MAGNETIC RESONANCE IN MEDICINE

28,328-338 ( 1992)

Magnetic Resonance Neurography F. A. HOWE,*A. G . FILLER,**^'$ B. A. BELL,+AND J. R. GRIFFITHS* *CRC Biomedical Magnetic Resonance Research Group, Division of Biochemistry, St. George> Hospital Medical School, London, S W I 7 ORE, England; 7 Department of Neurosurgery, University of Washington, Seattle, Washington 98109; and $Atkinson Morley’s Hospital, London, United Kingdom Received August 6, 1992; revised September 25, 1992, accepted September 26, 1992 We have made cross-sectional image “neurograms” in which peripheral nerve has a greater signal intensity than that of other tissue. Neurographic images of the rabbit forelimb were obtained using a spin-echo magnetic resonance imaging (MRI) technique that combines fat suppression and diffusion weighting. After fat suppression the nerve shows up in relative isolation and is brighter than the surrounding tissue due to its longer T2relaxation time of =SO ms compared to =27 ms for muscle. The addition of pulsed gradients for diffusion weighting of the MR signal further enhances the intensity of the nerve signal relative to that of surrounding muscle tissue. The greater diffusional anisotropy of nerve tissue ( Dl,/D, = 3.1 )compared to that of muscle ( Dl,/D, = 1.9) allows further enhancement of the nerve by a subtraction of two diffusion-weighte,d images, one with the gradients oriented parallel and one with the gradients oriented perpendicular to the nerve orientation. We show that by manipulation of the MRI parameters, either echo time or pulsed gradient strength, the nerves can be made to show up as the most intense feature. This verifies the feasibility of generating three-dimensional“neurographic” images, analogous to angiograms, but which demonstrate the peripheral nerve tracts in apparent isolation. 0 1992 Academic Press, Inc.

INTRODUCTION

The fine spatial resolution required to create detailed images of peripheral nerves is well within the physical range of current clinical magnetic resonance ( M R ) imagers. The improved imaging of nerves would be a great advantage both in the diagnosis of nerve compression and injury and in the planning of surgery either upon nerves or near them. For example, a clear and distinct image of the course of the recurrent laryngeal nerve would be of great help to a surgeon operating in the anterior neck. Other routine diagnostic problems concerning the peripheral nerves passing through the carpal tunnel are similarly hampered at present by the difficulty of readily identifying and following the courses of nerves as they travel among other structures. There has recently been interest in the use of contrast agents to study material transport in nerve by axoplasmic flow (1, 2) and that type of study would also be greatly facilitated by the use of an imaging sequence with strong nerve contrast. Nerves have not been susceptible to imaging with X-ray or nuclear medicine techniques, and previous MR techniques for nerve have been hampered by poor signal-to-noise ( S / N ) and by the very poor conspicuity of nerves relative to the other tissues through which they pass. Ideally, it is desirable to generate a “neurographic” image analogous to an angiogram, but which highlights nerve rather than blood vessels. To this end we have used a spin0740-3 194192 $5.00 Copyright 0 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.

328

COMMUNICATIONS

329

echo-type imaging sequence to study the NMR properties of nerve that can be used to provide contrast and so discriminate the neural tissue from that surrounding it. A cross-sectional neurogram in which the nerve is left as the brightest feature in the image is produced. Standard techniques, such as maximum intensity projection ( MIP), could then be used to reconstruct a 3D representation of the nerve tracts. Nerves do not exist in isolation. In the brain, the white matter nerve tracts run through gray matter tissue; in the limbs, the peripheral nerve fibers are typically surrounded by fat or muscle. The physical structure of nerve is long axonal pathways surrounded by a myelin sheath. As a consequence there is considerable anisotropy in the measured diffusion coefficients of nerve tracts in the brain ( 3 ) .This could be due to the physical restriction of the axonal water or to the decreased mobility of water across the lipid bilayers (4-6). Anisotropic diffusion has been used to map the white matter tracts in the brain ( 7) and could also prove useful for nerve contrast in peripheral limb studies. Anisotropy in the diffusion coefficient has also been observed, but not quantified, in the human tibia1 nerve ( 3 ) . In peripheral nerve, it is possible that the measured diffusion anisotropy arises from the water in the endoneurial space, which is comparable to extracellular water in CNS tracts. However, the fibrous structure of muscle also leads to diffusional anisotropy (8, 9 ) , although to a lesser extent than in nerve tracts. To utilize the effect of anisotropic diffusion to provide enhanced peripheral nerve contrast, there must be a significant difference in the absolute value either of diffusion coefficientsbetween nerve and muscle or of the anisotropy in these coefficients. Diffusion weighting of MR images can easily be achieved by the incorporation of additional pulsed gradients in a standard spin-echo sequence ( 4 ) .Either the apparent T 2relaxation time or the apparent diffusion coefficient D can then be measured using the same sequence. In spin-echo images fat is generally the brightest feature, and initial studies of peripheral nerves in rabbit forelimbs showed a very large fat component surrounding or adjacent to what were thought to be nerves ( 1). Fat suppression has been shown to enhance visualization of the optic nerve ( 1 0 ) and a fat-suppression sequence has been used for all image data reported here. METHODS

