacity for accurate diffusion imaging. The second problem results from eddy currents generated mainly in the cryostat when switching rapidly large gradient pulses. Eddy currents may also be a major cause of gradient pulse mismatch, leading to signal loss. The best solution to this problem is undoubtedly to remove eddy currents at source by using actively shielded gradient coils, which have no fringe fields and therefore do not generate eddy currents. The use of a gradient coil of small dimensions, which remains at a fair distance from the magnet core and with which it is easier to generate large gradient amplitudes, is certainly an attractive alternative.21 2.4.2 Motion Artifacts
As the sequences used are deliberately sensitized to motion by the addition of large gradients, a major problem occurring with in vivo imaging of diffusion arises from irregular motion of the object. Artifacts result from discontinuities that occur between successive cycles of an imaging sequence. Results of such temporal incoherence are commonly visible as ‘ghosts’ along the phase-encoding direction. These ‘ghosts’ are particularly intense in the presence of the diffusion gradients and render the diffusion measurements meaningless. Cardiac gating has been used to mitigate this problem, but even this motion is not strictly cyclic. It is difficult to compensate diffusion imaging sequences for motion, since the use of successive bipolar gradient pulses considerably reduces the
value of the gradient factor b.2 Chenevert et al.24 have suggested eliminating any phase encoding from the acquisition by sacrificing one dimension of the image, which is then reduced to a single line. Ultimately, however, the best way to avoid motion artifacts is to use a single-shot technique, such as EPI, and to secure patients comfortably within the magnet using cushions or inflating devices. 2.5 Diffusion in Biological Systems
The diffusion coefficient of water in tissues has been found to be 2–10 times less than that of pure water (Table 1). This can be largely understood considering that water molecules are obliged to divert tortuously around obstructions presented by fibers, intracellular organelles, or macromolecules. Biological systems thus differ greatly from an ‘infinitely large medium’; they are very heterogeneous and made of multiple subcompartments (microstructure). Depending on the permeability of the barriers that limit these compartments, we have to consider exchanges and transport Table 1 Diffusion Coefficients of Water in Human Braina ×103 mm2 s−1 Cerebrospinal fluid Gray matter White matter Corpus callosum Axial fibers Transverse fibers
2.94 ± 0.05 0.76 ± 0.03 0.22 ± 0.22 1.07 ± 0.06 0.64 ± 0.05
a Measurements obtained in normal volunteers using a diffusion sensitized EPI sequence.21 Direction of fibers refers to z axis (direction of the diffusion sensitizing gradient pulses).
DIFFUSION AND PERFUSION IN MRI
5
between them (microdynamics). A classical treatment of the NMR signal may not then reflect properly the tissue structure or properties. In these conditions diffusion coefficients may become meaningless if the measurement timescale or the measurement direction are not provided. The main difficulty is that the structure of the medium is generally unknown in detail. 2.5.1 Restricted Diffusion
Diffusion is restricted when boundaries in the medium prevent molecules from moving freely.2,25,26 When measurement times are very short, most molecules do not have enough time to reach boundaries, and so they behave as if diffusing freely. Once the diffusion time increases, an increasing fraction of molecules will strike the boundaries, and diffusion will deviate from the free behavior. The diffusion distance, calculated as for free diffusion using Einstein’s relation ζ 2 = 2 DT d , shows a curvature and, finally, a levelling-off when the diffusion distance reaches the size of the restricting compartment. The effects of restriction, which depend on the shape of the restricting volumes and the types of restriction,2 will thus appear in the NMR signal for diffusion times such that molecular displacements are of the order of the size of the restricting volumes. When the restrictive barriers become permeable to diffusing molecules, the restricted diffusion pattern changes. An example has been given by Tanner27 for the case of equally spaced, plane, and permeable barriers. When the diffusion time is increased, the apparent diffusion coefficient decreases, as expected for restricted diffusion, but saturates at a level which depends on the permeability constant. 2.5.2 Anisotropic Diffusion
Diffusion in tissues may also be anisotropic if hindrance or restriction is not the same for different directions of motion. The measured diffusion coefficient then varies according to the direction of measurement. Examples include muscle28 and brain white matter (see Figure 6).29 To consider anisotropy, the diffusion ‘coefficient’ must be replaced by a tensor, D.1,5 The diagonal terms of the diffusion tensor D xx , D yy , and D zz represent molecular mobility along the axes x , y, and z of the laboratory frame of reference, while the off-diagonal terms indicate how displacements are coupled in directions which are mutually perpendicular. The echo attenuation must be rewritten as: bij Dij (10) M/M0 = exp − i=x,y,z j =x,y,z
where b ij are components of a matrix b and D ij are components of the diffusion tensor.5 Off-diagonal terms in the diffusion tensor may be important when the gradient directions do not coincide with the orthotropic directions of the tissue. The echo attenuation then depends on a combination of the diagonal and nondiagonal terms of the diffusion tensor. The three orthotropic axes of the tissue coincide with the eigenvectors of D. The diffusivities along these principal directions are the eigenvalues of D. The main direction of the tissue, e.g. the fiber-tract direction for muscle of white
Figure 6 Diffusion image showing anisotropic diffusion in white matter. Excellent contrast is achieved, although T 1 and T 2 effects have been removed. Corpus callosum and temporal white matter fibers, which are horizontal, are dark. Vertical corona radiata frontal fibers and internal capsule fibers are bright. In brainstem, vertical fast-conducting motor and somatosensory tracts are also bright
matter, is given by the eigenvector with the largest principal diffusivity. Also associated with D are scalar invariants, such as Tr(D), which are independent of the orientation of the sample with respect to the laboratory frame of reference; they depend only on the tissue microstructure.5 In white matter, the results of the measurements depend on the respective orientation of the myelin fiber tracts and the gradient direction at each different image location. Diffusion coefficients are significantly lower when the myelin fiber tracts are perpendicular to the direction of the magnetic field gradient used to measure molecular displacements (Figure 6). Considering that most NMR visible water is in the axons, a simple model would be to consider that water molecules are enclosed in the axonal spaces and that water diffusion outside the axons is prevented by the myelin sheath. Recent diffusion experiments have not been able to show complete restricted diffusion, since the diffusion distance increases linearly with the square root of the diffusion time, as for free diffusion30 (Figure 7). The reduced value of the diffusion coefficient across myelin fibers can thus only reflect a decreased water mobility through the successive lipid layers. 2.5.3 Diffusion in Multiple Compartment Systems
Most diffusion measurements in biological tissues refer to an ‘apparent’ diffusion coefficient (ADC), and it is generally considered that diffusion in the measurement volume (the ‘voxel’, in imaging) has a unique diffusion coefficient: Da =
ρi Di
(11)
i
This simplification may not always be legitimate, since most tissues are made of multiple subcompartments—at the least, intracellular and extracellular compartments. Assuming measurement times are short, so that diffusion is unrestricted
6 DIFFUSION AND PERFUSION IN MRI 37
12 10 8 6 Gray matter White matter, x White matter, z
4 2 0 0
2
4
6
8
10
Td (ms1/2)
Figure 7 Plot of the diffusion distance against diffusion time in gray and white matter. The diffusion distance was obtained from the diffusion coefficient using Einstein’s relationship. There is no leveling-off of the diffusion distance, suggesting that water diffusion is not completely restricted
Temperature (MRI) (˚C)
Diffusion distance (m m)
14
32
T1 Diffusion
27
22 22
in each subcompartment i , and there is no exchange, the signal attenuation is: ρi exp(−bDi ) (12) M/M0 =
27 32 Temperature (Luxtron) (˚C)
37
Figure 8 Comparison of temperature measurements using T 1 and diffusion MRI, and using probes (Luxtron) in a phantom. The accuracy was found to be 0.2◦ C with diffusion and 0.5◦ C with T 1
i
where ρ i is the density of molecules diffusing in compartment i , and D i is the associated diffusion coefficient. In this case, the ‘apparent’ diffusion coefficient that would be measured would depend on the range used for b and would not reflect properly the diffusion in the voxel. Measurements with low b values would then be more sensitive to fast diffusion components. The ideal approach would be to separate all subcompartments by fitting the data with a multiexponential decay. Unfortunately, the values for D i are often low and not very different from each other, so that very large b values and very high signal-to-noise ratios would be required.
2.6
Artery
Microcirculation exchanges
Vein
Blood flow
Conclusion
Diffusion thus appears to be a powerful source of contrast for tissue characterization or functional studies. It has been shown that there is no correlation between the diffusion coefficient and the relaxation times T 1 and T 2 . T 1 and T 2 may be normal or elevated in diseased states, while diffusion is lowered, as shown in early brain ischemia.31 – 33 Other applications include temperature imaging, a technique based on the sensitivity of diffusion coefficients to temperature34 (Figure 8). Temperature imaging is especially useful for monitoring interventional procedures.
3
Capillaries
PERFUSION MRI
Classical techniques used to measure and image perfusion can be split into three categories: methods based on diffusible tracers (Figure 1); methods based on intravascular tracers (blood pool tracers); and deposition techniques (microspheres) (Figure 9). Conventional perfusion measurement techniques based on radionuclide tracers have led to the identification
Tissue (mass M) Microspheres Diffusible tracers Pure intravascular tracers
Figure 9 Principles of conventional perfusion measurements. Microspheres are particulates which, because of their size, are trapped at the entry of the capillary network; a count of their deposition pattern gives the flow rate. Diffusible tracers cross the capillary wall; their concentration in the tissue reflects blood flow. Intravascular tracers remain inside the capillary compartment and may be used to evaluate blood flow (under some assumptions)
of ‘perfusion’ with ‘blood flow’ in the tissue. However, perfusion has also often been used simply to denote the degree (density) of normal or abnormal microvasculature in a tissue as seen, for instance, with conventional catheter angiography.
DIFFUSION AND PERFUSION IN MRI
Diffusible Tracers
0.2
Blood flow measurements using diffusible tracers were first introduced by Kety and Schmidt.35 Blood flow can, in principle, be determined from two parameters: the arterial input C a (t), which is assumed to be uniform across all voxels, and the tissue tracer concentration, measured by means of the external imaging detector. The tracer tissue concentration at time T , C 0 (T ), is related to the normalized arterial flow F (in mL min−1 per 100 g of tissue): C0 (T ) = (F /λ) exp[−(F /λ)T ]
pCO2 = 48.5 T CBF = 72 ml min–1 per 100 g
0.1 pCO2 = 23.5 T CBF = 29 ml min–1 per 100 g
T
Ca (t) exp[(F /λ)T ] dt 0
(13) where λ is the blood/tissue partition coefficient of the tracer. With NMR, the tracers are nonproton nuclei, such as 19 F compounds, D2 O or H2 17 O; so far these have been used in animal studies. Tissue concentration is determined from integrated spectral line areas with spectroscopy techniques or from pixel signal intensities with MRI methods after tracer has been administered, and the arterial input function is derived from blood samples or, in some cases, from monitoring the NMR signal in a peripheral artery (either on-line or off-line). In general, one has to deal with multiple compartment fitting, e.g. gray/white matter in the brain, signal contamination from surrounding muscles, and tracer recirculation. 19 F has a sensitivity and a resonant frequency which are close to those of the proton, and is thus easily observed by NMR. Most studies have used inhalation of inert Freon gases, such as FC22 (CHClF2 )36 or FC23,37 – 39 which appear to be less toxic and induce fewer changes in brain physiology. Deuterium is used as a tracer injected intraarterially in the form of deuterated water (D2 O), which quickly exchanges protons with the surrounding water to become HOD.40 HOD behaves very similarly to normal water, and the potential toxicity is limited at low concentration. The low sensitivity of the nucleus and its short T 2 result in low signal-to-noise ratios. 17 O is a stable isotope of natural abundance of 0.0037%, is chemically equivalent to 16 O and has a very low toxicity, even at somewhat high concentrations. The current limitation to its use is its low availability and high cost. While the 17 O NMR signal has been observed directly,41 other workers have based their approach on the scalar coupling of 17 O which relaxes water protons efficiently.42,43 Much higher signal-tonoise ratios may thus be reached by observing 17 O water through proton–hydrogen MRI (Figure 10). 17 O may also be administered as H2 17 O and used similarly to H2 15 O with PET to evaluate perfusion, or as a gas to evaluate tissue oxygen extraction.41 3.2
Relative concentration (AR2)
3.1
7
0.0 0
2
4
6
Minutes
Figure 10 Perfusion measurements based on relaxivity effects of [17 O]water. Cerebral blood flow (CBF) was found to increase with pCO2 , as expected (from ref. 43, with permission)
amplification effect over the simple relaxivity effect. These gradients are responsible for spin dephasing which leads to significant signal loss, especially with long echo time gradient echo sequences and high field strengths (Figure 11). At ‘clinical’ concentrations, i.e. 0.1 mM kg−1 , the relationship between the tissue agent concentration and the NMR signal loss is experimentally linear:44 R2∗ = (1/T2∗ ) = k[Gd]
(14)
The constant k has been empirically determined in the brain, but potential variations between tissues, and especially abnormal tissues, remains an important issue.
‘Blood Pool’ Agents
If a nondiffusible NMR paramagnetic agent, such as gadolinium diethylenetriaminepentaacetic acid (Gd-DTPA), is injected into the blood stream in the form of a bolus, the difference in magnetic susceptibility between the tissue and the blood where the contrast agent is compartmentalized results in internal magnetic field gradients, which extend far beyond the capillary limits and, therefore, represent a significant
Figure 11 Set of 16 echo planar images acquired after injection of a bolus of Gd-DTPA. Images were obtained at 1-s intervals. The transit of the contrast agent bolus through the brain capillaries produces a signal drop which lasts a few seconds. This drop is higher in gray matter than in white matter. The limited signal drop in the low grade astrocytoma suggests that the tumor has a low capillary density (blood volume)
8 DIFFUSION AND PERFUSION IN MRI Capillaries
0.2 +
log (Signal)
0
++ ++ +
+ ++ ++ +
+++
++ + +
++ ++
++ + ++ + ++ ++
+ + ++ ++ + ++ + ++ + + + + ++ ++ + ++ + +++ ++
Artery
Vein
–0.2
v
–0.4 +
–0.6
Gray matter SR2 = 150 Tumor 1 SR2 = 121 White matter SR2 = 44 Tumor 2 SR2 = 456
L
–0.8 60
Arterial input
70
Figure 12 Typical magnetic resonance signal time course study after injection of a bolus of Gd-DTPA in a patient with a high-grade glioma. Each curve corresponds to a given region of interest. The necrotic center of the tumor is completely avascularized. The tumor wall has a very large blood volume. Note that, due to the breakdown of the blood–brain barrier in the tumor, the contrast agent starts leaking inside the tumor, where T 1 effects (which counteract the susceptibility-related signal drop) may lead to underestimation of the blood volume SR2, area under first-pass curve [proportional to cerebral blood volume (CBV)].
As with other blood pool tracer approaches,45 information on perfusion (i.e. blood flow, blood volume, or mean transit time) may, in theory, be derived from the tissue concentration–time curve. In practice, the local blood volume V derived from the magnetic resonance signal time course is proportional to the area (integral) under the tissue concentration–time curve: (15) V ∝ Fa /Q · S(t) dt
Tissue residue, t = L/v
CT (t)
30 40 50 Time after injection (s)
Time
Time
Actual bolus
Actual tissue concentration
CT (t)
20
CA(t)
10
CA(t)
0
Time
Time
Figure 13 Diagram showing that the actual tissue concentration time course is the product of convolution of the tissue residue function and the arterial input function. (Adapted from Axel45 )
where Q is the total amount of tracer injected as a tight bolus, and F a is the arterial blood flow. Only the relative blood volume is determined but, assuming the same blood comes from a common feeding artery, comparison can be made between different parts of the brain (Figure 12). Several difficulties have to be overcome when integrating the tissue concentration over time, such as the recirculation of the tracer.46 As for the blood flow, in principle the recirculation of the tracer could be determined using the Stewart–Hamilton theorem: Fa = V /MTT
(16)
where MTT is the mean transit time, i.e. the average time required by the tracer to pass through the tissue.47 Ideally, the tracer should be introduced into the vascular compartment as an impulse bolus. In practice, the mean transit time could be determined by deconvolving the tissue concentration–time curve by the arterial input function obtained by monitoring the contrast agent concentration in a feeding artery using MRI (Figure 13). Without the knowledge of the arterial input function, the area under the tissue tracer concentration–time course only gives the regional blood volume, and not blood flow.48 This technique has been used clinically to produce semiquantitative maps of the regional cerebral volume, for instance in brain tumors (Figure 14). Under some circumstances, however, such as brain activation studies, blood volume evaluation may provide useful information on hemodynamic changes.49
Figure 14 Blood volume map calculated from a set of echo planar images obtained during the transit of a bolus of Gd-DTPA. The tumor shows a rim with very high blood volume, while the necrotic center remains black. The blood volume is 2–4 times larger in gray matter than in white matter
3.3 Magnetic Labeling
3.3.1 Methods Based on Phase Shifts
Spins moving in the presence of a magnetic field gradient get out of phase. Depending on the capillary geometry and circulation conditions on the one hand, and the MRI acquisition parameters on the other, perfusion can be viewed
DIFFUSION AND PERFUSION IN MRI
as a predominantly incoherent motion (intravoxel incoherent motion, IVIM)50 or a predominantly coherent motion (intravoxel coherent motion, IVCM).51 3.3.1.1 IVIM. When the capillary bed is sufficiently tortuous, phase shifts of flowing spins are distributed within each voxel and result in signal attenuation. A simple approach considers microcirculation as a pseudodiffusion process.50 The signal attenuation A depends then on a ‘pseudodiffusion coefficient’ D*: A ∝ exp(−bD ∗ )
(17)
D* = lv /6, where l is the mean capillary segment length, and v is the average blood velocity; b is the gradient factor defined in equation (7). Pseudodiffusion has, in practice, to be separated from bulk water diffusion, which requires a high signal-to-noise ratio52 and hardware stability, given the very small blood volume fraction f in the brain (2–4%). The use of EPI allows one to obtain a large number of images in a short time and significantly reduces the occurrence of motion artifacts21 (Figure 15). The IVIM parameters, f and D*, may be related to actual perfusion in terms of blood flow F under some assumptions about capillary network organization:53 ∗
F = 6f D /(ρlt l)
(18)
where ρ is tissue density and l t is the average total capillary length (between the arteriole and the venule). An interesting and successful approach is to use 19 F blood substitutes (perfluorocarbons), which remain in the vascular compartment.54 Paramagnetic contrast agents may also be used to enhance signal from flowing blood.55 3.3.1.2 IVCM. Considering now the coherent part of microscopic flow, even echo rephasing has been used in an attempt to observe perfusion in first and second echo difference images,56 but variations in T 2 tend to mask the Intercept [f (flowing blood volume) = 8 ml per 100 g] Initial slope (D* = 11 × 10–3 mm2 s–1)
4.7 4.5
ln (S)
4.3 Slope at high b values (D = 1.1 × 10–3 mm2 s–1)
4.1
9
effects of perfusion. A more powerful technique uses velocity compensation. It is possible to devise two sequences with the same echo time, one of which refocuses the effects of blood moving with constant velocity along capillaries (‘flow compensated’), while the other gives an additional attenuation resulting from this motion. A difference image then highlights the effects of perfusion. This effect has been shown in a model of rat brain ischemia, where the perfusion deficit is clearly visible as the disappearance of the difference between the two images.57 This technique, however, requires excellent hardware stability and high signal-to-noise ratio. The order of flow compensation by gradient nulling may be increased to take into consideration the tortuosity of the capillary network. 3.3.2 Arterial Water Spin Labeling
In the experiment proposed by Williams et al.58 to look at brain perfusion, a series of rf inversion pulses is used to saturate continuously or invert spins of arterial blood water protons. The inflow of these inverted spins in the imaging volume results in a small signal loss S /S 0 , which depends on blood flow F : F = (λ/T1app )(S/S0 − 1)/2α
(19)
where λ is the brain/blood partition coefficient, α is the effective degree of arterial inversion, and T 1app is the apparent spin–lattice relaxation time in the presence of arterial inversion. In practice, one has also to consider T 1 recovery of labeled flowing spins between the excitation and the detection slices, and cross relaxation between tissue water and macromolecules. A critical parameter is the degree of arterial spin labeling α , which depends on the flow velocity and the technique used to label spins. Steady-state spin inversion has been successfully achieved using adiabatic fast passage (AFP) spin inversion. Quantitative perfusion maps have been obtained with this technique at high magnetic fields in animals, which have a short distance between neck and brain. High magnetic fields may be required to achieve a sufficient signal-to-noise ratio and dispose of longer T 1 values. A variant of this technique is to invert or saturate spins in the imaged slice and to monitor entry of fresh spins. Results may not be quantitative, but variations in blood flow during brain activation tasks may be observed.59
3.9 Gray matter
3.7
3.3.3 Blood Oxygenation Level Dependent (BOLD) Contrast
White matter 3.5 3.3 0
200
400 600 b (s mm–2)
800
1000
Figure 15 Plot of the logarithm of the signal decay versus gradient b factor in the presence of IVIM (experiment carried out at 1.5 T on a 70year-old human volunteer using IVIM EPI). A series of 16 differently diffusion-sensitized spin echo images was collected. For white matter, the plot follows a straight line, as expected for a pure diffusion process. The slope of the straight line gives the diffusion coefficient D. For gray matter, the plot clearly shows a curvature effect for low b values. This curvature has been ascribed to microcirculation effects. The straight line obtained for large b values gives the diffusion coefficient. The deviation from the diffusion asymptote, as measured at the intercept, gives the perfusion fraction f . D*, blood velocity
Although the approach based on Gd-DTPA was a significant breakthrough for functional brain imaging, it had several limitations, in particular the requirement to use an external contrast agent. The number of measurements is therefore limited (for instance, to one baseline and one activation image) and brain function cannot be evaluated in real time. Based on earlier works describing the magnetic properties of hemoglobin60 and their effect on the magnetic resonance signal,61 it was recently suggested that deoxyhemoglobin (which is paramagnetic, although oxyhemoglobin is not) in red blood cells could be used as an endogenous contrast material.62,63 Kwong et al.,59 closely followed by others,64,65 showed that MRI could be used to monitor in real time the local modulation of the level of blood oxygenation associated
10 DIFFUSION AND PERFUSION IN MRI 4
Signal change (%)
3 2 1 0
–1 Stimulus –2 On
Off
–3 0:06
0:36 Time (min:sec)
1:06
Figure 16 Typical magnetic resonance (MR) signal time course study during brain activation by a visual stimulus using the BOLD method (1.5 T). Gradient echo EPI images were obtained at 3-s intervals. During the stimulus the MR signal increases. The amplitude of the signal change is usually small, but largely exceeds baseline fluctuations. After stimulation, the MR signal sometimes goes below baseline. There is a delay of several seconds between the beginning of the activation task and the MR signal change, due to the delayed hemodynamic response over the neuronal response
with brain activity on an individual basis, without the need for intersubject averaging. During brain activation tasks, there is a large local increase in blood flow which overcomes a small increase in tissue oxygen uptake. There is, therefore, a surplus of oxyhemoglobin over deoxyhemoglobin, resulting in a reduction of the susceptibility effects and thus in increased MRI signal (Figure 16). The effect is usually small (a few per cent) and depends on the sequence parameters, the voxel size, and the field strength.66,67 Recent articles have demonstrated the tremendous potential of MRI to aid in our understanding of how the brain works, with unmatched combined spatial and temporal resolution68 – 71 (Figure 17). Besides exploration of the normal brain, potential clinical applications include presurgical mapping,72 monitoring of recovery after stroke or head trauma, and monitoring the fate of neuropharmacological agents.73 Presurgical mapping consists in functional evaluation of brain tissue to be removed or spared during surgery, for instance in cases of intractable complex partial seizures if language is to be spared during lobectomy. So far, hemispheric language dominance is determined from invasive and risky tests, such as the Wada test which is based on intracarotid amobarbital injection, and electrical cortical recording through electrodes positioned intraoperatively. It is clear that functional MRI (FMRI) language studies have the potential to replace such invasive tests.68,74 On the other hand, FMRI could also be used to visualize epileptic foci with, or even without, associated clinical seizure. The use of FMRI for monitoring of functional recovery after stroke or head trauma appears especially important for patients with hemiplegia or aphasia. Studies may be repeated at will on individuals during the recovery period without concerns about irradiation. MRI could also be used to visualize regions of the brain where receptors have been activated or inhibited by neuropharmacological agents through effects on blood flow and blood oxygenation, as we recently suggested using
Figure 17 Activation map of the visual cortex. Activation maps can be calculated by identifying pixels where significant magnetic resonance signal change occurred during stimulation. The degree of activation is shown using a color scale. The raw BOLD images were obtained using a spoiled-GRASS gradient echo (SPGR) sequence with a spatial resolution of 0.8 mm × 1.6 mm. The activation map is overlayed on a high resolution anatomical image to show the exact location of activation
neurochemical activation by arecholine in monkey brain.73 Also, the modulation by such drugs of the response of the brain to specific stimuli or cognitive tasks could be evaluated. There are, of course, some limitations when using FMRI. Patient cooperation is particularly necessary. Not only is their active participation required for the activation tasks, but they are also required to stay absolutely immobile during scanning in order to avoid misregistration artifacts. Registration algorithms are being evaluated, as well as sophisticated statistical analysis packages. Another concern regards the size of the vessels (veinules or feeding arteries) responsible for the effect. If the vessels are not small enough, they may correspond to a fairly large territory, so that the area of activation may differ from the location where activation is seen on the images.