Images were obtained with a 4.7-T, 33-cm SISCO system fitted with a 10 G/cm high-performanceauxiliary gradient insert ( 12-cm inner bore). A three-turn solenoid, 3 cm in diameter and 2.5 cm long, was placed around the upper portion of the forelimb of 2- to 2.5-kg rabbits. The forelimb was taped to the side of the supporting cradle to minimize motion artifacts. The animals were maintained under balanced continuous intravenous infusion of an anesthetic mixture containing 1 mg of medazolam, 1.5 mg of fluanisone, and 50 pg fentanyl/ml at rates of 4 to 10 ml/h to achieve a deep anesthesia which further minimized motion from respiration. The imaging sequence shown in Fig. 1 is a standard multislice spin-echo imaging sequence which has been modified to incorporate elements for fat suppression and diffusion weighting. A chemical-shift-selective(CHESS) ( I I ) pulse sequence was used for fat suppression. A 3-ms Gaussian pulse (with a -3-dB bandwidth of 600 Hz) was used for selective excitation of the fat resonance, followed by a dephasing gradient pulse of 5 G/cm for 3 ms. This CHESS sequence was repeated three times with orthogonal gradients (12) prior to each spin-echo sequence.

330

COMMUNICATIONS 90'

CHESS

~

TE/Z

~

echo

180° -TTEZ/ -

RF Dephasing gradients Slice select Readout gradient

i

Phase encoding

-

Diffusion weighting

A .

6

n

FIG. I. RF and gradient pulse timing of the spin-echo sequence modified to incorporate fat suppression and diffusion weighting. Note the position of the readout dephasing gradient, which is placed toward the end of the sequence to minimize the effects of coupling between the imaging and diffusion-weightinggradients.

Pulsed gradients for diffusion weighting of the image were placed symmetrically around the inversion pulse. For an echo time (TE) of 50 ms the diffusion gradients had a duration of 6 = 10 ms and a separation A = 20 ms and could be oriented parallel or perpendicular to the image plane to measure the anisotropy in the diffusion coefficient. To reduce the effect of cross-terms between the imaging gradients and the diffusion gradients the original spin-echo sequence was modified (13, 14). The readout gradient rephasing lobe was placed directly before the echo acquisition instead of after the slice-selective excitation pulse. However, a consequence of this was the appearance of artifacts in the non-diffusion-weighted images, due to an unwanted echo, presumably formed from imperfections in the slice-selection profiles, which was dephased by the readout gradient in the original sequence. To overcome this a second modification was to split the phase-encoding gradient into two sections and acquire two or four transients (dependent on the S / N ) with phase cycling. It can be calculated ( 13)that

TABLE 1 Apparent T2Relaxation Time and Diffusion Coefficients (Oil, 0,) of Nerve and Muscle Measured in Vivo in the Rabbit Forelimb