3.4 Conclusion
MRI appears to be a very promising imaging modality for looking at microcirculation and perfusion. Compared with other approaches, MRI may provide images of high temporal and spatial resolution totally noninvasively. However, today, methods which give the best spatial and/or temporal resolution are not always quantitative, at least in terms of blood flow or other physiological parameters, often due to the lack of an adequate model. On the other hand, there are techniques, most of them using the basic principles of more conventional methods, which are potentially quantitative, but remain difficult to implement and give images with limited spatial and/or temporal resolution. It is likely that progress
DIFFUSION AND PERFUSION IN MRI
will be made in both directions, so that physiologists and clinicians will soon have available to them methods which give quantitative information on perfusion with unmatched spatial and temporal resolution.
4
RELATED ARTICLES
Anisotropically Restricted Diffusion in MRI; Brain: Sensory Activation Monitored by Induced Hemodynamic Changes with Echo Planar MRI; Diffusion: Clinical Utility of MRI Studies; Whole Body Magnetic Resonance Angiography.
5
REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9.
10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28.
E. O. Stejskal and J. E. Tanner, J. Chem. Phys., 1965, 42, 288. D. Le Bihan, Magn. Reson. Q., 1991, 7, 1. H. C. Torrey, Phys. Rev., 1965, 104, 653. D. Le Bihan, E. Breton, D. Lallemand, P. Grenier, E. A. Cabanis, and M. Laval-Jeantet, Radiology, 1986, 161, 401. P. J. Basser, J. Mattiello, and D. Le Bihan, J. Magn. Reson., 1994, in press. E. L. Hahn, Phys. Rev., 1950, 80, 580. H. Y. Carr and E. M. Purcell, Phys. Rev., 1954, 94, 630. J. E. Tanner, J. Chem. Phys., 1970, 52, 2523. K. D. Merboldt, W. Hanicke, and J. Frahm, J. Magn. Reson., 1985, 64, 479. R. Kaiser, E. Bartholdi, and R. R. Ernst. J. Chem. Phys., 1974, 60, 2966. D. Le Bihan, Magn. Reson. Med., 1988, 7, 346. D. Le Bihan, R. Turner, and J. R. MacFall, Magn. Reson. Med., 1989, 10, 324. K. D. Merboldt, W. Hanicke, M. L. Gyngell, J. Frahm, and H. Bruhn, J. Magn. Reson., 1989, 82, 115. K. D. Merboldt, H. Bruhn, J. Frahm, M. L. Gyngell, W. Hanicke, and M. Deimling, Magn. Reson. Med., 1989, 9, 423. E. X. Wu and R. B. Buxton, J. Magn. Reson., 1990, 90, 243. D. Canet, B. Diter, A. Belmajdoub, J. Braondeau, J. C. Boubel, and K. Elbayed, J. Magn. Reson., 1989, 81, 1. G. S. Karczmar, D. B. Twieg, T. J. Lawry, G. B. Matson, and M. W. Weiner, Magn. Reson. Med., 1988, 7, 111. D. Le Bihan and E. Breton. CR Acad. Sci. Paris, 1985, 301, 1109. M. Neeman, J. P. Freyer, and L. O. Sillerud, J. Magn. Reson., 1990, 90, 303. D. G. Taylor and M. C. Bushell, Phys. Med. Biol., 1985, 30, 345. R. Turner, D. Le Bihan, J. Maier, R. Vavrek, L. K. Hedges, and J. Pekar, Radiology, 1990, 177, 407. R. Turner and D. Le Bihan, J. Magn. Reson., 1990, 86, 445. P. Mansfield, J. Phys. C: Solid State Phys., 1977, 10, L55. T. L. Chenevert, J. G. Pipe, D. M. Williams, and J. A. Brunberg, Magn. Reson. Med., 1990, 17, 197. E. O. Stejskal, J. Chem. Phys., 1965, 43, 3597. R. L. Cooper, D. B. Chang, A. C. Young, C. J. Martin, and D. Ancker-Johnson, Biophys. J., 1974, 14, 161. J. E. Tanner, J. Chem. Phys., 1978, 69, 1748. G. G. Cleveland, D. C. Chang, C. F. Hazlewood, and H. E. Rovschach, Biophys. J., 1976, 16, 1043.
11
29. M. E. Moseley, Y. Cohen, J. Kucharczyk, J. Mintorovitch, H. S. Asgari, M. F. Wendland, J. Tsuruda, D. Norman, and P. Weinstein, Radiology, 1990, 176, 439. 30. D. Le Bihan, R. Turner, and P. Douek, NeuroReport, 1993, 4, 887. 31. M. E. Moseley, J. Kucharczyk, J. Mintorovitch, Y. Cohen, J. Kurhanewicz, N. Derucin, H. Asgari, and D. Normou, Am. J. Neuroradiol., 1990, 11, 423. 32. S. Warach, D. Chien, W. Li, M. Ronthal, and R. R. Edelman, Neurology, 1992, 42, 1717. 33. D. Le Bihan, R. Turner, P. Douek, and N. Patronas, Am. J. Radiol., 1992, 159, 591. 34. D. Le Bihan, J. Delannoy, and R. Levin, Radiology, 1989, 171, 853. 35. S. S. Kety and C. F. Schmidt, Am. J. Physiol., 1945, 143, 53. 36. N. M. Bolas, A. J. Petros, D. Bergel, and G. K. Radda, Proc. Soc. Magn. Reson. Med. 1985 Meet., p. 315. 37. S. Eleff, M. Schnall, L. Ligetti, M. Osbakken, H. Subramanian, B. Chance, and J. S. Leigh, Magn. Reson. Med., 1988, 7, 412. 38. D. Barranco, L. N. Sutton, S. Florin, J. Greenberg, T. Sinnwell, L. Ligetti, and A. C. McLaughlin, J. Cereb. Blood Flow Metab., 1989, 9, 886. 39. J. R. Ewing, C. A. Branch, S. C. Fagan, J. A. Helpern, R. T. Simkins, S. M. Butt, and K. M. A. Welch, Stroke, 1990, 21, 100. 40. J. J. H. Ackerman, C. S. Ewy, N. N. Becker, and R. A. Shalwitz, Proc. Natl. Acad. Sci. USA, 1987, 84, 4099. 41. J. Pekar, L. Ligeti, Z. Ruttner, R. C. Lyon, T. M. Sinnwell, P. van Gelderen, D. Fiat, C. T. W. Moonen, and A. C. McLaughlin, Magn. Reson. Med., 1991, 21, 313. 42. A. L. Hopkins, E. M. Haacke, J. Tkach, R. G. Barr, and C. B. Bratton, Magn. Reson. Med., 1988, 7, 222. 43. K. K. Kwong, A. L. Hopkins, J. W. Belliveau, D. A. Chesler, L. M. Porkka, R. C. McKinstry, D. A. Finelli, G. J. Hunter, J. B. Moore, R. G. Barr, and B. R. Rosen, Magn. Reson. Med., 1991, 22, 154. 44. C. R. Fisel, J. L. Ackerman, R. B. Buxton, L. Garrido, J. W. Belliveau, B. R. Rosen, and T. J. Brady, Magn. Reson. Med., 1991, 17, 336. 45. L. Axel, Radiology, 1980, 137, 679. 46. J. W. Belliveau, B. R. Rosen, H. L. Kantor, R. R. Rzedzian, D. N. Kennedy, R. C. McKinstry, J. M. Vevea, M. S. Cohen, I. L. Pykett, and T. J. Brady, Magn. Reson. Med., 1990, 14, 538. 47. K. L. Zierler, Circ. Res., 1962, 10, 393. 48. N. A. Lassen, J. Cereb. Blood Flow Metab., 1984, 4, 633. 49. J. W. Belliveau, D. N. Kennedy, R. C. McKinstry, R. M. Buchbinder, R. M. Weisskoff, S. M. Cohen, J. M. Vevea, T. J. Brady, and B. R. Bosen, Science, 1991, 254, 716. 50. D. Le Bihan, E. Breton, D. Lallemand, M. L. Aubin, J. Vignaud, and M. Laval-Jeantet, Radiology, 1988, 168, 497. 51. I. R. Young, A. S. Hall, D. J. Bryant, D. G. T. Thomas, S. S. Gill, M. S. Dubowitz, F. Cowan, and J. M. Pennock, J. Comput. Assist. Tomogr., 1988, 12, 727. 52. J. Pekar, P. C. M. van Zijl, and C. T. W. Moonen, Magn. Reson. Med., 1992, 23, 122. 53. D. Le Bihan and R. Turner, Magn. Reson. Med., 1992, 27, 171. 54. J. J. Neil and J. J. H. Ackerman, J. Magn. Reson., 1992, 97, 194. 55. J. J. Neil, L. A. Scherrer, and J. J. H. Ackerman, J. Magn. Reson., 1991, 95, 607. 56. C. B. Ahn, S. Y. Lee, O. Nalcioglu, and Z. H. Cho, Med. Phys., 1987, 14, 43. 57. J. H. Maki, J. R. MacFall, and G. A. Johnson, Magn. Reson. Med., 1991, 17, 95.
12 DIFFUSION AND PERFUSION IN MRI 58. D. S. Williams, J. A. Detre, J. S. Leigh, and A. P. Koretsky, Proc. Natl. Acad. Sci. USA, 1992, 89, 212. 59. K. K. Kwong, J. W. Belliveau, D. A. Chesler, I. E. Goldberg, R. M. Weisskoff, B. P. Poncelet, D. N. Kennedy, B. E. Hoppel, M. S. Cohen, R. Turner, H. M. Cheng, T. J. Brady, and B. R. Rosen, Proc. Natl. Acad. Sci. USA, 1992, 89, 5675. 60. L. Pauling and C. D. Coryell, Proc. Natl. Acad. Sci. USA, 1936, 22, 210. 61. K. R. Thulborn, J. C. Waterton, P. M. Matthews, and G. K. Radda, Biochim. Biophys. Acta, 1982, 714, 265. 62. S. Ogawa, T. M. Lee, A. S. Nayak, and P. Glynn, Magn. Reson. Med., 1990, 14, 68. 63. R. Turner, D. Le Bihan, C. T. W. Moonen, D. Despres, and J. Frank, Magn. Reson. Med., 1991, 22, 159. 64. P. A. Bandettini, E. C. Wong, R. S. Hinks, R. S. Tikofsky, and J. S. Hyde, Magn. Reson. Med., 1992, 25, 390. 65. S. Ogawa, D. W. Tank, R. Menon, J. M. Ellerman, S. G. Kim, H. Merkle, and K. Ugurbil, Proc. Natl. Acad. Sci. USA, 1992, 89, 5951. 66. J. Frahm, H. Bruhn, K. D. Merboldt, and W. Hanicke, J. Magn. Reson. Imag., 1992, 5, 501. 67. R. Turner, P. Jezzard, H. Wen, K. K. Kwong, D. Le Bihan, T. Zeffiro, and R. S. Balaban, Magn. Reson. Med., 1993, 29, 277. 68. G. McCarthy, A. M. Blamire, D. L. Rothman, R. Gruetter, and R. G. Schulman, Proc. Natl. Acad. Sci. USA, 1993, 90, 4952.
69. D. Le Bihan, R. Turner, T. A. Zeffiro, C. A. Cuenod, P. Jezzard, and V. Bonnerot, Proc. Natl. Acad. Sci. USA, 1993, 90, 1 1802. 70. S. Kim, J. Ashe, K. Hendrich, J. M. Ellermann, H. Merkle, K. Urgubil, and A. P. Georgopoulos, Science, 1993, 261, 615. 71. W. Schneider, D. C. Noll, and J. D. Cohen, Nature, 1993, 365, 150. 72. C. R. Jack, R. M. Thompson, R. K. Butts, F. W. Sharbrough, P. J. Kelly, D. P. Hanson, S. J. Rieckrer, R. L. Ehmon, N. J. Hangianoreou, and G. D. Cascino, Radiology, 1994, 190, 85. 73. C. A. Cuenod, M. C. J. Chang, T. Arai, L. Pannier, S. Posse, D. DesPres, J. A. Frank, S. Rapoport, and D. Le Bihan, Proc. Soc. Magn. Reson. Med. 1993 Meet., p. 1387. 74. L. Rueckert, I. Appollonio, J. Grafman, P. Jezzard, R. Johnson, and D. Le Bihan D, J. Neuroimag, 1994, 4, 67.
Biographical Sketch Denis Le Bihan. b 1957. Ph.D., 1987, Physics, M.D., 1984, Radiology, University of Paris, France. Chief, Diagnostic Radiology Research Section, National Institutes of Health, Clinical Associate Professor of Radiology, Georgetown University, USA. Approx. 250 articles, book chapters and abstracts. Research specialties: diffusion, temperature, interventional, perfusion and functional MRI.
DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
Diffusion: Clinical Utility of MRI Studies Ole Henriksen Danish Research Center of Magnetic Resonance, Hvidovre University Hospital, Denmark
1 INTRODUCTION Tissue characterization by MRI has been a subject of great interest because the multiparameter dependence of the MRI signal makes it possible to improve the differentiation between various pathologies in vivo, which may lead to an improved speci®city of clinical MRI investigations. So far, these approaches have mostly been based on signal contrast behavior due to differences in proton densities, and T1 and T2 relaxation behavior between different tissue types. In recent years, methods for studying diffusion processes in vivo by means of MRI have been developed, and possible clinical applications are currently under investigation. The studies published so far have mostly been directed to imaging of diffusion processes of water in the human brain. The results are very promising with respect to future clinical utility. These studies will be described in this article, outlining possible major areas of application for diffusion studies in clinical practice.
2 MRI MEASUREMENTS OF WATER DIFFUSION IN BIOLOGICAL TISSUES IN VIVO A detailed description of the factors which in¯uence the diffusion of water molecules in biological tissues is given in other articles (see the list of related articles). Only those aspects relevant for the interpretation of clinical in vivo diffusion studies will therefore be considered here.
The water diffusion in the cytosol may be hindered by the presence of barriers such as cell membranes, intracellular organelles, the presence of macromolecules, electrical charges, etc. The consequence is that the diffusion path of the water molecule becomes geometry-dependent, and the concept of a diffusion coef®cient depending only on the properties of the diffusion medium breaks down. The practical consequence is that the diffusion coef®cients obtained in MRI studies also include prolonged diffusion paths encountering such barriers (tortuosity), and they should thus be regarded as `apparent diffusion coef®cients', and are denoted D* in the following discussion. The signal contrast in diffusion-weighted imaging depends on the design of the pulse sequence because D* depends on the time (diffusion time) in which the diffusion processes in¯uence the signal obtained by diffusion-sensitized pulse sequences. If the diffusion time is very short, the probability that a water molecule meets a barrier is low compared with that of diffusion over a longer period of time. Thus, D* decreases monotonically from values close to D for short diffusion times to a constant smaller value as the diffusion time is extended to very long periods. The fractional signal attenuation, A, due to diffusion in a given time, , in pulsed gradient experiments, neglecting the small effects of the imaging gradients, can be written as A expÿbD
Translational molecular diffusion processes are a phenomenon in which molecules in a ¯uid of in®nite extent move along random paths, colliding with and moving past each other. Diffusion can be described in statistical terms. The diffusion coef®cient, D, describes the mobility at the molecular level, and has the unit of length squared per unit of time. D is speci®c for a given molecule, and varies with viscosity and temperature. For the free diffusion of water molecules in water the diffusion coef®cient D is approximately 3 10ÿ9 m2 sÿ2 at 37 C. If the diffusion coef®cient of water in a given tissue is less than this value at 37 C, and it varies with the diffusion time, the diffusion is said to be restricted, and depends on a number of factors related to the microstructure of the tissue in question, which will be described brie¯y below.
1
where b is the diffusion attenuation factor in units of time per length squared [equal to 2G22( ÿ /3)] and D*() is the apparent diffusion coef®cient at the diffusion time, ( is the effective gradient duration and is the time between the leading edges of the pulses). Thus, in images displaying D*, tissues with high D* will appear with a high signal intensity. In diffusion-weighted images the situation is the opposite, as tissues with high D* values tend to appear with low signal intensity, but the interpretation is more dif®cult because the signal will also be in¯uenced by T1- and T2-weighting factors. Even though the D* in cerebrospinal ¯uid (CSF) is very high and equals D, the signal appears bright in diffusionweighted images using a diffusion-sensitized spin echo sequence because the T2 of CSF is very long compared with that of the brain tissue.
2.2 2.1 The Concept of Restricted Diffusion
1
Sources of Errors
Apart from imperfections in the MRI equipment, macroscopic movements are a main source of error in diffusion studies in vivo. Patient movements may be minimized by careful ®xation, but one still has to take physiological movements induced by arterial pulsation, respiration, etc., into account. Studies of brain motion using the phase contrast method1 have indicated an average velocity of about 0.5 mm sÿ1 during the cardiac cycle, corresponding to distances of at least 10 m in 20 ms, which are very similar to the expected values for the apparent average diffusion distance with of 20 ms. Cardiac and respiratory synchronization and compensation for ®rst order bulk movements by gradient moment nulling techniques reduce the movement problems considerably when spin echo sequences are used.2,3 Another way is to use singleshot sequences with very short imaging times.3
2 DIFFUSION: CLINICAL UTILITY OF MRI STUDIES Table 1 Clinical MRI studies of diffusion in human brain tissues Tissue CSF White matter Perpendicular Parallel Gray matter
Apparent diffusion coef®cient D*/(10ÿ9 m2 sÿ1) 2.9±3.5 0.2±1.7 0.2±0.9 0.8±1.5 0.7±2.3
References 3,4,5,6 4,7,10,11 2,7 2,7 3,6,8,9,10
3 CLINICAL MRI STUDIES OF DIFFUSION IN BRAIN 3.1 Normal Subjects So far the vast majority of in vivo diffusion studies applied to human beings have been carried out on the brain. The reasons are that the brain movements are small, and the relatively long T2 values of brain tissue make it possible to achieve a reasonable signal-to-noise ratio in high-resolution images (voxel size 8 mm3). Relatively few studies where quantitative measurements of the apparent diffusion coef®cient have been performed in healthy humans have been published so far. The results are summarized in Table 1. The main ®ndings are that the apparent diffusion coef®cient is reduced in gray and white matter compared with the diffusion coef®cient of water molecules in water at 37 C (Figure 1). Studies of white matter
indicate that the diffusion is anisotropic, as the apparent diffusion coef®cient is smaller for diffusion perpendicular to than parallel to the direction of the axons.11 Another important ®nding is that the diffusion appears to be restricted in white matter perpendicular to the direction of the nerve ®bers as D* decreases for diffusion times above 26 ms, which corresponds to an average diffusion distance of about 4±5 m.5 No sign of restricted diffusion is seen in the direction of the ®bers up to distances of 9 m.5 The same seems to be true for gray matter.5 The apparent diffusion coef®cients obtained in CSF may be regarded as a kind of internal control of the validity of the quantitative measurements, since the expected values should be equal to the diffusion coef®cient of water molecules in water at 37 C (D = 3 10ÿ9 m2 sÿ1). In the early study of Thomsen et al.,8 D* in CSF is higher than the expected value of free diffusion in water, indicating that bulk movements contributed to the measurement. In a later study where compensation for movements of constant velocity is incorporated, D* in CSF corresponds to the expected value.2 3.1.1
Age Dependency
Gideon et al.6 have observed that the apparent diffusion coef®cient for water in deep white matter increases with age from 0.7 10ÿ9 m2 sÿ1 (age ~20 years) to approximately 1.0 10ÿ9 m2 sÿ1 (age 60 ±75 years), using a ¯ow-compensated, diffusion-sensitized spin echo sequence with a diffusion time of 19 ms (Figure 2). This relationship between water diffusion and age has not been seen in gray matter.6 Measurements on
Figure 1 Images of the calculated apparent diffusion coef®cient of water in the brain of (a) an adult healthy volunteer and of (b) a premature neonate (gestational age 33 weeks). The calculated images are based on diffusion-weighted spin echo sequences with compensation for ®rst-order bulk movements. ECG triggering was performed as well. The color scale shown to the right of the images is linear, corresponding to D* = 0 (magenta) to D* = 3.5 10ÿ9 m2 sÿ1 (red) which corresponds to the free diffusion coef®cient of water in water at 37 C. The voxel size is 1 1 8 mm3. Note the relatively low D* in white matter perpendicular to the ®ber direction in the adult brain and that D* is considerably less than in the CSF, which corresponds to the free diffusion of water molecules in water. In contrast to the adult brain, the D* in white matter is higher than in gray matter in the neonatal premature brain
DIFFUSION: CLINICAL UTILITY OF MRI STUDIES 1.4
ADC (10–9 m2 s–1)
1.2 1 0.8 0.6 0.4 0.2 0 0
10
20
30 40 50 Age (years)
60
70
80
Figure 2 D* of water in periventricular white matter of healthy subjects plotted against age. Regression analysis shows a signi®cant correlation between D* and age. (Adapted from Gideon et al.6)
neonates and small infants suggest that the degree of water diffusion anisotropy in frontal white matter and optic radiation is less in neonates than in young adults.12 No clear difference is seen between the neonates and adults when D* is measured parallel to the direction of the ®bers.12 3.1.2
Diffusion-weighted Imaging
The studies mentioned so far in which the apparent water diffusion in human brain has been estimated in quantitative terms indicate that the water diffusion in white matter is restricted and there is an appreciable degree of anisotropy. The possible mechanisms underlying the diffusion anisotropy in brain white matter is discussed in Anisotropically Restricted Diffusion in MRI, but the observations in neonates12 suggest that the myelin sheaths may at least be partly responsible. The practical consequence is that anisotropically restricted diffusion creates a signal contrast between nerve ®ber bundles running in different directions in relation to those of the diffusion-sensitizing gradients. This has been shown in a number of studies where major ascending and descending projection tracts are visualized.5,13 It should be recalled that nerve ®bers running perpendicular appear brighter than those running in parallel with the direction of the diffusion-sensitizing gradients. 3.2 Brain Pathologies 3.2.1
Brain Ischemia
A number of animal studies show that the apparent diffusion of water is reduced in experimentally induced ischemic areas of the brain.14±16 These alterations seem to be detectable within the ®rst half hour after the onset of brain ischemia, and occur a long time before changes are detectable in T2-weighted images.15,17 In diffusion-weighted images, focal ischemic areas appear brighter than normal brain tissue in a gray±white scale, because the smaller apparent diffusion coef®cient of water causes less reduction of the signal due to diffusion processes. The animal studies also indicate good agreement between the anatomical spreading of focal brain ischemia as visualized by diffusion-weighted imaging, perfusion-related imaging of the passage of a gadolinium compound (input function is not taken into account), and histological evaluation based on staining
3
with Evans blue.14,17 Diffusion-weighted imaging of photochemically induced ischemia in rat brain shows hyperintense signals corresponding to the ischemic area on the ®rst day. During the following days the signal intensity fades, indicating that the diffusion of water becomes less impeded in the subacute stage.16 Taken together, the animal studies mentioned above clearly demonstrate that the apparent diffusion of water is reduced in ischemic brain tissue and that these alterations occur in the very early phase. The reason for the impeded diffusion of water is unclear at present. It cannot be explained by a decrease in tissue temperature because this has to drop about 10 C in order to explain the observed decrease in the apparent diffusion coef®cient of water of about 30%. It seems to be associated with energy failure secondary to severe ischemia.18 The impediment of diffusion probably re¯ects alterations of the water balance between tissue compartments, swelling of membranes, and accumulation of macromolecules. Only a few studies of water diffusion in the brain of stroke patients have been published. As early as 1987 Thomsen et al.8 had presented evidence suggesting that the apparent diffusion coef®cient of water is reduced in subacute brain infarcts. Warach et al.19 have studied 32 patients with stroke using a highspeed diffusion-weighted stimulated echo sequence (turboSTEAM). The time after the onset of symptoms ranged between 1 h and about 4 months. Diffusion-weighted images con®rm the results of the animal studies described above, as a hyperintense signal is seen in the ischemic area, and these changes can be seen in the very acute stage where no changes are seen in T2-weighted images (Figure 3). In accordance with this, parametric images of the calculated apparent diffusion coef®cient of water show that this is reduced compared with the contralateral side in the acute and subacute stages, whereas the apparent diffusion coef®cient in infarcted tissue is increased in the chronic stage (Figure 4). The absolute values of D* may have been in¯uenced as bulk movements due to gross head movements, brain movements caused by arterial pulsation,1,2,20 and respiration may not have been suf®ciently controlled. However, another study using a diffusion-sensitized spin echo pulse sequence and cardiac triggering also suggests a reduced apparent diffusion of water in ischemic areas following acute stroke.20 These results show a considerable interindividual variation, and they do not indicate a signi®cant correlation between D* and the duration of ischemia. Taken together, the experimental animal and human studies have established that the apparent diffusion coef®cient of water is reduced in brain tissue subjected to severe acute ischemia, and that these alterations can be visualized at a very early stage. The evidence obtained suggests that the apparent water diffusion coef®cient increases within the lesion to supranormal values in the chronic stage of stroke. As mentioned before, the exact pathophysiological events underlying these observations are still poorly understood. When the diffusion values obtained in these often elderly patients are compared with those obtained in healthy volunteers, the latter should be carefully age matched because the apparent diffusion coef®cients in white matter have been shown to increase with age6 (cf Figure 2). Further studies on anisotropy as well as on the relationship between water diffusion characteristics and regional cerebral perfusion are warranted in order to obtain a better understanding of the underlying pathophysiological mechanisms in acute
4 DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
Figure 3 T2-weighted (left) and corresponding diffusion-weighted (right) images obtained (a, b) 105 min, (c, d) 13 h, (e, f ) 9 days after an acute stroke incident. Note that the ischemic tissue located in the left hemisphere can already be clearly visualized approximately 1.5 h after the clinical attack, where no changes are seen in the T2-weighted image. (Reproduced by permission from Warach et al.19)
DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
by using a ¯ow refocusing sequence and cardiac triggering. It is of interest that the apparent diffusion coef®cient is higher in acute lesions than in chronic but stable lesions, and that there is a signi®cant increase in D* in apparently normal white matter in the conventional spin echo images21 (Figure 6). Finally, no relationship between the apparent diffusion coef®cient and the T2 relaxation rate measured by means of a multi-spin echo technique is seen, stressing that these phenomena depend on different mechanisms.21 Diffusion-weighted imaging of a patient with chronic multiple sclerosis shows a high signal in the lesion compared with apparently normal white matter.21 Since D* of water is increased, as shown by Christiansen et al.,21 a low signal should be expected in the diffusion-weighted images. The most likely reason for this is that the T2 weighting dominates compared with the effect of increased diffusion in lesions with high T2 values when pulsed gradient, diffusionsensitized spin echo pulse sequences are used. Even though our experience is relatively limited, the results obtained so far are promising with respect to differentiation between acute and chronic multiple sclerosis lesions.21 The higher D* value in acute than in chronic lesions may re¯ect an increase in water content due to edema and demyelinization in acute lesions, whereas gliosis dominates in the chronic stage. To what extent diffusion studies may contribute to the differential diagnosis between lesions due to multiple sclerosis and other diseases including, for example, subcortical infarcts and viral infections, is still too early to say, as systematic comparative diffusion studies are still lacking. A single case study suggests that diffusion-weighted imaging may give further information in patients with progressive multifocal leukoencephalopathy.13
ADC ratio, lesion/control
, ,, ,,,
2
5
1
0
0 –12 h
12–24 h
24–48 h
5–12 d
24–60 d >4 months
Time
Figure 4 Calculated apparent diffusion coef®cients (ADC) ratios of water in ischemic and contralateral nonischemic brain tissue plotted against time after a stroke incident. Note that the ratios seem to increase from values <1 in the acute and subacute stages to values >1 in the chronic stage. (Adapted from Warach et al.19)
cerebral ischemia, and to be able to evaluate the prognostic information in diffusion-weighted images of brain ischemia with respect to tissue survival. 3.2.2
White Matter Disorders
Quantitative water diffusion studies indicate that the apparent diffusion coef®cient is increased in multiple sclerosis lesions as compared with healthy age-matched volunteers,21 in agreement with preliminary results published by Larsson et al.,22 (Figure 5). Possible in¯uences by movement due to brain pulsation and ¯ow phenomena have been taken into account
Figure 5 (a) T2-weighted and (b) calculated apparent water diffusion coef®cient images obtained in a patient with an acute attack of multiple sclerosis. The color scale shown to the right is similar to that described for Figure 1. Note the increase in diffusion corresponding to the acute lesion seen at the left posterior horn of the lateral ventricles. There also seems to be an increased diffusion in the apparently normal periventricular white matter. (Reproduced by permission from Christiansen et al.21)
6 DIFFUSION: CLINICAL UTILITY OF MRI STUDIES Controls
Multiple sclerosis patients
2.00 1.90 1.80 1.70 1.60 1.50 1.40 1.30 1.20 1.10 1.00 0.90 0.80 0.70 0.60 0.50
Normal white matter
White matter without plaques
Chronic multiple sclerosis plaques
Acute multiple sclerosis plaques
Figure 6 Apparent diffusion coef®cient of water in white matter obtained in age-matched controls (left) and patients with chronic and acute multiple sclerosis lesions. Note that the average diffusion coef®cient obtained in apparently normal white matter of multiple sclerosis patients is signi®cantly increased compared with the control group. (Adapted from Christiansen et al.21)
3.2.3
Intracranial Tumors
Only a few studies on a limited number of patients with intracranial tumors have been reported so far.5,23,24 The ®ndings suggest both an increase and decrease in D* within, as well as between, tumors compared with normal brain tissue
Figure 7 (a) T2-weighted and (b) calculated apparent water diffusion coef®cient images, obtained from a patient with a cystic astrocytoma. Note that D* in the central part of the tumor is close to the free diffusion of water in water, indicating a cystic component of the lesion containing free ¯uid
(Figure 7). The increase in D* is most prominent in cystic parts of the tumor approaching that of CSF and water. Diffusion measurements seem to make it possible to differentiate between solid tumor tissue and cystic lesions having the same appearance on conventional T1- and T2-weighted images due to a high content of paramagnetic protein. In the
DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
latter case, diffusion studies will reveal D* values corresponding to cyst ¯uid.25 Diffusion studies may also improve the assessment of the delineation between epidermoid tumor and adjacent CSF.23 The extent to which diffusion studies may improve the tissue characterization of solid brain tumors is as yet unclear. Another important clinical aspect is whether diffusion studies allow differentiation between solid tumor and surrounding edema. Anecdotal evidence suggests that the signals from presumed edema in diffusion-weighted images are isointense with the solid tumor components but the edema may be associated with a higher degree of diffusion anisotropy.13 It is, however, much too early to draw any de®nite conclusions. 3.2.4
Disturbances in CSF Hydrodynamics
Preliminary studies indicate that the apparent diffusion coef®cient of water is signi®cantly higher in white as well as gray matter of patients with normal-pressure hydrocephalus than in a group of healthy age-matched controls26 (Figure 8). Surprisingly, the diffusion coef®cients in the CSF of the patients seem also to be higher than in the controls, even though the diffusion-sensitive pulse sequences were compensated for ®rst-order motion, and ECG triggering was performed. Since pulsatile motion of CSF is increased in these patients,27 a higher degree of complex ¯ow patterns may have contributed to the signal loss. This may explain the apparent elevation of the diffusion coef®cients in CSF of patients with normal-pressure hydrocephalus. Examination of two patients with high-pressure hydrocephalus also suggests a higher apparent diffusion coef®cients in both white and gray matter compared with normal subjects.26 The increase in the apparent diffusion of water in brain white matter of patients with normal-pressure hydro-
7
cephalus may re¯ect the loss of myelin sheaths, especially in the periventricular regions. Soelberg Sùrensen et al.28 have observed a pronounced increase in the apparent diffusion coef®cient of water in patients with benign intracranial hypertension or pseudotumor cerebri. The alterations in water diffusion are most pronounced in the periventricular white matter. In some cases the calculated apparent diffusion coef®cient exceeds that of free diffusion of water in water, indicating that the obtained diffusion values may to some extent have been in¯uenced by pulsatile movements of the brain as well as by convective water movements in the brain tissue. In a later study where ®rst order movement was corrected for in the pulse sequence, Gideon et al.26 con®rm that the apparent diffusion is increased in periventricular white matter of patients with benign intracranial hypertension, whereas no signi®cant changes are seen in gray matter (Figure 9). Thus, it seems that the increased apparent diffusion coef®cient of water in gray matter observed by Soelberg Sùrensen et al.28 can most likely be ascribed to pulsatile brain movements, which may be more pronounced in these patients than in healthy volunteers. Quantitative measurements of T1 and T2 of gray and white matter, respectively, show no differences between patients with benign intracranial hypertension and healthy volunteers.28 In accordance with this, T1- and T2weighted images do not show any abnormalities, indicating that quantitative diffusion imaging is much more sensitive to alterations in intra- and extracellular water balance and dynamics. 3.2.5
Brain Edema
Sevick et al.29 have studied the in¯uence of cytotoxic edema secondary to hyponatremia on the apparent diffusion
Figure 8 (a) T2-weighted and (b) calculated apparent water diffusion coef®cient images obtained from a patient with normal-pressure hydrocephalus. There is an increased diffusion in the periventricular white matter as compared with normal subjects. (Reproduced by permission from Gideon et al.6)
8 DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
Figure 9 (a) T2-weighted and (b) calculated apparent water diffusion coef®cient images obtained in a patient with benign intracranial hypertension. Note the marked increase in D* in the periventricular white matter in the diffusion images, whereas the T2-weighted image appears normal. (Reproduced by permission from Gideon et al.6)
coef®cient of water in rat brain. The results suggest that cytotoxic edema reduces the apparent diffusion coef®cient of water in the brain by about 10%, and there seems to be an inverse relationship between the calculated D* value and the total water content of the brain tissue. It cannot, however, be totally excluded that the results may at least partly have been in¯uenced by a concomitant reduction in pulsatile movements of the edematous brain, since no cardiac triggering or compensation for ®rst order movements was performed. Anyway, the relatively small decrease in the apparent diffusion coef®cient observed in cytotoxic brain edema does not explain the alterations seen in acute brain ischemia described in a previous section. A number of studies indicate that the apparent diffusion coef®cient of water is higher in tumor-induced vasogenic brain edema than in normal gray and white matter13 (cf Figure 6). Taken together, the preliminary evidence suggests that quantitative diffusion studies may make it possible to differentiate between cytotoxic (D* reduced) and vasogenic brain edema (D* increased). 3.2.6
Pediatric Disorders
The experience with diffusion studies of infants is very limited at present. Rutherford et al.30 have recently published a study in which they examined 31 neonates and infants with various disorders of the brain. They did not calculate the apparent diffusion coef®cient. Cysts and ¯uid collections are recognized by their hypointense signals, indicating an increase in the D* value of water. Low signal intensities are seen in chronic infarcts, consistent with a relatively free isotropic diffusion, whereas the signals are low in intracerebral
hematomas, suggesting restricted isotropic diffusion. Patients with leukodystrophy associated with congenital muscle dystrophy reveal an anisotropic diffusion pattern. In three cases, abnormalities can be seen in the diffusion-weighted images despite conventional imaging being normal. Finally, abnormalities in the corticospinal tract are frequently seen in children with a history of birth asphyxia. Even though the retrospective character of the study does not allow any de®nite conclusions to be drawn, the results are very promising, indicating that diffusion studies may contribute to the diagnosis of a number of pediatric brain disorders. Before reaching that point, however, a large number of systematic studies with quantitative measurements of the apparent diffusion coef®cient of water are warranted in order to interpret the signal changes which are seen in the diffusion-weighted images.
4
OTHER ORGANS
Only a very limited number of diffusion studies of organs other than the brain are available at present. The reason is that diffusion studies may be hampered by considerable gross movements of the organ, caused by respiration and arterial pulsation. Evidence of anisotropic water diffusion has been obtained in human skeletal muscle. The apparent diffusion coef®cients of water range from 1.1 10ÿ9 to 1.5 10ÿ9 m2 sÿ1 perpendicular to and are about 2 10ÿ9 m2 sÿ1 parallel to the muscle ®bers, giving an anisotropic ratio of about 0.7.31,32
DIFFUSION: CLINICAL UTILITY OF MRI STUDIES
The ability to map changes in tissue temperature noninvasively in vivo in been investigated in human skeletal muscle during hyperthermia.33±35 Noninvasive mapping of tissue temperatures in vivo may offer great clinical potential in the control of hyperthermia therapy. The question is whether diffusion measurements as described above are suf®ciently accurate and reproducible. A study by Morvan et al.35 suggests that it is possible to measure temperature differences of approximately 2 C. This approach should be compared with an alternative method where changes in resonance frequencies induced by alterations in tissue temperature are measured.36 The latter method may turn out to be more sensitive and reproducible. Preliminary studies indicate that it may be possible to obtain meaningful diffusion-weighted images of human kidneys in vivo by means of an ultrafast diffusion-sensitized echo planar imaging sequence.37 It is much too early to speculate on the clinical signi®cance, however, because water transport in the kidney is very complex. Furthermore, substantial sources of error due to motion have to be taken into account. Ultrafast techniques combined with breath holding may be a part of the solution. 5 DIFFUSION-WEIGHTED SPECTROSCOPY Recent developments in localized spectroscopic techniques have made it possible to take advantage of chemical shift information to extend diffusion measurements to include molecules other than water. By means of 31P spectroscopy it is possible to study the diffusion of various phosphorus compounds, including phosphocreatine, in skeletal muscle.38 Diffusion characteristics of N-acetylaspartate, total creatine, and cholinecontaining compounds have been measured in anesthetized rat brain,39 and recently also in human brain,40 using a diffusionsensitized localized proton spectroscopic sequence with water suppression. The calculated apparent diffusion coef®cients (mean 1 SD) of N-acetylaspartate, choline-containing compounds, and total creatine in normal human brain are (0.18 0.02) 10ÿ9, (0.13 0.03) 10ÿ9, and (0.15 0.03) 10ÿ9 m2 sÿ1, respectively.40 Applications of diffusion-weighted spectroscopic techniques are very interesting because compounds which do not penetrate the cell membranes can be used to study the intracellular medium, including viscosity, cell dimensions, etc. At present, the drawback is relatively poor spatial resolution, which means that the calculated D* values represent a weighted average of diffusion coef®cients in a mixture of tissue types consisting of many different cell types. 6 CLINICAL UTILITY OF MRI DIFFUSION STUDIES: A GENERAL VIEW Although clinical experience with diffusion imaging is rather limited, it is possible to outline some general trends with respect to clinical utility. So far the method has almost exclusively been applied to the brain, because problems due to movement can be dealt with, and secondly, the relatively high T2 values make it possible to achieve a reasonable signal-tonoise ratio at the long echo times which are necessary in order to obtain suf®cient diffusion sensitivity.
9
One of the most striking areas for the use of diffusionweighted imaging is early visualization of brain ischemia. This will most likely have a great impact on treatment strategies, where success depends critically on a very early start to intervention. Further research is needed in order to evaluate the prognostic information of alterations in diffusion characteristics, with particular emphasis on questions related to the penumbra zones.41,42 Studies on anisotropy during the course of infarction may give new information about pathophysiological mechanisms operating during severe brain ischemia. Secondly, diffusion studies may contribute to tissue characterization in the brain and thus improve the speci®city of MRI investigations, taking advantage of the information related to anisotropy. Such studies may contribute to the separation of solid tumors and edema, identi®cation of ¯uid compartments, and maybe also to differentiation between cytotoxic and vasoactive edema. Quantitative diffusion imaging is very sensitive to subtle changes in water balance and viscosity, as observed in patients with benign intracranial hypertension, where imaging by all other imaging modalities, including radio isotope techniques, is normal. Thirdly, it is most likely that diffusion studies will contribute signi®cantly to the characterization of brain white matter disorders. Alterations in diffusion anisotropy may give information of lesions in speci®c nerve tracts. Quantitative measurements of the apparent diffusion coef®cients may contribute, for example, to the differentiation between acute and chronic multiple sclerosis lesions. Diffusion-weighted spectroscopy may offer unique information about speci®c molecules. The study of chemical compounds which do not penetrate cell membranes and other structures makes it possible to obtain information about cell dimensions, intracellular ¯uid viscosity, etc. It is important to avoid errors due to motion in diffusion studies. Cardiac gating and compensating for ®rst-order movements are mandatory unless single acquisition techniques such as echo planar imaging are used. Quantitative measurements of apparent diffusion coef®cients in pathologies are needed in order to obtain the knowledge necessary for interpretating the diffusion-weighted images where the signal intensity is also in¯uenced by proton density, and T1 and T2 relaxation. Thus, diffusion-weighted imaging is a very challenging ®eld which offers great potential for biomedical research and diagnosis. Systematic clinical studies are now required in order to de®ne the clinical indications for diffusion-weighted imaging by magnetic resonance.
7
RELATED ARTICLES
Anisotropically Restricted Diffusion in MRI; Brain MRS of Infants and Children; Susceptibility and Diffusion Effects in NMR Microscopy.
8
REFERENCES
1. D. Greitz, R. Wirestam, A. Franck. B Nordell, C. Thomsen, and F. StaÊhlberg. Neuroradiol., 1992, 34, 370.
10 DIFFUSION: CLINICAL UTILITY OF MRI STUDIES 2. H. B. W. Larsson, M. Stubgaard, C. Thomsen, and The Scandinavian Flow Group, Proc. 9th Ann Mtg. Soc. Magn. Reson. Med., New York, 1990, p. 389. 3. D. Le Bihan, R. Turner, C. T. W. Moonen, and J. Pekar, JMRI, 1991, 1, 7. 4. D. Chien, R. B. Buxton, K. K. Kwong, T. J. Brady, and B. R. Rosen, J. Comput. Assist. Tomogr., 1990, 14, 514. 5. D. Le Bihan, R. Turner, and P. Douek, and N. Patronas, Am. J. Roentgenol., 1992, 159, 591. 6. P. Gideon, C. Thomsen, and O. Henriksen, JMRI, 4, 185. 7. P. Douek, R. Turner, J. Pekar, N. Patronas, and D. Le Bihan, J. Comput. Assist. Tomogr., 1991, 15, 923. 8. C. Thomsen, O. Henriksen, and P. Ring, Acta Radiol., 1987, 28, 353. 9. D. Le Bihan, E. Breton, D. Lallemand, P. Grenier, E. Cabanis, and M. Laval-Jeantet, Radiology, 1986, 161, 375. 10. K. Harada, N. Funita, K. Sakurai, Y. Akai, S. Kim, N. Nakanishi, and T. Kozuka, Proc. 9th Ann Mtg. Soc. Magn. Reson. Med., New York, 1990, 375. 11. T. L. Chenevert, J. A. Brunberg, and J. G. Pipe, Radiology, 1990, 177, 401. 12. H. Sakuma, Y, Nomura, K. Takeda, T. Tagami, T. Nakagawa, Y, Tamagawa, Y Ishii, and T. Tsukamoto, Radiology, 1991, 180, 229. 13. J. V. Hajnal, M. Doran, A. S. Hall, A. G. Collins, A. Oatridge, J. M. Pennock, I. R. Young, and G. M. Bydder, J. Comput. Assist. Tomogr., 1991, 15, 1. 14. M. E. Moseley, J. Kucharczyk, J. Mintorovitch, Y. Cohen, J. Kurhanewicz, N. Derugin, H. Asgari, and D. Norman, Am. J. Neuroroentg., 1990, 11, 423. 15. M. E. Moseley, Y. Cohen, J. Mintorovitch, L Chileuitt, H. Shimizu, J. Kucharczyk, M. F. Wendland, and P. R. Weinstein, Magn. Reson. Med., 1990, 14, 330. 16. N. van Bruggen, B. M. Cullen, M. D. King, M. Doran, S. R. Williams, D. G. Gadian, and J. E. Cremer, Stroke, 1992, 23, 576. 17. M. Maeda, S. Itoh, H. Ide, T. Matsuda, H. Kobayashi, T. Kubota, and Y. Ishii, Radiology, 1993, 189, 227. 18. A. L. Busza, K. L. Allen, M. D. King, N. van Bruggen, S. R. Williams, and D. G. Gadian, Stroke, 1992, 23, 1602. 19. S. Warach, D. Chien, W. Li, M. Ronthal, and R. R. Edelman, Neurology, 1992, 42, 1717. 20. D. Chien, K. K. Kwong, D. R. Gress, F. S. Buoanno, R. B. Buxton, and B. R. Rosen, Am. J. Neuroroentg., 1992, 13, 1097. 21. P. Christiansen, P. Gideon, C. Thomsen, M. Stubgaard, O. Henriksen, and H. B. W. Larsson, Acta Neurol. Scand., 1993, 87, 195. 22. H. B. W. Larsson, C. Thomsen, J. Frederiksen, M. Stubgaard, and O. Henriksen, Magn. Reson. Imag., 1992, 10, 7. 23. J. S. Tsuruda, W. M. Chew, M. E. Moseley, and D. Norman, Am. J. Neuroroentg., 1990, 11, 925.