T2(ms) Water Muscle Nerve

-

26 53

D, (

cm2/s) 2.26 1.17 0.65

D,, (

cm2/s) 2.20 2.18 2.00

D,,:D, 1.03 1.9 3.1

COMMUNICATIONS

33 1

the remaining cross-terms contribute less than 3% to the diffusion-weighting factor and measurements on pure water yield isotropic diffusion coefficients to this accuracy (see Table 1 ). To enable the determination of the apparent T2and D ,multiple images of the same transverse slice were acquired with either different TEs or different diffusion gradient strengths (GI).The data were obtained in an interleaved manner; the variable parameter (TE, GI, or the orientation of G,) was arrayed for each phase-encoding step, thus averaging any effects due to motion within the duration of the experiment. All images were acquired with a repetition time TR = 1.5 s over a 4- by 4-cm field of view (FOV) and a 2-mm slice thickness. Two hundred fifty-six phase-encoding steps were used and 5 12 data points were acquired, giving a basic in-plane image resolution of 156 by 78 pm. For image analysis the data were zero filled to 1024 by 1024 and Fourier transformed after applying a 2D Gaussian filter that attenuated the high-frequency time-domain components by 6 dB. This clarified the delineation of the very small regions of interest ( ROIs) (typically 0.001 to 0.003 cm2for nerves) but did not significantly change the average pixel intensities measured. Image analysis was performed using software supplied with the spectrometer. An arbitrarily shaped ROI could be specified for each nerve or muscle area from which the average pixel intensity was then determined. The ROI could then be stored to analyze each image in a particular data set. For a particular image feature it was assumed that the average image intensity could be represented by the equation

S

=

-

-

A . exp[ -TE/ T 2 ] exp [ --bD],

[ll

where the gradient factor is

b

y 2 -G; - a 2 - [ A - 6/31.

121 A linear regression analysis of the logarithmic plot of image intensity versus TE or b will then determine the values of the apparent T 2relaxation time or apparent diffusion coefficient D , respectively. =

RESULTS

Fat-suppressed transverse images through the rabbit upper forelimb, in which the nerves are primarily orthogonal to the image plane, were acquired with echo times from 30 to 100 ms. No diffusion weighting was used. Representative images from one rabbit are shown in Fig. 2, which includes a schematic anatomical diagram for comparison. The trace above each MR image represents the intensity along a horizontal line through the median nerve, and to the side a maximum-intensity (i.e., skyline) projection over the whole image is shown. Logarithmic plots of the average image intensity over muscle and nerve ROIs versus TE are shown in Fig. 3. The regression lines (correlation coefficient Y > 0.99) are for the average of the nerve or muscle data and calculated values of the apparent T2are given in Table 1. Fat-suppressed diffusion-weighted transverse images were acquired at a TE of 50 ms. Diffusion gradient strengths of Gi = 0, 3, 5, and 7 G/cm, corresponding to b = 0, 10.7, 29.8, and 58.5 X lo3 s/cm2, were applied parallel and perpendicular to the image plane. Representative images with diffusion gradients aligned in the image plane (orthogonal to the nerves and muscle fibers) are shown in Fig. 4. An intensity trace

332

COMMUNICATIONS

FIG. 2. ( A ) Schematic diagram of the transverse section through the upper forelimb of a rabbit. Fatsuppressed spin-echo images (no diffusion weighting) of the rabbit forelimb were acquired at TEs of ( B ) 30 ms, ( C ) 60 ms, and ( D ) 100 ms. The intensity is normalized to the radial nerve in each image and the intensity trace at the top of each image is taken through the median nerve as indicated by the arrows in ( D ) . To the side of each image a maximum-intensity (i.e., skyline) projection of the whole image is shown.

through the radial nerve and an MIP are shown with each image. Logarithmic plots of the average image intensity for muscle and nerve ROIs versus b are shown in Fig. 5. The regression lines ( r > 0.97) are for the averages of all the nerve or muscle data for each orientation of the diffusion gradients. The calculated apparent diffusion coefficients D,, and DL(where the subscripts refer to measurements parallel and perpendicular to the nerve or muscle fiber orientation) are given in Table 1. In Fig. 6D a diffusion “anisotropy” image is created by subtracting an image with diffusion gradients oriented perpendicular to the image plane (Fig. 6 C ) from one with the diffusion gradients oriented parallel to the image plane (Fig. 6B). Figure 6A shows the normal spin-echo image. The intensity traces are from a line through the median nerve and MIPS for each image are also shown. The muscle and nerve signal intensities calculated from Eq. [ 11 for selected echo times and diffusion gradient strengths are shown in Table 2. A parameter, R , has been defined as the ratio of nerve-to-muscle signal intensity. The contrast-to-noise, C / N , is calculated from the ratio of the difference between nerve and muscle signal intensity to the average background noise. (In the same units as those in Table 2