24. J. S. Tsuruda, W. M. Chiew, M. E. Moseley, and D. Norman, Magn. Reson. Med. 1991, 19, 316. 25. M. Maeda, Y. Kawamura, Y. Tamagawa, T. Matsuda, S. Itoh, H. Kimura, T. Iwasaki, N. Hayashi, K. Yamamoto, and Y. Ishii, J. Comput. Assist. Tomogr. 1992, 16, 514. 26. P. Gideon, C. Thomsen, F. StaÊhlberg, O. Henriksen, P. Soelberg Sùrensen, and F. Gjerris, Proc. XIth Ann Mtg. Soc. Magn. Reson. Med., Berlin, 1992, p. 1609. 27. P. Gideon, F. StaÊhlberg, C. Thomsen, F. Gjerris, P. Soelberg Sùrensen, and O. Henriksen, Neuroradiology, 1994, 36, 210. 28. P. Soelberg Sùrensen, C. Thomsen, F. Gjerris, J. Schmidt, L. Kjaer, and O. Henriksen, Neurol. Research, 1989, 11, 160. 29. R. J. Sevick, F. Kanda, J. Mintorovitch, A. I. Arieff, J. Kucharczyk, J. S. Tsuruda, D. Norman, and M. E. Moseley, Radiology, 1992, 185, 687. 30. M. A. Rutherford, F. M. Cowan, A. Y. Manzur, L. M. S. Dubowitz, J. M. Pennock, J. V. Hajnal, I. R. Young, and G. M. Bydder, J. Comput. Assist. Tomogr. 1991, 15, 108. 31. M. E. Moseley and M. F. Wendland, Proc. Xth Ann Mtg. Soc. Magn. Reson. Med., San Francisco, 1991, 108. 32. D. Morvan, A. Leroy-Willig, P. Jehenson, C. A. Cuenod, and A. Syrota, Radiology, 1992, 185, 871. 33. D. Le Bihan, J. Delannoy, and R. L. Levin, Radiology, 1989, 171, 853. 34. A. S. Hall, M. V. Prior, J. W. Hand, I. R. Young, and R. J. Dickinson, J. Comput. Assist. Tomogr., 1990, 14, 430. 35. D. Morvan, A. Leroy-Willig, A. Malgouyres, C. A. Cuenod, P. Jehenson, and A. Syrota, Magn. Reson. Med., 1993, 29, 371. 36. J. de Poorter, C. de Wagter, C. Thomsen, Y. de Deene, F. StoÊhlberg, and E. Acten, J. Magn. Reson., 1995, 33(1), 74. 37. R. Turner, D. Le Bihan, J. Maier, R. Vavrek, and J. Baich, Proc. 9th Ann Mtg. Soc. Magn. Reson. Med., New York, 1990, 1136. 38. C. T. W. Moonen, P. C. M. van Zijl, D. Le Bihan, and D. Des Pres, Magn. Reson. Med., 1990, 13, 467. 39. K.-D. Merboldt, D. HoÈrstermann, W. HaÈnicke, H. Bruhn, and J. Frahm, Magn. Reson. Med., 1993, 29, 125. 40. S. Posse, C. A. Cuenod, and D. Le Bihan, Radiology, 1993, 188, 719. 41. M. J. Quast, N. C. Huang, G. R. Hillman, and T. A. Kent, Magn. Reson. Imag., 1993, 11, 465. 42. J. Seega and B. Elger, Magn. Reson. Imag., 1993, 11, 401.
Biographical Sketch Ole Henriksen. b 1944. M.D., 1973. Specialist, Clinical Physiology and Nuclear Medicine, 1982. Doctor, Medical Science, 1977. Professor, MRI, 1990. Director, The Danish Research Center of Magnetic Resonance, 1985±present. Approx. 100 publications on MRI since 1985. Research interests are mainly in vivo studies of physiological and pathophysiological mechanisms in humans using MRI techniques including diffusion, ¯ow perfusion, and spectroscopy measurements.
1
Methods and Applications of Diffusion MRI
will be given to the dif®culties encountered when implementing diffusion imaging and the potential of diffusion measurements for clinical applications.
Denis Le Bihan Service Hospitalier FreÂdeÂric Joliot, CEA, Orsay, France
1 INTRODUCTION Diffusion is the process by which matter is transported from one part of a system to another as a result of random molecular motion, also called Brownian motion. This motion, of thermal origin, results in the macroscopic ¯ux of different molecular species, which can be observed in nonuniform systems and is characterized by a diffusion coef®cient. Classically, diffusion coef®cients may be determined by measuring the concentration of molecular species at different times using either physical or chemical methods based on the classical ®rst Fick Law.1 Such approaches rely on the introduction of a tracer in the medium, as similar as possible to the studied molecular species, and then monitoring the concentration of the tracer in the medium by chemical or radiotracer techniques.2 Microscopic displacements on the scale of millimeters can be seen with such tracers over diffusion times of minutes. Such tracer methods have been successfully applied to biological systems, such as the brain3 but are, of course, extremely invasive. An alternative in studying diffusion is to monitor the random walks of molecules. On a statistical scale, the diffusion re¯ects the meansquare distance traveled by molecules in a given interval of time (m2 sÿ1). Diffusion NMR, which relies on these principles, is the only method today available to provide noninvasively such information on molecular displacements which occur over diffusion distances that extend largely beyond elementary molecular jumps, justifying the considerable success of diffusion NMR in physics and chemistry, and more recently in biology.4 Taking typical values for water diffusion ~110ÿ3 mm2 sÿ1) and diffusion time (~20 ms) achievable on conventional MRI equipment free water molecules diffuse over distances on the order of 6 m, which is about the size of many tissue structures. NMR has, in principle, a displacement sensitivity of around 100 nm.5 Because of its noninvasive nature, it is especially suited to probing the molecular dynamics and structural information of biological systems, as well as transport processes. The recent combination of such principles with MRI6±8 represents a spectacular and somewhat unpredicted development in the ®eld of medical sciences. Measuring molecular displacements in biological tissues in vivo may have enormous value, from the determination of the molecular organization in tissues to the emergency management of stroke patients or the monitoring of laser surgery. In this chapter, the basic principles of diffusion measurements with NMR will be presented, as well as the various methods that have been proposed to produce images of diffusion. Some technical considerations will be discussed, followed by a description of the effects on diffusion measurements of microdynamic and microstructure in biological tissues. Emphasis
2
DIFFUSION MRI
Although the effects of diffusion on the NMR signal have been described very early,9,10 most diffusion NMR studies started after the seminal work of Stejskal and Tanner, who introduced the bipolar pulse ®eld gradient method (Figure 1).11 This approach not only provides much better control over what is actually measured but also simpli®es the understanding of the encoding of the diffusion process in the NMR signal. The purpose of these gradient pulses, the duration and the separation of which are classically represented, respectively, as and , is to label spins carried by diffusing molecules magnetically. To some extent, these labeled spins are like the endogenous tracers that can be monitored with classical diffusion measurement tracer methods. However, the behavior of these labeled spins in an NMR experiment leads to more peculiar and speci®c results that are unique to diffusion NMR, and diffusion data obtained with NMR may not always be directly comparable with those obtained by other means. The ®rst gradient pulse induces a phase shift '1 of the spin transverse magnetization, which depends on the spin position z1:
1
' Gz1
where is the gyromagnetic ratio and G is the gradient strength (here, along the z axis). After the 180 rf pulse, '1 is inverted to ÿ'1. The second pulse will produce a phase shift '2: '2 Gz2
2
where z2 is the spin position during the second pulse. The net transverse magnetization Mxy is: Mxy=M0 expi
'2 ÿ '1 expi G
z2 ÿ z1
3
where all relaxation effects have been incorporated into M0.
90°
180°
Echo
G
G
d
d D
Figure 1 Stejskal and Tanner diffusion spin-echo sequence. Gradient pulses, G, are separated by a time interval . Their duration is
For References see p. 16
2 METHODS AND APPLICATIONS OF DIFFUSION MRI Obviously for `static' spins, z1=z2 and the bipolar gradient pair has no effect. For `moving' spins, however, there is a net dephasing that will depend on the spin history during the time interval between the pulses. Indeed, we must now consider a population of spins, which may have different motion histories. This leads to an attenuation of the Mxy and the resulting signal. Assuming << (negligible displacement during ), these equations give a time-dependent part that represents the attenuation owing to diffusion, A:
4 A d z1 P
z1 expi G
z2 ÿ z1 P
z2 ; z1 ; d z2
the selected slice pro®le. In addition, the relaxation time T2 of the medium must not be too short to allow enough diffusion attenuation to occur during TE.
where P(z1) is the probability for a labeled spin to be at initial position z1 and P(z2, z1, ) is the diffusion propagator, i.e., P(z2, z1, )dz2 is the conditional probability to ®nd a labeled spin initially at position z1 in position z2 after a time interval . For a true, isotropic diffusion process:
A expÿ 2 G2 D2
ÿ =3
@P
z2 ; z1 ; t=@t Dr2 P
z2 ; z1 ; t
5
and P is Gaussian, at t=:
6
where D is the diffusion coef®cient. Combining Equations (1) and (2) gives: A expÿ
G2 D expÿ
G2 hz2 i=2
7
where the root mean square diffusion displacement hz2i that occurred during the diffusion time is easily interchangeable with D according to Einstein's equation, hz2i=2DTd. Diffusion coef®cients or root mean square diffusion distances are classically obtained by varying either G or and by measuring the slope of the semilogarithmic plot of the signal intensity versus (G)2. However, it is the diffusion path that the NMR experiment is actually sensitive to, not the diffusion coef®cient. Therefore, in this simple bipolar gradient experiment, clear information on spin displacements can directly be obtained from the signal attenuation A, assuming P is Gaussian. This model is valid only for diffusion in an in®nite and homogeneous medium. If diffusion is restricted or anisotropic (see below), the signal attenuation must be calculated using a more general formalism based on diffusion tensors.12 2.1 Diffusion and the NMR Signal: Different Approaches Almost any NMR sequence can be designed to measure diffusion.
Bipolar Gradient Pulse Spin Echo Technique
The bipolar gradient pulse spin echo technique is, by far, the sequence most widely used for NMR diffusion measurements. In practice, as is usually not negligible compared with , movement during the application of the gradient pulses can no longer be neglected. Taking into account spin diffusion during , Equation (7) becomes:11
9
The diffusion time can be formally de®ned as (ÿ/3), but its physical signi®cance is clear only when is suf®ciently short compared with . Tissues with short T2 times require short TE values, which may not allow suf®ciently long diffusion times to produce enough diffusion effects. 2.1.3
P
z2 ; z1 ;
4Dÿ1=2 expÿ
z2 ÿ z1 2 =4D
2.1.1
2.1.2
Stimulated Echo Technique
A stimulated echo is generated from a sequence consisting of three rf pulses separated by time intervals 1 and 2. Gradient pulses must be inserted within the ®rst and the third periods of the stimulated echo sequence (Figure 2). The diffusion time now includes 2 and can be much longer than with the spin echo sequence without T2-related signal decay because only longitudinal relaxation occurs during this time interval. The Stejskal±Tanner relation [Equation (9)] still applies, provided that the period 2 is included in .13 The longer diffusion time is useful for studying very slow diffusion rates or for compensating for the unavailability of large gradients.7 Unfortunately, the signal is reduced by 50% compared with the spin echo signal. 2.1.4
Gradient Echo Technique
The effect of diffusion on the amplitude of a gradient echo formed by a bipolar gradient pulse pair of reversed polarity does not differ from that of a spin echo sequence.14 However, if the gradient echo is part of a steady-state free precession (SSFP) sequence, where some degree of phase coherence is
90°
90°
Stimulated echo
90°
t1
t2
t1
G
G
d
d
Constant Field Gradient Spin Echo Method
In the presence of a simple constant linear gradient G0, the echo attenuation is:9,10 A
nTE exp
ÿ
2
G20 DTE 3 =12n
8
where TE is the echo time and n the echo number in a multiple echo experiment. This simple approach is dif®cult to combine with NMR imaging because the presence of a constant gradient during the rf pulses of the imaging sequence severely impairs For list of General Abbreviations see end-papers
D
Figure 2 Diffusion-stimulated echo sequence. The stimulated echo results from the spin excitation by three rf pulses separated by time intervals 1 and 2. During 2, relaxation is driven by T1 and not T2. Diffusion effects can be enhanced by placing gradient pulses within the 1 periods, where transverse magnetization is sensitive to ®eld inhomogeneities
METHODS AND APPLICATIONS OF DIFFUSION MRI
2.1.5
0 –0.1 –0.2 log (S/S0)
propagated throughout successive cycles, multiple echo paths with different diffusion times and different diffusion weighting must be considered.15 In MRI, the contrast-enhanced (CE-Fast) scheme has been proposed.16±18 Unfortunately, this sequence remains very sensitive to motion artifacts.19 Moreover, the effects of diffusion and relaxation are mingled so that diffusion measurements are always contaminated to some degree by relaxation effects.20
3
–0.3 –0.4 –0.5
Diffusion Measurements with B1 Field Gradients
Diffusion measurements can also be achieved by means of the rf (B1) ®eld produced by a NMR rf coil oriented perpendicularly to the main transmit/receive NMR coil.21±24 With rf gradients, extremely short switching times can be achieved since there are no eddy currents. Furthermore, substantial gradient strength may be produced from the rf transmitter, allowing measurements of very low diffusion coef®cients.
–0.6 –0.7 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 b (¥103 s mm–2)
Figure 3 The relationship of the logarithm of the signal intensity in brain white matter with the gradient factor b is linear; the slope is equal to the diffusion coef®cient D=0.60 (0.01) 10ÿ3 mm2 sÿ1
2.2 Combining Diffusion NMR with MRI 2.2.1
Principles
Diffusion-weighted (DW) images are obtained by inserting gradient pulses into any NMR imaging sequence. For quanti®cation, it is necessary to determine the degree of diffusion weighting of the sequence. When inserting gradient pulses for diffusion, however, the combination of the imaging and the diffusion gradient pulses produces cross-terms. These cross-terms, which depend on the sequence design, may also lead to signi®cant diffusion-related attenuation effects,14,25 which must be taken into account; consequently the Stejskal±Tanner relation [Equation (9)] is incorrect in most cases. A more general formalism must then be developed to solve the Bloch±Torrey equation.26 Solutions may become analytically complex,27 and it has been suggested that all gradient effects should be combined in a term generally known as the `b factor':14,28 b 2
TE
t 0
0
2 G
t0 d t0 d t
10
The signal attenuation for, unrestricted, isotropic diffusion is then reduced to a simple, convenient expression: A exp
ÿbD
11
This b factor characterizes the diffusion sensitivity of the sequence, as TE characterizes the degree of T2 weighting of a spin echo sequence. Without additional gradients, typical spin echo imaging sequences have intrinsically very low b values, typically less than 1 s mmÿ2; consequently diffusion effects are negligible (for pure water at room temperature the attenuation is less than 1%).14,29 As the raw signal depends not only on diffusion but also on other MR parameters, such as T1 or T2, a quantitative determination of the diffusion coef®cient requires at least two images acquired with different diffusion sensitivities (i.e., b factor values), but better accuracy is obtained when more than two images are used (see below). With the b factor values associated with each image, it is possible to compute diffusion images, i.e., images where the diffusion coef®cient is determined for each pixel. The computation of such diffusion
images is obtained by ®tting the signal intensity of each pixel obtained for different b factor values with Equation (11)4,6,7,14,28 using a regression analysis (Figure 3). On diffusion images, diffusion values are displayed using a gray scale where brightness corresponds to high, fast diffusion and darkness to low, slow diffusion (Figure 4). To save on acquisition and processing time, some investigators have proposed to limit the diffusion studies to the analysis of the raw images obtained with some degree of diffusion sensitivity or weighting. For this reason, these images are called DW images. Contrast in these images is opposite to that found in true diffusion images: regions with high diffusion have more pronounced MRI signal attenuation and appear dark, while regions with low diffusion appear bright. The DW image may be convenient to use, but its content is affected by many parameters other than diffusion, as it is also usually strongly T1 and T2 weighted. As these parameters may not have the same behavior as diffusion, variations in image intensity may be dif®cult to interpret. For instance, acute stroke lesions have decreased diffusion (see below) and appear bright on DW images, while subacute infarcted areas, which are associated with increased diffusion, may also appear bright on DW images because T2 has increased (T2 `shine-through' effect). DW images may be convenient to use in clinical practice, but, whenever possible, absolute diffusion images should be preferred. 2.2.2
Diffusion MRI Sequences
Using the spin echo sequence, it is possible to vary the strength or the duration of the diffusion-sensitizing gradients or their direction in order to enhance anisotropic diffusion effects14 (see below). The spin echo two-dimensional Fourier (2DFT) method is certainly the simplest to implement.6,8,28,29 Other sequences4 that have been investigated are stimulated echo sequences,7 line-integral projection reconstruction,30±32 variants of the SSFP technique; and `turbo' sequences such as HASTE,33 TurboFlash34 and fast spin echo.35 These schemes have been suggested to overcome, at least partially, some of the problems encountered with the spin echo 2DFT method. For References see p. 16
4 METHODS AND APPLICATIONS OF DIFFUSION MRI
Figure 4 Diffusion image (left) calculated from a set of 16 diffusion-weighted coronal echo-planar images of the brain of a normal volunteer. These are 64 64 pixel images with a ®eld of view of 16 cm and, therefore, an in-plane resolution of 2.5 mm 2.5 mm. The section thickness was 10 mm. The diffusion gradients varied from 0 to 38 mT mÿ1 in the z direction; the duration of each lobe was 20 ms and their separation was 6 ms (b=0 to 872 s mmÿ2). The acquisition time for each image was about 50 ms. Note that diffusion is the highest in ventricular cavities containing cerebrospinal ¯uid (free water). Because of diffusion anisotropy, diffusion along the x axis in corpus callosum is very low, while vertical frontal white matter tracts has the higher diffusion (see Table 1)
180°
,, ,,
rf
Slice select Readout
Phase encode
Figure 5 Diffusion echo planar imaging sequence. The spin echo is split into a series of gradient echoes by quickly switching the readout gradient. Sensitization to diffusion of an echo planar imaging sequence can be easily achieved by additional gradient pulses (dashed boxes)
For list of General Abbreviations see end-papers
With EPI, motion artifacts are virtually eliminated. The accuracy of assessments of diffusion achieved with EPI is generally extremely good, as many images differently sensitized to diffusion can be generated because of the very short acquisition time (typically less than 100 ms). The EPI technique is the method of choice for in vivo diffusion imaging, although it is very vulnerable to susceptibility artifacts, which are responsible for image distortion or signal dropout and to chemical shift artifacts, which require ef®cient fat suppression. Despite this, diffusion-EPI has become widely available on many clinical scanners and has been successfully used for measuring diffusion of water in the human brain in volunteers (Figure 4)37,41 and patients (see below). Recently, alternative fast acquisition schemes, such as BURST42,43 and spiral MRI,44 have been proposed to overcome some EPI artifacts, such as ghosting or susceptibility artifacts. 2.2.3
Echoes
, ,, ,
90°
,, ,,
However, echo planar imaging (EPI) remains generally the imaging method of choice.36±39 With EPI, the entire set of echoes needed to form an image is collected within a single acquisition period (single shot) of 25±100 ms. This is obtained by switching the echo signal formation in a train of gradient echoes by means of a large gradient in which polarity is very rapidly inverted as many times as is required to achieve the desired image resolution.40 EPI may easily be sensitized to diffusion (Figure 5).36±39 Sensitization consists of providing a pair of large compensated gradients for a period of time before rapid gradient switching and data acquisition. The refocusing may be achieved either by simply reversing the polarity of the gradient halfway through the period over which it is applied or by inserting a 180 rf refocusing pulse at the midpoint without reversing the gradient polarity.
Sequence Optimization
Although the b factor can be exactly determined for any pulse sequence,27,45 the diffusion time is impossible to de®ne accurately for a MR sequence where gradient pulses of ®nite duration are scattered all over the pulse sequence. A mixture of diffusion-related events with differently weighted contributions are obtained depending on their sequence in time with respect to the many pulse gradients. That is why it has been suggested that the parameter derived from the NMR signal attenuation using Equation (7) should be called an `apparent diffusion coef®cient' (ADC).28 MR sequences should be designed to minimize these problems, i.e., by placing gradient pulses to get as close as possible to the simple Stejskal±Tanner bipolar pulse sequence. However, the use of the ADC term remains justi®ed,
METHODS AND APPLICATIONS OF DIFFUSION MRI
as biological tissues usually differ signi®cantly from the ideal `free, unlimited, isotropic' medium. Another issue is to optimize the values and the number of b factors to be used. A simple calculation4 shows that the best accuracy for the diffusion coef®cient is obtained from a set of two acquisitions if the two b factors, b1 and b2 differ by about 1/D.46 In the brain, this translates to (b2ÿb1)~1000 to 1500 s mmÿ2. If more than two acquisitions are to be used, error propagation theory47 shows that it is better to accumulate n1 and n2 measurements at each of the small and large b factors, b1 and b2, than to use a range of b factors. The accuracy dD/D can be obtained from the raw image signal-to-noise ratio (SNR): dD=D 1=n1 exp2D
b1 ÿ b2 =n2 1=2 =
SNRD
b1 ÿ b2
12 Acquisition performed with linear sets of b value ranges remains, however, extremely useful to validate the sequence/ hardware quality and to assess the nature of the diffusion process or the presence of multiple diffusion compartments as it allows the exponential diffusion attenuation to be visualized according to Equation (11).
2.3 Experimental Considerations in Diffusion MRI 2.3.1
Gradient Hardware
Since the minimum length of molecular diffusion paths detectable with gradient-pulsed NMR is primarily determined by the intensity of the gradient pulses, hardware must be capable of providing stable gradients of the utmost intensity.48 This requirement may be extremely challenging when considering whole-body instruments designed for clinical studies. The lack of gradient power is usually compensated by using somewhat long gradient pulse widths; as a result, the classical condition << is not satis®ed. The case of such ®nite pulse widths is signi®cantly more dif®cult to treat and to interpret compared with the situation with -sharp pulses. If the duration of the pulses is ignored, signi®cant underestimation of the diffusion distances may result.49 For instance, to achieve at least a 20% signal attenuation and limit signal loss from transverse relaxation, the minimum diffusion coef®cient Dmin measurable is 3710ÿ3 mm2 sÿ1 using =/10, =T2=100 ms, and Gmax=10 mT mÿ1, which is more than 16 times the diffusion coef®cient of free water!14 Maximum sensitivity is achieved when = (equivalent to a constant gradient) and =T2/2. Then Dmin is 0.410ÿ3 mm2 sÿ1. There are partial solutions to this problem, such as using stimulated echoes to increase the effective diffusion time without penalizing the signal by T2 relaxation effects,13 or, better, by building high-performance gradient coils (e.g., up to 50, 100 or even 200 mT mÿ1). Safety should not be too much a concern, considering that slew rates will remain low, as there is no need to switch such gradient coils very rapidly (which also limits eddy currents). Hence, diffusion effects during the application of the gradient pulses can no longer be neglected, although their contribution to the overall signal attenuation does not scale the same way with time as
5
for diffusion occurring between pulses. The diffusion time then becomes more dif®cult to de®ne, although (ÿ/3) is often taken instead of as the ef®cient diffusion time.11 Another concern is that any mismatch between the diffusion-sensitizing gradient pulses may cause artifactual signal losses through an improper spin rephasing. Two major sources of problem are commonly seen. One is gradient instability, which may arise when gradient ampli®ers are driven hard for fast switching of large gradient intensities. Variations from shot to shot of the bipolar gradient balance result in widely distributed ghost artifacts. While such artifacts do not occur when a single-shot technique such as EPI is used, it is still essential to have high-quality gradient ampli®ers with some reserve capacity for accurate diffusion imaging. The second problem results from eddy currents generated mainly in the cryostat when switching large gradient pulses rapidly. Eddy currents may also be a major cause of image distortion and misregistration between images obtained with different b values, leading to ADC miscalculation. The best solution to this problem is undoubtedly to remove eddy currents at the source by using actively shielded gradient coils, which have no fringe ®elds and, therefore, do not generate eddy currents. The use of a gradient coil of small dimensions, which remains at a fair distance from the magnet core and with which it is easier to generate large gradient amplitudes, is certainly an attracting alternative.21 If residual eddy current effects persist, other approaches can be combined.50±52 In summary, there is clearly a need for very-high-performance gradient coils. With the hardware that is now becoming available,53,54 shorter diffusion times or larger b values can be reached while maintaining high SNR, giving access to slow diffusion components (e.g., intracellular water, metabolites).
2.3.2
Motion Artifacts
As the sequences used are deliberately sensitized to motion by the addition of large gradients, a major problem occurring with in vivo imaging of diffusion arises from motion of the object. Artifacts result from discontinuities that occur between successive cycles of an imaging sequence. Results of such temporal incoherence are commonly visible as `ghosts' along the phase-encoding direction. These ghosts are particularly intense in the presence of the diffusion gradients and render the diffusion measurements meaningless. Cardiac gating has been used to mitigate this problem, but even this motion is not strictly cyclic. It is dif®cult to compensate diffusion imaging sequences for motion, since the use of successive bipolar gradient pulses considerably reduces the value of the b factor.4,14 Chenevert et al. have suggested eliminating any phase encoding from the acquisition by sacri®cing one dimension of the image, which is then reduced to a single line.55,56 Ultimately, however, the best way to avoid motion artifacts is to use a single-shot technique, such as EPI, and to secure patients comfortably within the magnet using cushions or in¯ating devices. In spite of this, motion artifacts may persist, especially when multiple shots are used, as with diffusion spectroscopy (see below) or segmented EPI. Motion effects can be estimated by monitoring the phase of the signal and corrected.57,58 Navigator echoes have been used with some success by several investigators.59±62 For References see p. 16
6 METHODS AND APPLICATIONS OF DIFFUSION MRI 2.3.3
Background Inhomogeneities
In addition to the magnetic ®eld gradient pulses, residual ®eld inhomogeneities arising from the imperfect shimming of the magnet and from inhomogeneities in the sample must be considered. These inhomogeneities result from susceptibility variations, especially at high ®eld, and may not be negligible.63 By adding a constant, linear gradient G0, it is clear that two contributions of the residual gradients coexist.11 One is a unique contribution (term in G20 ); the other results from a cross-term between the residual and the applied gradients (GG0 term).14 While the ®rst contribution can easily be eliminated from a series study where only G is varied, the second term is more dif®cult to handle. There are however, diffusion pulse sequences that have been designed to reduce signi®cantly or to eliminate effects of background gradients.64±70 Most of them are based on multiple rf pulses. For instance, cross-terms can be completely eliminated by alternating the gradient pulses between the successive 180 rf pulses of a quadrupole spin echo sequence.71 If cross-terms with the background gradients are not eliminated, the logarithm of A is no longer linear with the b factor and the ADC overestimates the true diffusion coef®cient.25 The ADC will, furthermore, appear as a function of the diffusion time. The examination of the linearity of the relationship between log A and b is therefore, the ®rst step in interpreting any NMR diffusion study in biological tissues, although there are several possible causes that would explain a deviation from linearity. Notably, similar nonlinear behavior will be seen if the b factor is miscalculated, i.e., cross-terms coming from all applied (diffusion and imaging) gradient pulses are not appropriately considered in the calculation.27 The b factor must be carefully calculated before the presence of background gradients can be assessed. However, ®nding a linear plot for log A versus b does not exclude background gradients. First, the effect may be small. Second, the interpretation of the data may become very complicated when gradients are nonuniform or nonlinear and slowly varying in space, as in the case of capillaries ®lled with materials with a high susceptibility constant (e.g. contrast agent). Simulations and experiments have shown, in this case, that the ADC is decreased and that the parameter to consider is the variance of the internal gradients.72,73 The ultimate test to evaluate background gradients, however, is to vary TE while and are kept constant. The measured ADC should normally be independent of TE. Otherwise, the contribution of background gradients must be suspected. Furthermore, as self-induced gradients may depend on the orientation of the sample with respect to the direction of B0, they may cause isotropic diffusion to appear anisotropic.65 In this case also, the measured diffusion coef®cient depends on the diffusion time. Once it is established that background gradients are present, it may be desirable to estimate them from the attenuation behavior in order to obtain information on the structure of the medium. For instance, in the ideal case where the internal gradients result from inhomogeneities caused by the presence of closely packed particles, a rough estimation of the mean size R, of these particles can be obtained from:74 R 60 B0 =G0 For list of General Abbreviations see end-papers
13
where 0 is the magnetic susceptibility constant. Furthermore, these susceptibility effects may be of biological interest,75 as variations in the oxygenation level in brain capillaries and small vessels following brain activity have been shown to induce susceptibility effects measurable by T2 .76,77 A signi®cant drop of the measured ADC has also been observed during status epilepticus,78 cortical electroshocks79 and spreading depression.80±82 In all cases, large increases in oxygenated blood ¯ow could result in important reductions in local internal gradients and apparent variations to diffusion, although cytotoxic edema and cell swelling associated with energy failure has been mainly evoked as the reason for this diffusion drop. 2.4 2.4.1
Other Approaches to Diffusion Diffusion Tensor Imaging
Diffusion is a three-dimensional process. However, the molecular mobility may not be the same in all directions. This anisotropy may be a result of the physical arrangement of the medium (liquid crystal) or the presence of obstacles that limit diffusion in some directions. The result is that the ADC appears different when pulse gradients are imposed in different directions; this has been observed in muscle83 and, more recently, in brain white matter.37,56,84 The proper way to study anisotropic diffusion is to consider the diffusion tensor.1,11 Diffusion is no longer characterized by a single scalar coef®cient but by a tensor D, which fully describes molecular mobility along each axis and the correlation between these axes: Dxx D Dyx Dzx
Dxy Dyy Dzy
Dxz Dyz Dzz
14
In the reference frame [x', y', z'] that coincides with the principal or main directions of diffusivity, the off-diagonal terms do not exist and the tensor is reduced only to its diagonal terms Dx'x', Dy'y', Dz'z', which represent molecular mobility along axes x', y', and z', respectively. The echo attenuation then becomes: A exp
ÿbx0 x0 Dx0 x0 ÿ by0 y0 Dy0 y0 ÿ bz0 z0 Dz0 z0
15
where bii are the elements of the b matrix (which now replaces the b factor) expressed in the coordinates of this reference frame. In practice, measurements are made in the reference frame [x, y, z] of the gradients, which usually does not coincide with that of the tissue. Therefore, one must also consider the coupling of the nondiagonal elements bij of the b matrix with the nondiagonal terms Dij (i6j ), of the diffusion tensor (now expressed in the gradient frame), which re¯ect correlation between molecular displacements in perpendicular directions:85 ! X X A exp ÿ bij Dij
16 ix;y;z jx;y;z
Calculation of the value of b may quickly become very complicated when many gradient pulses are used,45 however, the full
METHODS AND APPLICATIONS OF DIFFUSION MRI
determination of the diffusion tensor is necessary to assess anisotropic diffusion properly and fully. To determine the diffusion tensor, DW images are collected along several gradient directions. As the diffusion tensor must be symmetric, measurements along only six directions are mandatory (instead of nine), along with an image acquired without diffusion sensitivity (b=0). A typical set of gradient combinations that preserves uniform space sampling and similar b values along each direction is as follows (coef®cients for gradient pulses along the x, y, z axes, normalized to a given amplitude, G): p p p p p p
1= 2; 0;1= 2;
ÿ1= 2; 0;1= 2;
0; 1= 2;1= 2; p p p p p p
0; 1= 2; ÿ1= 2;
1= 2;1= 2; 0;
ÿ1= 2;1 2; 0 This minimal set of images may be repeated for averaging, as the SNR may be low12,85 In the case of `axial symmetry', only four directions are necessary (tetrahedral encoding),86 as suggested for the spine.87 The acquisition time and the number of images to process is then reduced, but efforts should be made to collect data along as many directions in space as possible to avoid sampling direction biases for applications such as the determination of ®ber orientation and gain in SNR (see below).88 A second step is to estimate the Dij values from the set of diffusion/orientation-weighted images, generally by multiple linear regression, using Equation (23) below. The ®nal step is to determine the main direction of diffusivities in each voxel and the diffusion values associated with these directions. This is equivalent to determine the reference frame [x', y', z'] where the off-diagonal terms of D are null. This `diagonalization' of the diffusion tensor provides `eigen-vectors' and `eigen-values', , which correspond, respectively, to the main diffusion directions and associated diffusivities,12,85 (the eigendiffusivities are not to be confused with the tortuosity factors; see below). As it is dif®cult to display tensor data with images (multiple images would be necessary), the use of `diffusion ellipsoids' has been proposed.12,85 Ellipsoids are a three-dimensional representation of the diffusion distance covered in space by molecules in a given diffusion time (Td). These ellipsoids, which can be displayed for each voxel of the image, are easily calculated from the eigen-diffusivities 1, 2, and 3 (corresponding to Dx'x', Dy'y', and Dz'z'): x02 =
21 Td y02 =
22 Td z02 =
23 Td 1
17
where x', y', and z' refer to the frame of the main diffusion direction of the tensor. These eigen-diffusivities represent the unidimensional diffusion coef®cients in the main directions of diffusivity of the medium. The main axis of the ellipsoid gives the main diffusion direction in the voxel (coinciding with the direction of the ®bers), while the eccentricity of the ellipsoid provides information about the degree of anisotropy and its symmetry (isotropic diffusion would be seen as a sphere). The length of the ellipsoids in any direction in space is given by the diffusion distance covered in this direction. In other words, the ellipsoid can also be seen as the three-dimensional surface of constant mean squared displacement of the diffusing molecules.
7
Recent work has demonstrated the feasibility of diffusion tensor imaging (DTI) in animal models, as well as in the human brain.41,86,89,90
2.4.2
The Diffusion Propagator and q-space Imaging
In the case where diffusion deviates from the Gaussian model, as with restricted diffusion (see below), it becomes extremely dif®cult to get any meaningful information on molecular displacements from the measured NMR signals. The distribution of the propagator is no longer Gaussian; consequently the value obtained for the diffusion coef®cient using Equation (11) may not necessarily properly re¯ect tissue structure of properties. Diffusion coef®cients may become, in these conditions, meaningless. The main dif®culty is that the medium structure is generally not known in detail and modeling is required. In particular, one must be cautious in the use of Fick's Law and derived relations. The probability distribution driven by the diffusion process (Equation (6)] may now signi®cantly deviate from a Gaussian distribution; as a result the relationship between A and b is no longer exponential as would be expected from Equation (7). A successful analysis should, therefore, start from a reformulation of the probability distribution, taking into account the medium structure and particular boundary or limit conditions. Hence, more useful information may be obtained by directly inferring molecular displacements from the MR signal attenuation without going through a diffusion coef®cient formulation. Molecular displacement pro®les, or more exactly their probability density distribution, can be obtained using q-space imaging.91,92 This does not differ, in principle, from diffusion imaging. Assuming P(z1) is homogeneous throughout the medium, Equation (4) can be rewritten by substituting the relative diffusion displacement z for (z2ÿz1) and by calling q the quantity ( G) as:
A
q; expiqzP
z; d z
18 It now appears that P(z,) is just the Fourier transform of the attenuation A(q,). By inverting Equation (15), direct and unequivocal data on molecular displacements can be obtained. In practice, data are collected for a series of different gradient strengths and an inverse Fourier transform is applied to the corresponding set of signal attenuation values. The experiment can be repeated with different values of in order to characterize the probability distribution for different diffusion times.93 With q-space imaging, a more comprehensive picture of the system under study is obtained than when using a unique ADC.92 It then becomes straightforward to verify whether P(z,) is a Gaussian distribution and to determine the value of the diffusion time when the value of P(z,) may start to deviate from a Gaussian behavior, suggesting that molecules have reached another structural order in the medium, or to obtain the molecular displacement distribution in different conditions. Localized q-space imaging has been used, for instance, to obtain water-displacement pro®les from normal and ischemic mouse brain, showing an increase of the fraction of molecules undergoing displacements of less than 10 m.94 Such ®ndings are very useful for examining the microstructural changes occuring during acute brain ischemia (see below). For References see p. 16
8 METHODS AND APPLICATIONS OF DIFFUSION MRI 2.4.3
Diffusion Spectroscopy
Water may not be the most suitable molecule for diffusion measurements in biological tissues because of its ubiquity and the permeability of most tissue interfaces, such as membranes, to water molecules. Consequently, diffusion measurements of larger molecules that are more speci®c to a tissue or compartment appear to be more promising for tissue characterization. However, such measurements are technically more challenging, because of the low concentration of such molecules and their relatively low diffusion coef®cients. Further, this requirement for signal averaging, and, therefore, long acquisition times, makes in vivo diffusion spectroscopy particularly sensitive to motion.57,95 Recent progress made in in vivo Fourier NMR spectroscopy allows the extension of diffusion measurements to molecules other than water. For a given species, the chemical shift can be used to determine independent diffusion coef®cients of compounds in complex mixtures.96 Diffusion of phosphocreatine, for instance can be studied by 31P spectroscopy.97.98 Phosphocreatine is a true intracellular space probe, in contrast to water, which diffuses across cell membranes; as a result it can be used to observe true restricted diffusion. Phosphocreatine may be used to provide unique information on the intracellular medium, such as viscosity or geometry. Similarly, diffusion of Nacetylaspartate (NAR) and myo-inositol may provide useful information on neuronal and glial cell populations, respectively. Monitoring of exchanges of metabolites or drugs through cell membranes could also bene®t from diffusion ®lter techniques designed to separate intracellular from extracellular compounds.99 Although diffusion coef®cients of metabolites are low because of their size, for nuclear species with spin >1/2 or for coupled-spin systems, multiple quantum experiments are less demanding on gradient hardware because the effective gradient amplitudes are increased by the power of the coherence order (n). With n=2, a fourfold increase in the diffusion effect would occur. Such spectral editing techniques have been used for selective measurement of diffusion properties of J-coupled compounds such as lactate.100,101 3 DIFFUSION IN BIOLOGICAL SYSTEMS: EFFECTS OF MICRODYNAMICS AND MICROSTRUCTURE Signi®cant technical progress has been achieved in biological systems to ensure that meaningful data can be acquired with minimal artifacts. However, even in the ideal case with perfect data, the interpretation of diffusion NMR experiments performed in biological tissues remains challenging. The diffusion coef®cient of water in tissues has been found to be two to ten times less than that of pure water (Table 1).14 This can be largely understood by considering that water molecules are obliged to divert tortuously around obstructions presented by ®bers, intracellular organelles, or macromolecules. Water molecules may be con®ned in bounded compartments or retained by attractive centers or surfaces. Biological systems, therefore, differ greatly from an `in®nitely large medium'. They are very heterogeneous and made of multiple subcompartments (microstructure). Depending on the permeability of the barriers that limit these compartments, exchanges and transport between them (microdynamics) may occur. A classic treatment of the NMR signal may not then re¯ect properly tissue structure or For list of General Abbreviations see end-papers
Table 1 Diffusion Coef®cients of Water in Human Brain Mean diffusivity Anisotropy (10ÿ3 mm sÿ1) (1ÿvolume ratio) Cerebrospinal ¯uid Gray matter (frontal cortex) Caudate nucleus White matter Pyramidal tract Corpus callosum (splenium) Internal capsule Centrum semiovale
3.19 0.10 0.83 0.05 0.67 0.02
0.02 0.01 0.08 0.05 0.08 0.03
0.71 0.04 0.69 0.05 0.64 0.03 0.65 0.02
0.93 0.04 0.86 0.05 0.70 0.08 0.27 0.03
Measurements (mean SD) were obtained in normal volunteers using diffusion tensor MRI. (With permission from Pierpaoli et al.41)
properties. Water molecules will `sense' all these obstacles only if the time over which diffusion is measured is made suf®ciently long that signi®cant interaction of the water molecules with cellular compartments may occur. The concept of `diffusion time' is, therefore, central to any diffusion study in biological tissues. While it may appear that using long diffusion times may reveal more tissue interactions, it is perhaps even more interesting to make the diffusion time as short as possible, up to the point where the diffusion coef®cient will reach its value in pure water. By doing so, and plotting measured diffusion coef®cients versus diffusion times, the many mechanisms contributing to in vivo diffusion may be revealed and identi®ed. This approach, which will bene®t from the newly developed gradient hardware allowing very large gradient strength, may be compared in some ways to that used in NMR dispersion studies, where multiple elementary relaxation mechanisms can be identi®ed and separated by studying dispersion curves.102 Diffusion coef®cients may become, in these conditions, meaningless if the measurement time scale or the measurement direction are not provided. The main dif®culty is that the medium structure is generally unknown in detail. The issue is, therefore, to infer meaningful information on tissue from the measured NMR signals. Initially, this problem could be reversed to consider how known tissue features relevant to tissue microstructure and dynamics affect diffusion NMR signals. Diffusion NMR has been used to study ¯uid-®lled porous media and derive information on microstructure (shape of the pore space) and ¯uid permeability (porosity).103±105 It is clear that many of the questions addressed in those studies are relevant to diffusion measurements in biological media.5 Differences in diffusion coef®cient and available diffusion space can be used to distinguish compartments and exchanges between them.99 3.1
Effect of Temperature
The ®rst obvious effect on diffusion is that of temperature, as diffusion directly results from the molecular thermal motion. The diffusion sensitivity to temperature is high, about 2.4% for 1 C change (Figure 6),106,107 consequently temperature must be carefully controlled in diffusion MRI experiments. Based on the strong and unique relationship that exists between temperature and molecular diffusion, diffusion imaging has been proposed for the real-time and noninvasive monitoring of tem-
METHODS AND APPLICATIONS OF DIFFUSION MRI
Temperature (MRI) (°C)
37
32
T1 Diffusion
27
22 22
27 32 Temperature (Luxtron) (°C)
37
Figure 6 Comparison of temperature measurements using T1 and diffusion MRI, and using probes (Luxtron) in a phantom. The accuracy was found to be 0.2 C with diffusion and 0.5 C with T1
perature. Noninvasive and nondestructive temperature imaging in biological systems may be particularly useful to monitor hyperthermia treatments in real-time, whether using rf electrical ®elds108 or focused ultrasound.109,110 It could also be used to study and control tissue interactions in surgical and medical laser procedures. However, in the context of this chapter, the diffusion±temperature relationship also provides some indication about diffusion mechanisms in tissues. This relationship has been established semi-empirically in liquids1 and veri®ed experimentally.106,111 D D0 exp
ÿEa =kT
19
where k is the Boltzmann constant, T is the temperature, D0 is the diffusion that would be obtained at an in®nite temperature, and Ea is introduced as the translational diffusion activation energy, which is approximately equal to the energy required to break hydrogen bonds, both in pure water and in vivo (Ea=0.2 eV).107,111 This important result can be understood by considering that the mechanism at the molecular scale for water to move, and thus diffuse, involves continuous breaking and reforming of hydrogen bonds. The fact that Ea is identical for water molecules diffusing in vitro and in vivo is not really surprising as the laws of physics at this level should not be different. 3.2 Restricted Diffusion Diffusion is restricted when boundaries in the medium prevent molecules from moving freely.112,113 Restriction must be related to the experimental parameters. When measurement times are very short, most molecules do not have enough time to reach boundaries and so they behave as if diffusing freely. Once the diffusion time increases, an increasing fraction of molecules will strike the boundaries, and diffusion will deviate
9
from the free, Gaussian behavior. Hence, the usual way to check for restricted diffusion is to plot the diffusion distance, calculated as for free diffusion using Einstein's equation, as a function of the square root of the diffusion time. In the case of restricted diffusion, this plot shows a curvature and ®nally a leveling off when the diffusion distance reaches the size of the restricting compartment. The effects of restriction will, therefore, appear in the NMR signal for diffusion times such that molecular displacements are in the order of the size of the restricting volumes. These effects will depend on the type of restriction (impermeable or permeable barriers, attractive centers, etc.), the shape of the restricting volumes (spherical, cylindrical, parallel walls, etc.), and the type of NMR experiment (constant or pulsed gradients). As a result, there is not a unique analytical expression that could describe any con®guration. A simple example is represented by molecules diffusing between two impermeable parallel walls separated by a distance a.112 If the theoretical, free diffusion distance greatly exceeds a, the echo attenuation A in the case of the bipolar gradient pulse experiment signi®cantly deviates from an exponential decay and becomes independent of the diffusion time,112 implying that molecules are trapped in the direction of the applied gradient: A sin
G a=2=
G a=22
20
Another interesting, but somewhat more complicated, case is represented by diffusion restricted in a spherical cavity of radius R0. In the limit where the theoretical, free diffusion distance largely exceeds R0, the attenuation is again independent of the diffusion time112,114 and the measured ADC decreases when the diffusion time is increased: A expÿ
G2 R20 =5
21
corresponding to an asymptotic ADC value, Dasymp of R20 /5. The factor 5 would be replaced by 3 if diffusion was con®ned to the surface of the sphere. Whatever the geometry of the restrictive medium, the deviation from linearity in the semi-log plot of the signal attenuation versus b is crucial to determine whether diffusion is restricted, although other causes may be responsible, such as diffusion in inhomogeneous systems or anisotropic diffusion. The ultimate test is to show that the measured diffusion coef®cient or the signal attenuation varies when the diffusion time is changed. Such studies can theoretically lead to the determination of the geometry and size of the restricting boundaries, as with the q-space concept. To avoid restricted diffusion effects, the diffusion time must be decreased to ensure that the diffusion distance during that period remains less than the size R of the restricted region. Unfortunately, the diffusion effect in these conditions becomes small unless considerable gradient intensities are used. A further complication is that in biological tissues walls may not be re¯ecting boundaries but rather occur as partially absorbing borders.115 3.3
Permeable Barriers
When the restrictive barriers become permeable to diffusing molecules, the restricted diffusion pattern changes. The mathematical treatment of diffusion in systems partitioned by For References see p. 16
10 METHODS AND APPLICATIONS OF DIFFUSION MRI permeable barriers is far from simple. An example was given by Tanner for equally spaced, plane barriers having a permeability constant .116 For short diffusion times, the ADC is D0. When the diffusion time increases, the ADC decreases, as expected for restricted diffusion, but saturates at Dasymp, which depends on the permeability constant: Dasymp D0 =
1 D0 =a
22
where a is the barrier spacing. This spacing can be estimated by the equivalent free diffusion distance that would be obtained from Einstein's equation with D=(D0+Dasymp)/2 as the diffusion coef®cient and Td1=2 , the corresponding diffusion time.117 The plot of D versus Td then shows a typical sigmoid pattern and it becomes possible to estimate the barrier permeability by ®tting the data with Equation (22).118 This approach is, however, very optimistic, as the geometrical arrangement of the medium is generally not known. In particular, this formalism does not apply to the case where the system consists of spherical cavities separated by permeable barriers.