333

COMMUNICATIONS

20

30

40

SO

60

70

80

90

100

110

Echo time TE (ms)

FIG. 3. Logarithmic plot of image intensity (normalized to the intensity at TE = 30 ms) against echo time TE. Data are from two rabbits. Regions of interest were 17,radial nerve; A, median nerve; 0, ulnar nerve; V, muscle. The regression lines obtained from the averages of the nerve and muscle data are shown.

the average noise was 3 units for all images except for the subtraction image, for which it was 3\/2.) DISCUSSION

Simple spin-echo images of limb anatomy proved inadequate for definitive identification of peripheral nerves which are typically surrounded by high-intensity signal from fat deposits in the intermuscular spaces. Further, a variety of structures similar in nerve size and shape follow similar routes. However, we have found that the fat surrounding nerve is actually beneficial for nerve identification, because in a fat-suppressed image, a relatively high-intensity nerve signal stands out sharply within the low-intensity space left behind by the suppressed fat signal (see Figs. 2B, 4A, and 6A). Relaxation time measurements (using the standard inversion-recovery and CPMG sequences) made on freshly excised median nerve from a rabbit forelimb gave T I N 1.3 s and T 2 N 48 ms (our unpublished data). Our in vivo apparent T2 of nerve, obtained using the spin-echo imaging sequence, corresponds well with these previous measurements. The apparent T2of nerve is relatively long compared to that of muscle (Table 1) and so T2 weighting can be used to improve nerve visibility. (However, note from Table 2 that for TE < 30 ms the muscle signal intensity will be much greater than that of nerve.) The trace through the median nerve shows a large increase in nerve-to-muscle contrast as TE is increased. For TE = 100 ms there is only a small remaining contribution from muscle (Fig. 2D). However, although there is an initial increase in C / N as TE is increased, for TE greater than 70 ms, C/ N rapidly decreases and other bright features still remain. The MIP shows some ambiguity of identification and of the position of the ulnar nerve. Application of pulsed gradients to provide diffusion weighting of the image showed that the diffusion coefficient perpendicular to the nerve is much less than that of muscle and other tissue as shown in the sequence of images in Fig. 4. The application of diffusion weighting can therefore provide a relative enhancement of nerve signal

334

COMMUNICATIONS

FIG.4. Diffusion-weighted fat-suppressed images of a single transverse slice through the upper forelimb of a rabbit at different diffusion gradient strengths. TE = 50 ms and the gradient factor b is ( A ) 0, ( B ) 10.7 X lo’ s/cm2,(C) 29.8 X lo3s/cm2,(D) 58.5 X lo’ s/cm2.The diffusion gradient was oriented perpendicular to the nerve fiber, the intensity trace at the top of each image is taken through the radial nerve as indicated by the arrows in (D), and a maximum-intensity projection is shown to the side of each image. The additional nerve observed in the top half of ( D ) is the musculocutaneous nerve.

over that of muscle, as demonstrated by the increase in both R and C/ N (see data in Fig. 5 and Table 2). Also, very importantly, other bright features such as lymphatics and ligaments, which are observed in Fig. 4A, are reduced in intensity by diffusion weighting. Vessels generate little signal because the movement of blood is rapid enough to cause flow voids under conditions useful for enhancing nerve signal. In Fig. 4 as the diffusion gradient strength is increased, the peaks in the MIP are successively reduced until only peaks corresponding to nerve remain. At the maximum diffusion weighting used (with a 7 G/cm gradient strength in Fig. 4D) there are four nerves (the musculocutaneous nerve in addition to the median, radial, and ulnar nerves noted in Fig. 2A) that are clearly observed as the most intense features. Their positions are very well defined by the MIP in Fig. 4D. Analysis of the data in Fig. 5 shows that there is greater anisotropy in nerve than in muscle. Further enhancement of the nerve signal relative to the surrounding muscle can thus be obtained by the subtraction of two images, one with the diffusion gradients oriented perpendicular to the nerve tracts and one with the gradients oriented parallel. In practice, at the very high spatial resolution used in this study, this poses problems

335

COMMUNICATIONS A h

,---0.2 x

'Z-0.4.. 0 I

.