3.4
Hindered Diffusion
True `restricted' diffusion, which occurs in bounded media and leads to a decrease of diffusion with the diffusion time should be distinguished from `hindered' diffusion. With hindered diffusion, diffusion is decreased by the presence of obstacles but there is no limit to the diffusion distance; consequently diffusion does not change with the diffusion time (once the hindered diffusion regime has been reached). Perhaps the most powerful concept associated with hindered diffusion is that of `tortuosity', a concept that has been widely used in solid porous media studies and more recently in brain diffusion experiments using external tracers.119±121 The idea is that because of the presence of obstacles such as ®bers, macromolecules, and organelles water molecules must travel longer paths to cover any given distance. In other words, molecules can no longer go straight from A to B, but must diffuse around structures that are impermeable to them (Figure 7). This situation results in a longer diffusion time to diffuse from A to B, or to an apparent decrease in the diffusion distance covered in a given diffusion time and in the measured ADC. This `hindered'
Figure 7 Restricted and hindered diffusion in white matter. Several models have been suggested to explain white matter diffusion anisotropy. First, diffusion may simply be totally restricted in axons (diameter, d). Plots of diffusion distance versus diffusion time (as in Figure 8, below) however, do not support this hypothesis. One may also take into account permeability of membranes and myelin to water. Going from A to B, water molecules would have to cross several interfaces. However, one may argue that those results may also be compatible with a `tortuosity' or `hindrance' model, where water molecules going from A to B would have to move around ®bers. As there is no limit to diffusion, this model is totally compatible with Figure 8 if extracellular water diffusion predominates
For list of General Abbreviations see end-papers
METHODS AND APPLICATIONS OF DIFFUSION MRI
11
14 Diffusion distance (mm)
12 10 8 6 Gray matter White matter, x White matter, z
4 2 0 0
2
4
6
8
10
Td (ms1/2)
Figure 8 Plot of the diffusion distance against diffusion time in gray and white matter. The diffusion distance was obtained from the diffusion coef®cient using Einstein's relation. There is no leveling off of the diffusion distance, suggesting that water diffusion is not completely restricted
diffusion effect is classically expressed quantitatively using a `tortuosity' coef®cient, , such that: ADC D=2
Figure 9 Diffusion image showing anisotropic diffusion in white matter. Excellent contrast is achieved, although T1 and T2 effects have been removed. Corpus callosum and temporal white matter ®bers, which are horizontal, are dark. Vertical corona radiata frontal ®bers and internal capsule ®bers are bright. In brainstem, vertical fast-conducting motor and somatosensory tracts are also bright
23
where D would be the diffusion coef®cient observed in the absence of obstacles. Furthermore, as there is no real barrier, molecules can, in principle, diffuse over very large distances compared with those seen in restricted diffusion. Therefore, no curvature would be seen in the plot of the diffusion distance versus the square root of the diffusion time. Similarly, the measured ADC would not depend on the diffusion time, unless the diffusion time is very short, and hindered diffusion paths would not differ signi®cantly from free diffusion paths (Figure 8). It would then be dif®cult to distinguish between hindered diffusion and diffusion restricted by permeable barriers. 3.5 Anisotropic Diffusion Diffusion in tissues may be different for different directions of motion; this anisotropy will give rise to variations in the measured diffusion coef®cient with the direction of measurement. Diffusion anisotropy has been observed in muscle,83 in brain white matter,37,56,84 (Figure 9) and, recently, in gray matter.122,123 It may result from restriction of diffusion inside ®bers (intracellular water) or from an increased tortuosity when diffusion occurs around ®bers (extracellular water). Diffusion can be both anisotropic and unrestricted. This behavior is well known in nematic liquid crystals124 and can be found in the water lamellar phase of amphiphilic lyotropic systems.125 Using DTI, it appears that diffusion data can be analyzed in three ways to provide information on tissue microstructure and architecture for each voxel or region of interest:126 the mean diffusivity, which characterizes the overall mean-squared displacement of molecules (average ellipsoid size) and the overall presence of obstacles to diffusion; the degree of anisotropy, which describes how much molecular displacements vary in space (ellipsoid eccentricity) and is related to the presence of
oriented structures; and the main direction of diffusivities (main ellipsoid axes), which is linked to the orientation in space of the structures. These three DTI `meta-parameters' can all be derived from complete knowledge of the diffusion tensor. However because of the complexity of data acquisition and processing for full DTI, and the sensitivity to noise of the determination of the diffusion tensor eigen values, simpli®ed approaches have been proposed. 3.5.1
Mean Diffusivity
To obtain an overall evaluation of the diffusion in a voxel or region, anisotropic diffusion effects must be avoided and the result limited to an `invariant', i.e., a quantity that is independent of the orientation of the reference frame.12,86 Among several combinations of the tensor elements, the trace of the diffusion tensor, Tr
D Dxx Dyy Dzz
24
is such an invariant. The mean diffusivity is then given by Tr(D)/3. A slightly different de®nition of the trace has proved useful in assessing the diffusion drop in brain ischemia127 (see below). Unfortunately, the correct estimation of Tr(D) still requires the complete determination of the diffusion tensor. Diffusion coef®cients obtained by separately acquiring data with gradient pulses added along the x, y, and z axes cannot be used as these measured coef®cients usually do not coincide with Dxx, Dyy, and Dzz, respectively. The reason is that the diffusion attenuation that results, for instance from inserting gradients on the x axis, is A=exp[ÿbxxDxx+2bxyDxy+2bxzDxz] and not simply exp[ÿbxxDxx] unless diffusion is isotropic (no nondiagonal terms) or there are no gradient pulses at all on the other axes (y and z, here) (no localization) during the diffusion measurement time. For References see p. 16
12 METHODS AND APPLICATIONS OF DIFFUSION MRI To avoid this problem and to simplify the approach, several groups have designed sequences based on multiple echoes or acquisitions with tetrahedral gradient con®gurations to cancel nondiagonal term contributions to the MRI signal directly.86,128±130 3.5.2
Diffusion Anisotropy Indices
Several scalar indices have been proposed to characterize diffusion anisotropy. Initially, simple indices calculated from DW images84 or ADC values obtained in perpendicular directions were used, such as ADCx/ADCy and displayed using a color,131 scale. Other groups have devised indices mixing measurements along the x, y, and z directions, such as maximum [ADCx, ADCy, ADCz]/minimum [ADCx, ADCy, ADCz] or the standard deviation of ADCx, ADCy, and ADCz divided by their mean value.127 Unfortunately, none of these indices is really objective as they do not correspond to a single meaningful physical parameter; more importantly, they are clearly dependent on the choice of directions made for the measurements. The degree of anisotropy would then vary upon the respective orientation of the gradient hardware and the tissue frames of reference and would generally be underestimated. Here again invariant indices must be found to avoid such biases and provide objective, intrinsic structural information.132 Invariant indices are made of combinations of the terms of the diagonalized diffusion tensor, i.e., the eigen-values 1,2, and 3. The most commonly used invariant indices are the relative anisotropy, RA, the fractional anisotropy, FA, and the volume ratio, VR: q p
1 ÿ hi2
2 ÿ hi2
3 ÿ hi2 = 3hi
RA
25
where hi
1 2 3 =3 FA
q 3
1 ÿ hi2
2 ÿ hi2
3 ÿ hi2 = q 2
21 22 23
VR 1 2 3 =hi3
26
27
RA, a normalized standard deviation, also represents the ratio of the anisotropic part of D to its isotropic part. FA measures the fraction of the `magnitude' of D that can be ascribed to anisotropic diffusion. p FA and RA vary between 0 (isotropic diffusion) and 1 ( 2 for RA) (in®nite anisotropy). As for VR, which represents the ratio of the ellipsoid volume to the volume of a sphere of radius hi, its range is from 1 (isotropic diffusion) to 0; as a result some authors prefer to use (1ÿVR).133 Once these indices have been de®ned, it is possible to evaluate them directly from DW images, i.e., without the need to calculate the diffusion tensor.134 For instance, A which is very similar to RA, has been proposed as:135 A
s X p
Dii ÿ hDi2
D2xy D2xz D2yz = 6hDi
28 ix;y;z
with hDi=
X
Dii/3.
ix;y;z
For list of General Abbreviations see end-papers
Also, images directly sensitive to anisotropy indices, or anisotropically weighted images, can be obtained.136 Finally, the concept of these intravoxel anisotropy indices can be extended to a family of intervoxel or `lattice' measures of diffusion anisotropy, which allows neighboring voxels to be considered together, in a region of interest, without losing anisotropy effects resulting from different ®ber orientations across voxels.133 Clinically relevant images of anisotropy indices have been obtained in the human brain (Table 1).41,137 3.5.3
Fiber Orientation Mapping
The last family of parameters that can be extracted from the DTI concept relates to the mapping of the orientation in space of tissue structure. The assumption is that the direction of the ®bers is colinear with the direction of the eigen-vector associated with the largest eigen-diffusivity. This approach opens a completely new way to obtain directly in vivo information on the organization in space of tissues such as muscle, myocardium, and brain or spine white matter, which is of considerable interest, clinically and functionally. Direction orientation can be derived from DTI directly from diffusion/orientationweighted images or through the calculation of the diffusion tensor. A ®rst issue is to display ®ber orientation on a voxelby-voxel basis. The use of color maps was ®rst suggested,131 followed by representation by ellipsoids,12,133 octahedra138 or vectors pointing in the ®ber direction.139,140 A second issue is to assess connectivity from DTI data in order to visualize anatomical connections between different parts of the brain on an individual basis. Studies of neuronal connectivity are tremendously important to the interpretation of functional MRI data and for establishing how activated foci are linked together through networks.141,142 This issue is dif®cult, as continuity of ®ber orientation from voxel to voxel has to be inferred. Fiber orientation may appear to be varying because of the occurrence of noise in the data. In a given voxel ®bers may be merging, branching or dividing. In addition, several fascicles may cross in a given voxel, which cannot be detected with the diffusiontensor approach in its present form. Structures that exhibit anisotropic diffusion at the molecular level can be isotropically oriented at the microscopic level, resulting in a `powder average' effect that is dif®cult to resolve.143 The semilog plot of the signal attenuation versus b may not be linear in this case.125 This deviation from linearity can be ascribed to anisotropy and not to restricted diffusion because the diffusion measurements are independent of the diffusion time. Several groups have recently approached the dif®cult problem of inferring connectivity from DTI data in the rat brain144 in vitro and in the living human brain145 (Figure 10). 3.6
Diffusion in Multiple Compartment Systems
Most diffusion measurements in biological tissues refer to an ADC and yet it is generally considered that diffusion in the measurement volume (voxel) has a unique diffusion coef®cient. This simpli®cation may not always be legitimate, since partial volume effects may occur and most tissues are made of multiple subcompartments (with at least intracellular and extracellular components). Assuming measurement times are short and diffusion is unrestricted in each subcompartment i, and that there is no exchange, the signal attenuation is:
METHODS AND APPLICATIONS OF DIFFUSION MRI
A
X
i exp
ÿbDi
29
i
where i is the density of molecules diffusing in compartment i, and Di the associated diffusion coef®cient. In this case, the ADC that would be measured would depend on the range used for the b values and would not re¯ect properly the diffusion in the voxel. Measurements with low b values would then be
13
more sensitive to fast diffusion components. The ideal approach would be to separate all subcompartments by ®tting the data with a multiexponential decay. Unfortunately, the values for Di are often low and not very different from each other; consequently very large b values and very high SNR would be required. Relaxation effects must also be considered if compartments have different relaxation rates.146 If all spins do not have the same transverse relaxation during the pulse sequence, Equation
Figure 10 Diffusion tensor imaging and ®ber orientation mapping in the brain of a normal volunteer at 1.5 T. The direction of the white matter ®bers can be determined from the eigen vectors of the diffusion tensor for each voxel and superimposed on a 3D rendering of internal brain structures. (Contribution of J.F. Mangin, C. Clark, C. Poupon, D. Le Bihan)
For References see p. 16
14 METHODS AND APPLICATIONS OF DIFFUSION MRI (5) must be complemented with a decay term:115 @P
z2 ; z1 t=@t Dr2 P
z2 ; z1 t ÿ P
z2 ; z1 ; t=T2
30 For two compartments, assuming measurement times are short and, therefore, exchanges are small, the diffusion signal is given by: S F1 exp
ÿbD1 F2 exp
ÿbD2
31
where b is the sequence diffusion factor, D1, D2 are the diffusion coef®cients and F1, F2 are relative weights to the signal of the two compartments. Estimates of F1, F2 and D1, D2 are obtained by ®tting a series of signals acquired with multiple b values. For a spin echo sequence with TR>>T1, F depends on the volume fraction f and the relaxation time T2 of each compartment: F f exp
ÿTE=T2
32
Two diffusion compartments have been found in the rat brain.147 Assignment of these compartments to extra- and intracellular water is not, however, straightforward, as the values found for F1 and F2 are apparently opposite to the known volume fractions for the intra- and extracellular spaces (82.5% and 17.5%, respectively). A ®rst explanation for this discrepancy is that T2 effects must be taken into account when converting F values into f values. A second factor is that the exchange rate may not be negligible. It has been shown that if the residence rate in each compartment is considered, the contribution of the fast diffusing component is overestimated.148 It is, therefore, not yet certain that the two compartments that have been observed do correspond to the intra- and extracellular compartments. The situation is even more complex when measurement times are longer. First, restricted diffusion may be seen in the smallest subcompartments. Second, molecular exchanges may occur between communicating compartments. As a result, the analytic treatment becomes dif®cult. Applying the central limit theorem for statistically distributed compartments in the case of long diffusion times, the use of a single apparent diffusion constant Da can be justi®ed: X i Di
33 Da i
This result is also consistent with NMR dispersion studies,102 which consider that cell membranes can be ignored on the NMR time scale. In intermediate situations, the geometrical arrangement and the diffusion coef®cients of each compartment must be considered as well as their rates of exchange.149 The comprehensive analysis of the diffusion attenuation curves obtained with different diffusion times may lead to an accurate description of the medium microstructure. It is clear that the ADC depends on the range used for the b values and would not re¯ect properly diffusion in the tissue. Measurements with low b values (less than 1000 s mmÿ2) would be more sensitive to fast diffusion components, such as those occurring in the extracellular compartment. In clinical studies and most animal experiments, especially when data are ®tted to a single exponential, diffusion patterns observed in tissues have to be explained by features of the extracellular For list of General Abbreviations see end-papers
space, although this compartment is physically small. Changes in the ADC calculated in these conditions should be interpreted in terms of changes in the diffusion coef®cient of the extracellular space (tortuosity) and in its fractional volume relative to the intracellular volume. This explains why diffusion has appeared as a sensitive marker of the changes in the extracellular/intracellular volume ratio, as observed in brain ischemia, spreading depression,80±82 status epilepticus,78 or extraphysiological manipulations of the cell size through osmotic agents.150 Using large b values, it becomes increasingly possible to assess separately the intra- and extracellular compartments and their relative volumes in vitro151 and in vivo, as attempted in the rat brain147 and the human brain.152
3.7
Metabolites
Data on magnitudes and even directional anisotropy of diffusion coef®cients of molecules such as choline, creatine/ phosphocreatine and NAA in animals,97,153,154 and human brain95 have been made available. Although diffusion rates of metabolites in pure water are lower than that of water because of the difference in molecular weight and hydration layers, the ADC values of these metabolites in brain are considerably smaller. Typical values from a series of 10 normal volunteers (18 cm3 voxels located in white matter, diffusion time 240 ms) are 0.1310ÿ3 mm2 sÿ1 for choline and creatine and 0.1810ÿ3 mm2 sÿ1 for NAA compared with 1.2410ÿ3 mm2 sÿ1 for choline and 0.8510ÿ3 mm2 sÿ1 for NAA at 35 C in vitro.155 The fact that metabolites, as well as molecules such as ¯uorodeoxyglucose 6-phosphate,156 with different molecular weights had very similar ADC values in vivo is striking and remains to be explained. The very low values of the ADC obtained with long diffusion times are, of course, compatible with restriction of these metabolites in compartments that were about the same size. For instance, recent experiments varying the diffusion times have shown that diffusion of NAA is considerably restricted in a manner compatible with two compartments, one of 7±8 m and one of approximately 1 m, possibly representing the cell bodies and the intra-axonal space.157 However, intracellular obstacles may also play an important role, as water molecules could diffuse easily in small spaces between small obstacles, such as mitochondria, organelles or macromolecules, while larger metabolites would `feel' a more obstructed medium. Tortuosity factors may, therefore, be larger for metabolites than for water and this would explain, at least partially, the diffusion decrease for intracellular metabolites that has been observed during ischemia.158,159
3.8
Summary: Application to Brain White Matter
The concepts of restriction, hindrance, tortuosity, and multiple compartments are particularly useful to understand diffusion ®ndings in brain white matter. Water diffusion is highly anisotropic in white matter37,56,84 and this anisotropy is observed even before ®bers are myelinated, though at a lesser degree.137,160±166 ADC values obtained by measurements made parallel and perpendicularly to the ®bers do not seem to depend on the diffusion time167,168 at least for diffusion times longer than 20 ms (see Figure 8).
METHODS AND APPLICATIONS OF DIFFUSION MRI
Initial reports suggested that the anisotropic water diffusion could be explained by restriction of the water molecules to the axons (anisotropically restricted diffusion) by the myelin sheath.169,170 However, although restricted diffusion has been seen for intra-axonal metabolites, such as NAA, or for truly intra-axonal water,171 it now appears that most studies were performed with relatively low b values and are mostly sensitive to the extracellular, interaxonal space. In this condition, diffusion anisotropy in white matter should be linked to the anisotropic tortuosity of the interstitial space between the ®bers: diffusion would be more impeded perpendicular to the ®bers because of their geometric arrangement (Figure 7).167 Considering a bundle where ®bers are organized in the most compact way, molecules would have to actually travel over a distance of d/2, where d is the ®ber diameter, for an apparent diffusion distance equal to the ®ber diameter. Therefore, on average, the ADC measured perpendicularly to the ®bers would be reduced, whatever the diffusion time, to: ADC
2=2 D0 0:4D0
34
where the reference value, D0, is the diffusion value measured parallel to the ®bers. This ratio of about 0.4 ®ts reasonably very well with literature data,167 although much larger ratios have been reported.133 This rough model also implies that the tortuosity factor would be anisotropic in white matter with a perpendicular/parallel ratio of /2 (&1.15). Unfortunately no systematic measurements of have yet been made in white matter using ionic extracellular tracers,2,119 although anisotropy has been seen.172 Another interesting point about this model is that it is compatible with the fact that no true restricted effects are observed when the diffusion time is increased, as there are no actual boundaries to diffusing molecules. Also, the parallel organization of the ®bers may be suf®cient to explain the presence of anisotropy before myelination. However, as the axonal membranes should be more permeable to water than the myelin sheaths, the degree of anisotropy should be less pronounced in the absence of myelin, as signi®cant exchanges with the axonal spaces should occur. Oriented ®laments within the axoplasm do not seem to play an important role.173 Fiber orientation mapping and connectivity studies derived from anisotropic diffusion in white matter will clearly bene®t from a better understanding of the respective contributions of intraaxonal and extra-axonal compartments to anisotropy mechanisms.174
4 CLINICAL APPLICATIONS Diffusion imaging is a truly quantitative method. The diffusion coef®cient is a physical parameter that directly re¯ects the physical properties of the tissues in terms of the random translational movement of the molecules under study (most often water molecules, but sometimes metabolites. The diffusion coef®cient does not depend on the ®eld strength of the magnet or the pulse sequence used, which is not the case for the other classical MRI parameters such as T1 or T2. Diffusion coef®cients obtained at different times in a given patient, or in different patients, or in different hospitals can be compared without any need for normalization.