2

-0.6-

,

0 x, -0.8

.-

,

yi

C

,, -.

'u - 1 .C - . : v v

'-.

:

C -1.2! ~

1.4

L.. 0

,?J

I

._ . ~ _ _ _ ~

10

20

30

40

50

Gradient factor b (105sec/cm2 )

60

Gradient factor b (105sec/cm2)

FIG. 5. Logarithmic plots of image intensity (normalized to that at b = 0) against the gradient factor b for gradients applied perpendicular (closed symbols) and parallel (open symbols) to the nerve fiber orientation. A, muscle; ( B ) U,O, radial nerve; A, 0, median Data are from two rabbits and regions of interest are ( A ) 0, nerve; V,Q ulnar nerve. The regression lines are calculated from the average of the nerve or muscle data for each of the two gradient orientations.

of registration between the two images when the nerves have such small dimensions. Even if the two images are acquired in an interleaved manner, distortions in the image are possible due to the different eddy-current effects for the two diffusion gradient orientations. Edge artifacts are then observed and the contrast-to-noise of the nerves are reduced. A successful subtraction image is shown in Fig. 6. In just the fat-suppressed spin-echo image (Fig. 6A) the median and radial nerves are fairly well defined but are not the most intense features. In particular, there was imperfect fat suppression at the edge of the image (which in this case was very close to the windings of the solenoid). In Fig. 6B diffusion gradients of 7 G/cm were applied perpendicular to the nerve. The ulnar nerve is now clearly defined and, apart from the residual fat signal, unwanted high-intensity features have been suppressed. With the diffusion gradients applied parallel to the nerve direction the nerve signal is suppressed, and only muscle and residual fat remain, as seen in Fig. 6C. Subtraction of Fig. 6C from Fig. 6B yields a cross-sectional neurogram in which the nerves are shown in apparent isolation. There is also very good suppression of this residual fat in the subtracted image. The trace shows that although the C / N is reduced by subtraction of two images, there is still an overall enhancement of nerve-to-muscle contrast. The MIP of the subtraction image clearly shows that the three nerves are the most intense features and they have well-defined positions. From Eq. [ I], and using the data in Table 1, it is possible to calculate the nerveto-muscle contrast for any combination of diffusion gradient strength and echo time. Table 2 shows that the nerve-to-muscle contrast achieved at long echo times is greater than that achieved using diffusion weighting but there is a reduction in C / N .However, this will be dependent on the specific values of the apparent T2and diffusion coefficients and does not take into account tissue other than muscle that may also appear as bright as nerve. However, the use of diffusion weighting emphasizes the diffusional anisotropy, which is greater for nerve than for other tissue, and thus aids confirmation that the hyperintense features are in fact nerve. The larger diffusion coefficient anisotropy of

336

COMMUNICATIONS

FIG.6 . Utilization of the diffusion coefficient anisotropy of nerves to create a cross-sectional neurogram. (A) Fat-supressed spin-echo image of a transverse section through the forelimb of a rabbit. ( B ) as in ( A ) but with diffusion gradients ( b = 58.5 X lo3s/cmZ)oriented perpendicular to the nerve fiber. (C) as in ( A ) but with diffusion gradients oriented parallel to the nerve. ( D ) Cross-sectional neurogram obtained by subtracting image (C) from (A). The intensity traces are taken through the median nerve as indicated by the arrows in (D). To the side of each image is a maximum-intensity projection of the image data. In ( D ) the MIP clearly shows that the three nerves have the greatest intensity.

nerves relative to that of other tissue also allows a cross-sectional neurogram (Fig. 6D) to be created in which the nerves are almost the only feature observed. CONCLUSIONS