4.1
15
Central Nervous System
The most clinically relevant ®eld of application of diffusion MRI is in the nervous system. Diffusion MRI has already appeared as a breakthrough in two areas: early brain ischemia and white matter diseases. During the acute stage of brain ischemia, water diffusion is decreased in the ischemic territory by as much as 50%, as shown in cat brain models.175 This diffusion slowdown is linked to the cytotoxic edema that results from the energetic failure of the cellular membrane Na+/K+ pumping system. The exact mechanism by which diffusion is reduced is still unclear (increase of the slow-diffusion intracellular volume fraction, changes in membrane permeability,176 or shrinkage of the extracellular space resulting in increased tortuosity for water molecules.177,178 have been suggested). Diffusion MRI has been extensively used in animal models to establish and test new therapeutic approaches. These results have been con®rmed in patients with stroke, offering the potential to highlight ischemic regions within the ®rst hours of the ischemic event, when brain tissue might still be salvageable,179,180 well before conventional MRI becomes abnormal (vasogenic edema). Combined with perfusion MRI, diffusion MRI is under clinical evaluation as a tool to help clinicians to optimize their therapeutic approach to individual patients,181 to monitor patient progress on an objective basis, and to predict clinical outcome.182±185 In white matter, DTI has already shown its potential in diseases such as multiple sclerosis186±188 leukoencephalopathy,189 Wallerian degeneration, Alzheimer disease,190,191 and Creutzfeld±Jacob disease.192±194 Mean diffusivity indices, such as the trace of the diffusion tensor, re¯ect overall water content, while anisotropy indices indicate myelin ®ber integrity. It has been shown that the degree of diffusion anisotropy in white matter increases during the myelination process,137,162,195 and diffusion MRI could be used to assess brain maturation in children,196 newborns, or premature babies.137,197 Abnormal connectivity in frontal white matter based on DTI data has also been reported in schizophrenic patients.198 The potential of diffusion MRI has also been studied in brain tumor grading,199±201 trauma,202 hypertensive hydrocephalus,203 AIDS,204 eclampsia,205 leukoaraiosis,206,207 and the spinal cord.87,185,208,209 4.2
Body
The use of diffusion imaging has been less successful in areas of the body apart from the brain because of the occurrence of strong respiratory motion artifacts and of the short T2 values of body tissues, which require shorter TE than in the brain and, therefore, leaves less room for the diffusion gradient pulses. These obstacles, however, can sometimes be overcome with ad hoc MR sequences and hardware. Potential for tissue characterization has been shown in the extremity muscles,210 the spine,211,212 the breast,54,213,214 the kidney, and the liver.215±218 Muscle ®ber orientation can be approached using DTI in organs such as the tongue219 or the heart. Myocardium DTI220,221 has tremendous potential for providing data on heart contractility, a very important parameter, but remains technically very challenging to perform in vivo because of heart motion. Other applications include temperature imaging For References see p. 16
16 METHODS AND APPLICATIONS OF DIFFUSION MRI through the sensitivity of diffusion coef®cients to temperature.107,108,210,222
5 CONCLUSION Many tissue features at the microscopic level may in¯uence NMR diffusion measurements. So far, many theoretical analyses of the effect of restriction, membrane permeability, hindrance, anisotropy, or tissue inhomogeneity have been published. These analyses underline how much care is necessary to conduct diffusion NMR studies properly and to interpret the results. Although the results of these analyses have been applied to characterize nonliving systems, such as porous media, much work remains to be done to produce accurate information on microstructure and microdynamics in vivo in biological systems. Powerful tools, such as diffusion spectroscopy of metabolites, DTI or q-space imaging, that are still under development are expected to provide such information.
6 RELATED ARTICLES Anisotropically Restricted Diffusion in MRI; Diffusion: Clinical Utility of MRI Studies; Ischemic Stroke; Male Pelvis Studies Using MRI.
7 REFERENCES 1. W. Jost, `Diffusion in Solids, Liquids and Gases', Academic Press, San Diego, 1960. 2. C. Nicholson, `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', ed. D. Le Bihan, Raven Press, New York, 1995, p. 127 3. O. B. Paulson, M. M. Hertz, T. G. Bolwig, and N. A. Lassen, Microvasc. Res., 1977, 13, 113. 4. D. Le Bihan (ed.), `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', Raven Press, New York, 1995. 5. P. T. Callaghan, `Principles of Nuclear Magnetic Resonance Microscopy', Oxford University Press, Oxford, 1991. 6. D. Le Bihan and E. Breton, C. R. Acad. Sci. Paris, Series II, 1985, T.301, 1109. 7. K. D. Merboldt, W. Hanicke, and J. Frahn, J. Magn. Reson., 1985, 64, 479. 8. D. G. Taylor and M. C. Bushell, Phys. Med. Biol., 1985, 30, 345. 9. E. L. Hahn, Phys. Rev., 1950, 80, 580. 10. H. Y. Carr and E. M. Purcell, Phys. Rev., 1954, 94, 630. 11. E. O. Stejskal and J. E. Tanner, J. Chem. Phys., 1965, 42, 288. 12. P. J. Basser, J. Mattiello, and D. Le Bihan, Biophys. J., 1994, 66, 259. 13. J. E. Tanner, J. Chem. Phys., 1970, 52, 2523. 14. D. Le Bihan, Magn. Reson. Q., 1991, 7, 1. 15. R. Kaiser, E. Bartholdi, and R. R. Ernst, J. Chem. Phys., 1974, 60, 2966. 16. D. Le Bihan, Magn. Reson. Med., 1988, 7, 346. 17. D. Le Bihan, R. Turner, and J. R. MacFall, Magn. Reson. Med., 1989, 10, 324. 18. K. D. Merboldt, W. Hanicke, M. L. Gyngell, J. Frahm, and H. Bruhn, J. Magn. Reson., 1989, 82, 115.
For list of General Abbreviations see end-papers
19. K. D. Merboldt, H. Bruhn, J. Frahm, M. L. Gyngell, W. Hanicke, and M. Deimling, Magn. Reson. Med., 1989, 9, 423. 20. E. X. Wu and R. B. Buxton, J. Magn. Reson., 1990, 90, 243. 21. D. Canet, B. Diter, A. Belmajdoub, J. Braondeau, J. C. Boubel, and K. Elbayed, J. Magn. Reson., 1989, 81, 1. 22. G. S. Karczmar, D. B. Twieg, T. J. Lawry, and M. W. Weiner, Magn. Reson. Med., 1988, 7, 111. 23. F. Humbert, M. Valtier, A. Retournard, and D. Canet, J. Magn. Reson., 1998, 134, 245. 24. F. Humbert, B. Diter, and D. Canet, J. Magn. Reson., 1996, A123, 242. 25. M. Neeman, J. P. Freyer, and L. O. Sillerud, J. Magn. Reson., 1990, 90, 303. 26. H. C. Torrey, Phys. Rev., 1965, 104, 563. 27. J. Mattiello, P. J. Basser, and D. Le Bihan, J. Magn. Reson., 1994, 108, 131. 28. D. Le Bihan, E. Breton, D. Lallemand, P. Grenier, and E. A. Cabanis, Radiology, 1986, 161, 401. 29. G. A. Ojemann, J. Neurosci., 1991, 11, 2281. 30. R. V. Mulkern and R. G. S. Spencer, Magn. Reson. Imag., 1988, 6, 623. 31. H. Gubjartsson, S. E. Maier, and F. A. Jolesz, Magn. Reson. Med., 1997, 38, 101. 32. S. E. Maier, H. Gudbjartsson, S. Patz, L. Hsu, K. O. LoÈvblad, R. R. Edelman, S. Warach, and F. A. Jolesz, Am. J. Roentgenol., 1998, 171, 85. 33. K. O. LoÈvblad, P. M. Jakob, Q. Chen, A. E. Baird, G. Schlaug, S. Warach, and R. R. Edelman, Am. J. Neuroradiol., 1998, 19, 201. 34. D. L. Thomas, G. S. Pell, M. F. Lythgoe, D. G. Gadian, and R. J. Ordidge, Magn. Reson. Med., 1998, 39, 950. 35. F. Schick, Magn. Reson. Med., 1997, 38, 638. 36. R. Turner and D. Le Bihan, J. Magn. Reson., 1990, 86, 445. 37. R. Turner, D. Le Bihan, J. Maier, R. Vavrek, L. K. Hedges, and J. Pekar, Radiology, 1990, 177, 407. 38. R. C. McKinstry, R. M. Weisskoff, M. S. Cohen, J. M. Vevea, K. K. Kwong, R. R. Rzedzian, T. J. Brady, and R. R. Rosen, Proc. IXth Annu. Mtg. Soc. Magn. Reson. Med., New York, 1990, p. 5. 39. H. R. Avram and L. E. Crooks, Proc. VIIth Annu. Mtg. Soc. Magn. Reson. Med., San Francisco, 1988, p. 80. 40. P. Mans®eld, J. Phys. C., 1977, 10, L55. 41. C. Pierpaoli, P. Jezzard, P. J. Basser, A. Barnett, and G. DiChiro, Radiology, 1996, 201, 637. 42. J. S. Doran, P. Jakob, and M. DeÂcorps, Magn. Reson. Med., 1996, 35, 547. 43. J. S. Doran and M. DeÂcorps, J. Magn. Reson., 1995, 117, 311. 44. L. Li, Magn. Reson. Med., 1999, 41, 143. 45. J. Mattiello, P. J. Basser, and D. Le Bihan, in `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', ed. D. Le Bihan, Raven Press, New York, 1995, p. 77. 46. J. H. Burdette, A. D. Elster, and P. E. Ricci, J. Comput. Assist. Tomogr., 1998, 22, 792. 47. D. Xing, N. G. Papadakis, C. L. Huang, V. M. Lee, T. A. Carpenter, and L. D. Hall, Magn. Reson. Imag., 1997, 15, 771. 48. R. M. Vavrek and J. R. MacFall, in `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', ed. D. Le Bihan, Raven Press, New York, 1995, p. 67. 49. P. P. Mitra and B. I. Halperin, J. Magn. Reson., 1995, 113, 94. 50. J. C. Haselgrove and J. R. Moore, Magn. Reson. Med., 1996, 36, 960. 51. P. Jezzard, A. S. Barnett, and C. Pierpaoli, Magn. Reson. Med., 1998, 39, 801. 52. F. Calamante, Magn. Reson. Med., 1999, 41, 95.
METHODS AND APPLICATIONS OF DIFFUSION MRI 53. M. R. Thompson, R. Venkatesan, K. Kuppusamy, A. Celik, W. L. Lin, D. K. Kido, and E. M. Haacke, Radiology, 1999, 210, 253. 54. C. F. Maier, H. N. Nikolov, K. C. Chu, B. A. Chronick, and B. K. Rutt, Magn. Reson. Med., 1998, 39, 392. 55. T. L. Chenevert and J. G. Pipe, Magn. Reson. Med., 1991, 19, 261. 56. T. L. Chenevert, J. A. Brunberg, and J. G. Pipe, Radiology, 1990, 177, 401. 57. S. Posse, C. A. Cuenod, and D. Le Bihan, J. Magn. Reson., 1993, 102, 222. 58. A. M. Ulug, P. B. Barker, and P. C. M. van Zijl, Magn. Reson. Med., 1995, 34, 476. 59. R. J. Ordidge, J. A. Helpern, Z. X. Qing, A. Knight, and V. Nagesh, Magn. Reson. Imag., 1994, 12, 455. 60. A. W. Anderson and J. C. Gore, Magn. Reson. Med., 1994, 32, 379. 61. A. J. de Crespigny, M. P. Marks, D. R. Enzmann, and M. E. Moseley, Magn. Reson. Med., 1995, 30, 720. 62. M. P. Marks, A. De Crespigny, D. Lentz, D. Enzmann, G. W. Albers, and M. E. Moseley, Radiology, 1996, 199, 403. 63. Z. H. Endre, B. E. Chapman, and P. W. Kuchel, Biochim. Biophys. Acta, 1984, 804, 137. 64. D. W. Williams, E. F. W. Seymour, and R. M. Cotts, J. Magn. Reson., 1978, 31, 271. 65. J. D. Trudeau, W. T. Dixon, and J. Hawkins, J. Magn. Reson., 1995, B108, 22. 66. R. F. Karlicek and I. J. Lowe, J. Magn. Reson., 1980, 37, 75. 67. W. D. Williams, E. F. W. Seymour, and R. M. Cotts, J. Magn. Reson., 1978, 31, 271. 68. M. I. Hrovat and C. G. Wade, J. Magn. Reson., 1981, 44, 62. 69. R. M. Cotts, M. J. R. Hoch, T. Sun, and J. T. Marker, J. Magn. Reson., 1989, 83, 252. 70. L. L. Latour, L. Li, and C. H. Sotak, J. Magn. Reson., 1993, 101, 72. 71. E. von Meerwall and M. Kamat, J. Magn. Reson., 1989, 83, 308. 72. R. P. Kennan, J. Zhong and J. C. Gore, Magn. Reson. Med., 1994, 31, 9. 73. R. P. Kennan, J. Zhong, and J. C. Gore, in `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', ed. D. Le Bihan, Raven Press, New York, 1995, p. 110. 74. J. S. Murdey and R. M. Cotts, J. Chem. Phys., 1968, 48, 4938. 75. S. Majumdar and J. C. Gore, J. Magn. Reson., 1988, 78, 41. 76. R. M. Weisskoff, C. S. Zuo, J. L. Boxerman, and B. R. Rosen, Magn. Reson. Med., 1994, 31, 601. 77. J. L. Boxerman, L. M. Hamberg, B. R. Rosen, and R. M. Weisskoff, Magn. Reson. Med., 1995, 34, 555. 78. J. Zhong, O. A. C. Petroff, J. W. Prichard, and J. C. Gore, Magn. Reson. Med., 1993, 30, 241. 79. J. Zhong, O. A. C. Petroff, L. A. Pleban, J. C. Gore, and J. W. Prichard, Magn. Reson. Med., 1997, 37, 1. 80. L. L. Latour, Y. Hasegawa, J. E. Formato, M. Fisher, and C. H. Sotak, Magn. Reson. Med., 1994, 32, 189. 81. Y. Hasegawa, L. L. Latour, J. E. Formato, C. H. Sotak, and M. Fisher, J. Cereb. Blood Flow Metab., 1995, 15, 179. 82. E. Busch, M. Hoehn-Berlage, M. Eis, M. L. Gyngell, and K.-A. Hossmann, NMR Biomed., 1995, 8, 59. 83. G. G. Cleveland, D. C. Chang, and C. F. Hazelwood, Biophys. J., 1976, 16, 1043. 84. M. E. Moseley, Y. Cohen, and J. Kucharczyk, Radiology, 1990, 176, 439. 85. P. J. Basser, J. Mattiello, and D. Le Bihan, J. Magn. Reson., 1994, 103, 247. 86. T. E. Contouro, R. C. McKinstry, E. Akbudakr, and B. H. Robinson, Magn. Reson. Med., 1995, 35, 399. 87. V. Gulani, G. A. Iwamoto, H. Jiang, J. S. Shimony, A. G. Webb, and P. C. Lauterbur, Magn. Reson. Med., 1997, 38, 868.
17
88. N. G. Papadakis, D. Xing, C. L. H. Huang, L. D. Hall, and T. A. Carpenter, J. Magn. Reson., 1999, 137, 67. 89. P. J. Basser, J. Mattiello, R. Turner, and D. Le Bihan, Proc. XIIth Annu. Mtg. Soc. Magn. Reson. Med., New York, 1993, p. 584. 90. T. L. Davis, V. J. Wedeen, R. M. Weisskoff, and B. R. Rose, Proc. XIIth Annu. Mtg. Soc. Magn. Reson. Med., New York, 1993, p. 289. 91. P. T. Callaghan, J. Magn. Reson., 1990, 88, 493. 92. D. G. Cory and A. N. Garroway, Magn. Reson. Med., 1990, 14, 435. 93. M. D. King, J. Houseman, S. A. Roussel, N. Van Bruggen, S. R. Williams, and D. G. Gadian, Magn. Reson. Med., 1994, 32, 707. 94. M. D. King, J. Houseman, D. G. Gadian, and A. Connelly, Magn. Reson. Med., 1997, 38, 930. 95. S. Posse, C. A. Cuenod, and D. Le Bihan, Radiology, 1993, 188, 719. 96. T. L. James and G. G McDonald, J. Magn. Reson., 1973, 11, 58. 97. C. T. W. Moonen, P. C. M. van Zijl, D. Le Bihan, and D. Despres, Magn. Reson. Med., 1990, 13, 467. 98. K. Yoshizaki, Y. Seo, H. Nishikawa, and T. Morimoto, Biophys. J., 1982, 38, 209. 99. P. C. M. van Zijl, C. T. W. Moonen, P. Faustino, J. Pekar, O. Kaplan, and J. S. Cohen, Proc. Natl. Acad. Sci., USA, 1991, 88, 3228. 100. L. E. Kay and J. H. Prestegard, J. Magn. Reson., 1986, 67, 103. 101. C. H. Sotak, J. Magn. Reson., 1990, 90, 198. 102. S. H. Koenig, R. D. Brown, M. Spiller, and N. Lundbom, Magn. Reson. Med., 1990, 14, 482. 103. M. D. Hurlimann, K. G. Helmer, L. L. Latour, and C. H. Sotak, J. Magn. Reson., 1994, 111, 169. 104. P. T. Callaghan, A. Coy, T. P. J. Halpin, D. MacGowan, K. J. Packer, and F. O. Zelaya, J. Chem. Phys., 1992, 97, 651. 105. D. J. Bergman and K. J. Dunn, Phys. Rev., 1995, 51, 3401. 106. R. Mills, J. Phys. C., 1973, 77, 685. 107. D. Le Bihan, J. Delannoy, and R. Levin, Radiology, 1989, 171, 853. 108. J. Delannoy, D. Le Bihan, D. I. Hoult, and R. L. Levin, Med. Phys., 1990, 17, 855. 109. K. Hynynen, A. Darkazanli, and E. Unger, Med. Phys., 1993, 20, 7. 110. H. E. Cline, K. Hynynen, R. D. Watkins, W. J. Adams, J. F. Schenck, R. H. Ettinger, W. R. Freund, J. P. Vetro, and F. A. Jolesz, Magn. Reson. Imag., 1995, 194, 731. 111. J. H. Simpson and H. Y. Carr, Phys. Rev., 1958, 111, 1201. 112. E. O. Stejskal, J. Chem. Phys., 1965, 43, 3597. 113. J. E. Tanner, Biophys. J., 1979, 28, 107. 114. C. H. Neuman, J. Chem. Phys., 1974, 60, 4508. 115. A. V. Barzykin, K. Hayamizu, W. S. Price, and M. Tachiya, J. Magn. Reson., 1995, 114, 39. 116. J. E. Tanner, J. Chem. Phys., 1978, 69, 1748. 117. R. L. Cooper, D. B. Chang, and A. C. Young, Biophys. J., 1974, 14, 161. 118. E. von Meerwall and R. D. Ferguson, J. Chem. Phys., 1981, 74, 6956. 119. C. Nicholson and J. M. Philipps, J. Physiol., 1981, 321, 225. 120. C. Nicholson and E. Sykova, Trends Neurosci., 1998, 21, 470. 121. D. A. Rusakov and D. M. Kullmann, Proc. Natl. Acad. Sci., USA, 1998, 95, 8975. 122. M. F. Lythgoe, A. L. Busza, F. Calamante, C. H. Sotak, M. D. King, A. C. Bingham, S. R. Williams, and D. G. Gadian, Magn. Reson. Med., 1997, 38, 662. 123. M. Hoehn Berlage, M. Eis, and B. Schmitz, NMR Biomed., 1999, 12, 45.
For References see p. 16
18 METHODS AND APPLICATIONS OF DIFFUSION MRI 124. M. E. Moseley, J. Chem. Phys., 1983, 87, 18. 125. P. T. Callaghan and O. Soderman, J. Chem. Phys., 1983, 87, 1737. 126. P. J. Basser, `Imaging Brain Structure and Function', New York Academy of Sciences, New York, 1997, p. 123. 127. P. van Gelderen, M. H. M. De Vleeschouwer, D. Des Pres, J. Pekar, P. C. M. Van Zijl, and C. T. W. Moonen, Magn. Reson. Med., 1994, 31, 154. 128. S. Mori and P. C. M. van Zijl, Magn. Reson. Med., 1995, 33, 41. 129. T. Chun, A. M. Ulug, and P. C. M. van Zijl, Magn. Reson. Med., 1998, 40, 622. 130. E. W. Hsu and S. Mori, Magn. Reson. Med., 1995, 34, 194. 131. P. Douek, R. Turner, J. Pekar, N. J. Patronas, and D. Le Bihan, J. Comput. Assist. Tomogr., 1991, 15, 923. 132. P. J. Basser and C. Pierpaoli, J. Magn. Reson., 1996, B111, 209. 133. C. Pierpaoli and P. J. Basser, Magn. Reson. Med., 1996, 36, 893. 134. P. J. Basser and C. Pierpaoli, Magn. Reson. Med., 1998, 39, 928. 135. T. E. Conturo, R. C. McKinstry, E. Akbudak, and B. H. Robinson, Magn. Reson. Med., 1996, 35, 399. 136. R. I. Shrager and P. J. Basser, Magn. Reson. Med., 1998, 40, 160. 137. J. J. Neil, S. I. Shiran, R. C. McKinstry, et al., Radiology, 1998, 209, 57. 138. R. J. Gilbert, T. G. Reese, S. J. Daftary, R. N. Smith, R. M. Weisskoff, and V. J. Wedeen, Am. J. Physiol., 1998, 38, G363. 139. N. Makris, A. J. Worth, A. G. Sorensen, et al., Ann. Neurol., 1997, 42, 951. 140. D. J. Werring, C. A. Clark, G. J. Barker, D. H. Miller, G. J. M. Parker, M. J. Brammer, E. T. Bullmore, V. P. Giampietro, and A. J. Thompson, J. Neurol. Neurosurg. Psychiatr., 1998, 65, 863. 141. D. J. Werring, C. A. Clark, G. J. M. Parker, D. H. Miller, A. I. Thompson, and G. J. Barker, Neuroimage, 1999, 9, 352. 142. J. W. Meyer, N. Makris, J. F. Bates, V. S. Caviness, and D. N. Kennedy, Neuroimage, 1999, 9, 1. 143. J. Karger and H. Pfeifer, Magn. Reson. Imag., 1994, 12, 235. 144. S. Mori, B. J. Crain, V. P. Chacko, and P. C. M. van Zijl, Ann. Neurol., 1999, 45, 265. 145. C. Poupon, J. F. Mangin, V. Frouin, J. ReÂgis, F. Poupon, M. Pachot-Clouard, D. Le Bihan, and T. Bloch, `Regularization of MR Diffusion Tensor Maps for Tracking Brain White Matter Bundles', Springer-Verlag, 1998, p. 489. 146. E. D. von Meerwall, J. Magn. Reson., 1982, 50, 409. 147. T. Niendorf, R. M. Dijkhuizen, D. G. Norris, M. van Lookeren Campagne, and K. Nicolay, Magn. Reson. Med., 1996, 36, 847. 148. J. G. Li, G. J. Stanisz, and R. M. Henkelman, Magn. Reson. Med., 1998, 40, 79. 149. G. P. Zientara and J. H. Freed, J. Chem. Phys., 1980, 72, 1285. 150. J. Pfeuffer, U. Flogel, and D. Leibfritz, NMR Biomed., 1998, 11, 11. 151. U. Pilatus, H. Shim, D. Artemov, D. Davis, P. C. van Zijl, and J. D. Glickson, Magn. Reson. Med., 1997, 37, 825. 152. R. V. Mulkern, H. Gudbjartsson, C. F. Weston, et al., NMR Biomed., 1999, 12, 51. 153. K. D. Merboldt, D. HoÈrstermann, W. HaÈnicke, H. Bruhn, and J. Frahm, Magn. Reson. Med., 1993, 29, 125. 154. S. Posse, C. A. Cuenod, and D. Le Bihan, in `Diffusion and Perfusion Magnetic Resonance Imaging. Applications to Functional MRI', ed. D. Le Bihan, Raven Press, New York, 1995, p. 57. 155. M. K. Nishijima, R. C. Koehler, and P. D. Hurn, Am. J. Physiol., 1989, 257, H1860. 156. T. Q. Duong, J. J. Ackerman, H. S. Ying, and J. J. Neil, Magn. Reson. Med., 1998, 40, 1. 157. Y. Assaf and Y. Cohen, J. Magn. Reson., 1998, 131, 69. 158. M. Wick, Y. Nagatoma, F. Prielmeier, and J. Frahm, Stroke, 1995, 26, 1930.
For list of General Abbreviations see end-papers
159. A. van der Toorn, R. M. Dijkhuizen, C. A. F. Tulleken, and K. Nicolay, Magn. Reson. Med., 1996, 36, 914. 160. M. A. Rutherford, F. M. Cowan, and A. Y. Manzur, J. Comput. Assist. Tomogr., 1991, 15, 188. 161. P. B. Toft, H. Leth, B. Peitersen, H. C. Lou, and C. Thomsen, J. Comput. Assist. Tomogr., 1996, 20, 1006. 162. C. Baratti, A. S. Barnett, and C. Pierpaoli, Radiology, 1999, 210, 133. 163. I. Vorisek and E. Sykov, J. Neurophysiol., 1997, 78, 912. 164. D. Prayer, T. Roberts, A. J. Barkovich, L. Prayer, J. Kucharczyk, M. Moseley, and A. Arieff, Neuroradiology, 1997, 39, 320. 165. K. Takeda, Y. Nomura, H. Sakuma, T. A. Tagami, Y. Okuda, and T. Nakagawa, J. Comput. Assist. Tomogr., 1997, 21, 1. 166. C. Beaulieu, F. R. Fenrich, and P. S. Allen, Magn. Reson. Imaging., 1998, 16, 1201. 167. D. Le Bihan, R. Turner, and P. Douek, NeuroReport, 1993, 4, 887. 168. C. T. W. Moonen, J. Pekar, M. H. M. De Vleeschouwer, P. Van Gelderen, P. C. M. Van Zijl, and D. Des Pres, Magn. Reson. Med., 1991, 19, 327. 169. J. V. Hajnal, M. Doran, and A. S. Hall, J. Comput. Assist. Tomogr., 1991, 15, 1. 170. M. A. Rutherford, F. M. Conan, and A. Y. Manzur, Radiology, 1991, 180, 229. 171. Y. Assaf and Y. Cohen, J. Magn. Reson., 1996, B112, 151. 172. C. Nicholson and E. Sykova, Trends Neurosci., 1998, 21, 207. 173. C. Beaulieu and P. S. Allen, Magn. Reson. Med., 1994, 32, 579. 174. G. J. Stanisz, A. Szafer, G. A. Wright, and R. M. Henkelman, Magn. Reson. Med., 1997, 37, 103. 175. M. E. Moseley, J. Kuchareczyk, J. Mintorovitch, Y. Cohen, J. Kurhanewicz, N. Derucin, H. Asgari, and D. Normou, Am. J. Neuroradiol., 1990, 11, 423. 176. L. L. Latour, K. Svoboda, P. P. Mitra, and C. H. Sotak, Proc. Natl. Acad. Sci., USA, 1994, 91, 1229. 177. D. G. Norris, T. Niendorf, and D. Leibfritz, NMR Biomed., 1994, 7, 304. 178. J. Pfeuffer, W. Dreher, E. Sykova and D. Leibfritz, Magn. Reson. Imag., 1998, 16, 1023. 179. A. G. Sorensen, F. S. Buonanno, R. G. Gonzalez, et al., Radiology, 1996, 199, 391. 180. S. Warach, D. Chien, W. Li, M. Ronthal, and R. R. Edelman, Neurology, 1992, 42, 1717. 181. S. Warach, M. Boska, and K. M. A. Welch, Stroke, 1997, 28, 481. 182. K. O. LoÈvblad, A. E. Baird, G. Schlaug, et al., Ann. Neurol., 1997, 42, 164. 183. R. G. Gonzalez, P. W. Schaefer, F. S. Buonanno, L. H. Schwamm, R. F. Budzik, G. Rordorf, B. Wang, A. G. Sorensen, and W. J. Koroshetz, Radiology, 1999, 210, 155. 184. S. Warach, M. Boska, and K. M. Welch, Stroke, 1997, 28, 481. 185. W. Dreher, B. KoÈhn, M. L. Gyngell, E. Busch, T. Niendorf, K. A. Hossmann, and D. Leibfritz, Magn. Reson. Med., 1998, 39, 878. 186. J. Ono, K. Harada, T. Mano, K. Sakurai, and S. Okada, Pediatr. Neurol., 1997, 16, 63. 187. T. Iwasawa, H. Matoba, A. Ogi, H. Kurihara, K. Saito, T. Yoshida, S. Matsubara, and A. Nozaki, Magn. Reson. Med., 1997, 38, 484. 188. M. A. Hors®eld, H. B. Larsson, D. K. Jones, and A. Gass, J. Neurol. Neurosurg. Psychiatry, 1998, 64(Suppl 1), S80. 189. H. Ay, F. S. Buonanno, P. W. Schaefer, D. A. Le, B. Wang, R. G. Gonzalez, and W. J. Koroshetz, Neurology, 1998, 51, 1369. 190. H. Hanyu, H. Shindo, D. Kakizaki, K. Abe, T. Iwamoto, and M. Takasaki, Gerontology, 1997, 43, 343.
METHODS AND APPLICATIONS OF DIFFUSION MRI 191. H. Hanyu, H. Sakurai, T. Iwamoto, M. Takasaki, H. Shindo, and K. Abe, J. Neurol. Sci., 1998, 156, 195. 192. M. M. Bahn, D. K. Kido, W. L. Lin, and A. L. Pearlman, Arch. Neurol., 1997, 54, 1411. 193. S. Kropp, M. Finkenstaedt, C. Laske, W. SchulzSchaeffer, I. Zerr, A. Kretschmar, and S. Poser, J. Neurol., 1999, 246, 66. 194. P. Demaerel, L. Heiner, W. Robberecht, R. Sciot, and G. Wilms, Neurology, 1999, 52, 205. 195. T. Q. Li, Z. G. Chen, and T. Hindmarsh, Acta Radiol., 1998, 39, 460. 196. R. A. Zimmerman, J. C. Haselgrove, Z. Y. Wang, J. V. Hunter, M. C. Morriss, A. Hoydu, and L. T. Bilaniuk, Brain Dev., 1998, 20, 275. 197. P. S. Huppi, S. E. Maier, S. Peled, G. P. Zientara, P. D. Barnes, F. A. Jolesz, and J. J. Volpe, Pediatr. Res., 1998, 44, 585. 198. M. S. Buchsbaum, C. Y. Tang, S. Peled, et al., Neuroreport, 1998, 9, 425. 199. D. Le Bihan, P. Douek, M. Argyropoulou, R, Turner, N. Patronas, and M. Fulham, Top. Magn. Reson. Imag., 1993, 5, 25. 200. K. Ikezaki, M. Takahashi, H. Koga, J. Kawai, Z. KovaÈcs, T. Inamura, and M. Fukui, Acta Neurochir. Suppl. (Wien), 1997, 70, 170. 201. K. Krabbe, P. Gideon, P. Wagn, U. Hansen, C. Thomsen, and F. Madsen, Neuroradiology, 1997, 39, 483. 202. P. Barzo, A. Marmarou, P. Fatouros, K. Hayasaki, and F. Corwin, J. Neurosurg., 1997, 87, 900. 203. R. B. Schwartz, R. V. Mulkern, H. Gudbjartsson, and F. Jolesz, Am. J. Neuroradiol., 1998, 19, 859. 204. L. Chang and T. Ernst, Neuroimaging Clin. North Am., 1997, 7, 409. 205. P. W. Schaefer, F. S. Buonanno, R. G. Gonzalez, and L. H. Schwamm, Stroke, 1997, 28, 1082. 206. K. Okada, L. H. Wu, and S. Kobayashi, Stroke, 1999, 30, 478. 207. D. K. Jones, D. Lythgoe, M. A. Hors®eld, A. Simmons, S. C. R. Williams, and H. S. Markus, Stroke, 1999, 30, 393. 208. J. C. Ford, D. B. Hackney, E. Lavi, M. Phillips, and U. Patel, JMRI, 1998, 8, 775.
19
209. B. A. Inglis, L. Yang, E. D. Wirth, D. Plant, and T. H. Mareci, Magn. Reson. Imag., 1997, 15, 441. 210. D. Morvan, Magn. Reson. Imag., 1995, 13, 193. 211. A. Baur, A. Stabler, R. Bruning, R. Bartl, A. Krodel, M. Reiser, and M. Deimling, Radiology, 1998, 207, 349. 212. G. Akansel, V. M. Haughton, R. A. Papke, S. Censky, Am. J. Neuroradiol., 1997, 318, 443. 213. S. A. Englander, A. M. Ulug, R. Brem, J. D. Glickson, and P. C. M. van Zijl, NMR Biomed., 1997, 10, 348. 214. C. F. Maier, Y. Paran, P. Bendel, B. K. Rutt, and H. Degani, Magn. Reson. Med., 1997, 37, 576. 215. T. Namimoto, Y. Yamashita, S. Sumi, Y. Tang, and M. Takahashi, Radiology, 1997, 204, 739. 216. T. Ichikawa, H. Haradome, J. Hachiya, T. Nitatori, and T. Araki, Am. J. Roentgenol., 1998, 170, 397. 217. Y. Yamashita, Y. Tang, and M. Takahashi, JMRI, 1998, 8, 367. 218. I. Yamada, W. Aung, Y. Himeno, T. Nakagawa, and H. Shibuya, Radiology, 1999, 210, 617. 219. R. J. Gilbert, T. G. Reese, S. J. Daftary, R. N. Smith, R. M. Weisskoff, and V. J. Wedeen, Am. J. Physiol., 1998, 275, G363. 220. D. F. Scollan, A. Holmes, R. Winslow, and J. Forder, Am. J. Physiol., 1998, 44, H2308. 221. E. W. Hus, A. L. Muzikant, S. A. Matulevicius, R. C. Penland, and C. S. Henriquez, Am. J. Physiol., 1998, 274, H1627. 222. Y. Zhang, T. V. Samulski, W. T. Joines, J. Mattiello, R. L. Lebin, and D. Le Bihan, Int. J. Hyperthermia, 1992, 8, 263.
Biographical Sketch Denis Le Bihan. b 1957. Ph.D., 1987, Physics, M.D., 1984, Radiology, University of Paris, France. Former Chief of the Radiology Research Section, Clinical Center, NIH, Bethesda, USA. Presently Research Director, French Atomic Energy Commission, Director, Anatomical and Functional Neuroimaging Laboratory, Orsay, France, and Clinical Professor of Radiology, Harvard University, Cambridge, USA. Corresponding Member, French Academy of Sciences. Approx. 400 articles, book chapters, abstracts and patents. Research interests: diffusion, perfusion and functional MRI.
For References see p. 16
BRAIN MRS OF HUMAN SUBJECTS
Brain MRS of Human Subjects James W. Prichard Yale University, New Haven, CT, USA
1 INTRODUCTION A biomedical revolution based on NMR technology is under way. Many of the NMR methods described in this Encyclopedia are already in use for medical research and diagnosis, and many others will ®nd such application soon. As the full impact of mature NMR methods becomes evident in the latter 1990s, it will be measured on the same scale with such things as microscopy and genetics. Two well-established characteristics of biomedical NMR technology ensure this outcome: it is remarkably noninvasive (see Health and Safety Aspects of Human MR Studies) and it is remarkably versatile, as attested in many other chapters. In each respect separately, it exceeds most other technologies available for the study of living tissue. By their combination, it stands alone. This chapter deals with human brain research done by the branch of biomedical NMR technology designated magnetic resonance spectroscopy (MRS) to distinguish it from magnetic resonance imaging (MRI), which makes pictures of anatomical structure from the water proton signal. MRS obtains chemically speci®c information from biological tissue by the measurement of much smaller signals from 1H, 31P, 13C, 17O, and other magnetic nuclei in a variety of compounds. It is extending the range of biomedically useful NMR measurements to practical applications that would have courted dismissal as visionary a decade ago. In 1994, signals from more than two dozen compounds can be detected routinely in the human brain, and many more are in prospect. All report on biochemical phenomena previously accessible only by surgical removal of tissue or at autopsy. As various forms of MRS supported by MRI reach technical maturity, their collective contributions to an understanding of normal and deranged biochemistry in the brain will amount to nothing less than a new neurobiological discipline comparable in scope to today's neurochemistry, neuroanatomy, and neuropathology combined. All stages of normal human brain development and disease processesÐincluding intrauterine stagesÐwill be open to inspection along the many axes MRS can measure. The MRS studies mentioned in this chapter are intended to help scientists and physicians interested in the brain to evaluate the above assertions, which may appear extravagant to readers who are exploring modern NMR for the ®rst time. Many more studies appropriate for that purpose have been published than can be cited. Selection among them was intended to favor efforts that illuminate each other or set new problems by failing to do so, but it cannot have escaped all of the author's biases.
2
1
H AND
1
31
P SPECTRA
Most human brain MRS research to date has used these two nuclei. Figures 1 and 2 illustrate the normal features and some disease-related abnormalities of brain 1H and 31P spectra with simulations that are free of the wide technical variation among spectra in original publications. Differences in acquisition parameters and data processing can cause spectra that are biologically equivalent to look quite different from each other. For readers new to either NMR or neuroscience, the variation can easily distract attention from features of biological importance. The simulated spectra in Figures 1 and 2 were generated by Mathematica# and annotated in Adobe Illustrator#. Within each ®gure, random noise is of the same intensity in all spectra. Resonance variables were adjusted to mimic the appearance typical of many published spectra, in which less intense resonances are not evident because of such factors as long spin echo times, J coupling, and processing methods. The resonances shown are usually identi®able in all spectra from normal adult brain. The disease-related variations are large ones that have been reported by more than one group. They are presented together to emphasize their potential for metabolic ®ngerprinting of different disease processes.
3
NORMAL BRAIN FUNCTION: LACTATE IN HUMAN SENSORY SYSTEMS
A signal from lactate is detectable in 1H spectra from the normal human brain.1 Stimulation of the human visual2±4 and auditory5 systems can cause increases in lactate observable by 1 H MRS in the primary cortical receiving areas. These ®ndings are consistent with positron emission tomography (PET) data showing that similar stimulation can increase glucose uptake more than oxygen extraction in the regions affected.6 The PET workers interpreted their observations as evidence for preferential activation of nonoxidative glycolysis. The MRS data strengthen that view. The notable aspect of this research area is the evidence from at least ®ve different PET and NMR groups showing that glycolysis can increase more than respiration in response to some kinds of activation within the normal range of brain function. The phenomenon is well known in skeletal muscle. Selective activation of muscle cells with differing capacities for glycolysis and respiration is a normal feature of muscle function, and heterogeneous distribution of glycolytic and respiratory enzymes has long been employed for histochemical characterization of individual muscle cells. Evidence for similarly heterogeneous enzyme distribution already exists at the level of cell groups in the animal brain.7 Like muscle cells, neurons and possibly glia may well have variable capacity for glycolysis and respiration related to their physiological functions. These PET and MRS data prompt the question of what adaptive utility lactate elevation resulting from brain activity might have. Two novel possibilities that can be tested by experiment are the following: (1) Lactate may be a local energy storage medium that is increased in anticipation of sudden high energy demand. Ammunition stacked beside cannons defending a fort is more useful than ammunition in a remote storehouse if the fort is
2 BRAIN MRS OF HUMAN SUBJECTS
N-Acetyls
NORMAL Cholines
Creatines
Simulations of 1H SPECTRA from the brain
Lactate
4
3
2
1
ppm PATHOLOGICAL – filled resonances are abnormal Stroke
Glioma
Epilepsy – during seizure
Epilepsy – mesial temporal sclerosis
potentials a second and surrounding cells to recycle transmitter at proportional rates. The immediate metabolic burden of this intense activity falls principally on synaptic terminals, which are quite small and in consequence have low ratios of cytoplasmic volume to the area of excitable surface membrane that they must support. In such a structure, the Krebs cycle can respond to sudden maximum energy demand more quickly if it is not limited by the time necessary for activation of glycolysis, which is long on the scale of intense neuronal discharge and transmitter release. Local lactate elevation by previous activity would prime synaptic terminals to deliver their maximum response to sudden demand while glycolytic rate is increasing. Existing techniques for the measurement of glucose and oxygen uptake and lactate concentration would not detect most instances of this process, because the ones that can be used in vivo, including MRS, report on overall changes in volumes of tissue much larger than the scattered locations at which such priming would ordinarily occur, and the others are bedeviled by the possibility of agonal artifact. Only when it occurs in many cells at the same timeÐas during sensory barrages and seizure dischargeÐcould such a process be detected by present methods.
PCr
Alzheimer disease
Multiple sclerosis plaque
PDE Simulations of 31P SPECTRA from the brain
PME
10
Figure 1 Simulated 1H spectra from brain, created in Mathematica# and annotated in Adobe Illustrator#. Alterations typically caused by six pathological conditions are contrasted with each other and to normal. Spectral variables were adjusted to mimic the appearance common in published spectra. Random noise and each un®lled resonance are of the same intensity in all spectra. The `normal' spectrum represents resonances from choline-containing compounds (cholines), creatine and phosphocreatine (creatines), N-acetylcontaining compounds, mostly N-acetylaspartate (N-acetyls), and lactate methyl protons (lactate). Glutamate and glutamine signals are represented as unresolved low-intensity resonances between 2 and 3 ppm. The `pathological' spectra were created by varying the intensities of signals reported to be affected by the disease states indicated by labels on each spectrum. The abnormalities shown are the ones most often associated with each condition, but they are not all present in every case. The elevated lactate signal in four spectra displays the characteristic 7 Hz splitting caused by J coupling to the C-2 proton
suddenly attacked. Suspicion of imminent attack would certainly cause the fort's defenders to move extra ammunition from the storehouse to the cannons, to shorten the response time and lengthen the duration of maximum ®ring. In the brain, neurons can suddenly be called upon to ®re hundreds of action
Pi
Normal
ATP Alpha Gamma
0
ppm
–10
Beta
–20
PATHOLOGICAL – filled resonances are abnormal
Stroke – moderate
Stroke – severe
Epilepsy – during seizure Glioma
Figure 2 Simulated 31P spectra from normal and diseased brain, created and presented by the same procedures used for Figure 1. The resonances represented are from phosphomonoesters (PME), inorganic phosphate (Pi), phosphodiesters (PDE), phosphocreatine (PCr) and the , , and phosphates of adenosine triphosphate (ATP). In the `stroke - severe' example, all of the signal present is Pi, indicating total loss of high-energy phosphate compounds
BRAIN MRS OF HUMAN SUBJECTS
(2) Lactate may have a signaling function in the brain. Lactate is well suited to be a neuromodulatorÐa substance which alters the excitability of local neural ensembles. It is produced at increased rates whenever glycolysis accelerates in response to neural activity. It accumulates during pathological activation caused by ischemia or lack of oxygen, and, as described above, during some kinds of activation in the physiological range. IfÐupon escape from the active cell that produced itÐit reduced the excitability of adjacent cells, it could both inhibit seizure discharge and `sharpen' transmission along multi®ber pathways, a process known to occur by other mechanisms. A ®nding consistent with this idea is available: in Torpedo electroplax, lactate in concentrations that can be produced by brain activation inhibited release of acetylcholine.8
4 STROKE Stroke is the commonest major brain disease and a principal medical burden on society through the lost productivity it causes and its consumption of resources by extended disability. It is a natural place to test the bene®ts of new biomedical technology. MRS observations on human stroke are far enough advanced to allow initial assessment of the role MRS will come to have in patient management. Lactate and N-acetyl (NA) resonances in the 1H spectrum and intracellular pH (pHi) derived from the 31P spectrum appear to have the best prospects for clinical utility.
4.1
1
H MRS in Stroke
Figure 1 illustrates the kind of change commonly observed in 1H spectra of human brain infarction: loss of the NA signal and elevation of the signal from lactate. Loss of NA signal should ensue from death of neurons, which are the only cells in mature brain thought to contain N-acetylaspartate (NAA), the principal source of the NA resonance. The NA signal was depressed both in the ®rst few days after stroke onset and several weeks later in six patients studied at least twice.9 In 10 patients studied within 60 h of stroke onset, it was reduced relative to the homologous region of the contralateral hemisphere in all but two; in seven of the same patients, repeat examination 1±2 weeks later showed additional decline at an overall rate calculated at ÿ29% 9% per week.10 Similar ®ndings have been reported by other groups.11±13 Decline of the NA signal in the ®rst days after a stroke surely re¯ects, in part, clearing of debris from neurons killed at the outset, but it may also be evidence of continuing neuronal loss after the acute period, possibly in the ischemic penumbra.14 Therapy which prevents delayed neuronal loss would presumably also reduce eventual ®xed de®cit. Monitoring of the NA signal may prove useful as a surrogate endpoint for evaluation of therapies intended to reduce continuing neuronal loss in the subacute period. Lactate elevation by acute infarction is readily detectable by 1H MRS (Figure 1). However, the ®rst report of increased lactate associated with stroke14 was on patients studied months after the stroke occurred. Many observers including
3
the present writer doubted that stroke-associated lactate elevation would persist so long, but subsequent observations by several groups have shown that it does. Serial study of individual patients demonstrated continuous lactate elevation for weeks after stroke in most and for several months in some.9 Again, similar observations have been made by other groups.11±13 Multiple mechanisms of lactate elevation must be at work over so long a period. A newly infarcted region of brain accumulates lactate to a concentration of 15±30 mM within minutes of losing its blood supply, as glucose and glycogen in unperfused tissue are metabolized in the absence of oxygen, which is exhausted in the ®rst few seconds. If the region is never reperfused, the lactate in it can dissipate only by diffusion, but the process would not take weeks to months in the presence of active tissue repair processes. Other possible sources of elevated lactate associated with infarction include adjacent regions of surviving but impaired tissue referred to as an `ischemic penumbra',15 in®ltrating cells involved in the tissue's reaction to injury, and altered metabolism of surviving brain. MRS has provided evidence for the second of these:16 the brain of a patient studied two weeks after a stroke and autopsied a week later had large numbers of macrophages in the regions where 1H MRS had shown elevated lactate. The role of altered metabolism is conjectural at present; the longterm metabolic response of brain tissue that survives ischemic insult is not well understood. Lactate that accumulates in the killed core of an infarct is metabolically inert, while that in an ischemic penumbra turns over, albeit probably at a rate different from that of normal brain tissue. The anatomical boundary between pools of lactate in these two states need not be sharp; the pools may even be anatomically interdigitated. Their relative sizes re¯ect the proportion of killed to surviving tissue early in the history of the lesion, and may therefore indicate how a particular lesion is likely to evolve. An advanced form of combined 1H/13C MRS that has been shown to be feasible in humans can make a pertinent measurement. Inert and metabolically active lactate associated with a human stroke can be distinguished from each other by determining how much of the total stroke-associated lactate pool can be labeled with 13C from blood glucose. Explanation of this remarkable possibility requires a brief description of the NMR properties of 13C.
4.2
13
C MRS
Carbon in Nature is nearly 99% 12C, which is not magnetic and hence gives no NMR signal. The stable magnetic isotope 13 C is 1.1% naturally abundant. Its potential for biological studies was appreciated early in the second decade of empirical NMR work. In 1958, the following