In spin-echo images, fat suppression by CHESS enhances the visualization of peripheral nerves. Further, the application of diffusion weighting perpendicular to the nerve orientation yields a relative change in the muscle intensity far larger than that of the nerve. For the peripheral nerves observed in the rabbit forelimb there are three factors that allow the manipulation of the image contrast to make nerves the hyperintense features: (i) they have a longer apparent T 2than surrounding muscle; (ii) they have a much smaller perpendicular diffusion coefficient than muscle; and (iii) they have a larger diffusion coefficient anisotropy than muscle. One or more of these properties may be used to generate an image where a threshhold above which only the nerve is observed can be set. Thus, from a data set of cross-sectional neurograms, it is feasible to create a three-dimensional neurogram showing only the nerve tracts. The

337

COMMUNICATIONS TABLE 2 Calculated Image Signal Intensity of Nerve (S.) and Muscle (S,) for Selected Values of Echo Time (TE) and Diffusion Gradient Strength (GJ Using Eq. [ 11

Gi(G/cm)

TE (ms)

S,

0 0

0 50 50 50 70 100

120 44 30 17" 30 16

7 70

0 0

s m

205 30 15 7'

14

4.4

R 0.59 1.4 2.00 2.43" 2.16 3.65

CIN

-85 4.7

5.0 2.4" 5.3 3.9

Note. The nerve-to-muscle contrast parameter is calculated from R = SJS,. The nerve-to-muscle contrast-to-noise ratio is calculated from C/N = (S, - S,)/N for an average noise of N = 3 units (except for the subtracted data, where N = 3\12 units). Calculated from difference between signal intensities with pulsed gradients perpendicular and parallel to nerve orientation.

formation of 3D images will be aided by a sequence for acquisition of data more timeefficient than the simple spin-echo sequence used here. For example, a stimulated echo sequence can allow the acquisition of two independently diffusion-weighted images in which the pulse gradients have been applied aIong orthogonal axes ( 1 5 ) . Fast sequences using echo-planar or steady-state free precession have also been developed ( 4 ) . The MR characterization of nerves in the human forearm at clinical field strength is currently underway in our laboratory and we comment on the practicalities of transfemng this technique from a high-field experimental system. The typical 1 G/cm gradient strength of a clinical system can provide the large diffusion weighting achieved with our 7 G/cm gradient strength only by using a much longer TE so that both A and 6 can be increased. With a TE of 150 ms, a more typical echo time in clinical imaging, a fourfold increase in both A and 6 will produce a gradient factor of b = 76 X lo3 s/cm2. Alternatively, the cross-terms between the diffusion gradients and the imaging gradients ( 1 3 ) could be used to advantageously increase the diffusion weighting if properly calibrated. A preferrable solution would be to use small highstrength gradient inserts. This would also minimize the detrimental effects of eddy currents, but probably limit the studies to peripheral nerve in limbs. At the lower clinical field strength of 1.5 T there will be an approximate threefold decrease in S / N , and increasing the TE to 150 ms will reduce the nerve signal intensity by an additional factor of four (using our measured apparent T , value for nerve). However, this is offset by the much larger size of human nerve, typically 3 mm diameter for median nerve in the human forearm, compared to the median nerve in the rabbit, which is less than 0.5 mm in diameter. From these figures we might expect an actual increase in SI N if the pixel density per nerve is kept constant. Diffusion-weighted images are very prone to motion artifacts. As discussed earlier, this can produce serious artifacts if subtraction images are required. The fast imaging techniques mentioned

338

COMMUNICATIONS

above may alleviate this problem if the requirements of high diffusion weighting can also be met. Traumatic, neuropathic, and compressive pathology affecting peripheral, cranial, and autonomic nerves constitutes a very significant portion of the clinical work of neurologistsand neurosurgeons. However, because there have been no effective means of imaging nerves, radiologists have had only a very limited involvement in the evaluation of these conditions. We believe that the improvement in nerve image conspicuity yielded by neurographic techniques described in this paper will make possible an enhanced role for radiology in the diagnosis and management of disease and injury involving nerves. ACKNOWLEDGMENrS This work was supported by the Cancer Research Campaign, UK. A.G.F. acknowledges the award of a Hamson Lectureship from the Neuroscience Foundation while at AMH. REFERENCES 1. A. G . FILLER,H. R. WINN,F. A. HOWE,J. R. GRIFFITHS,B. A. BELL,AND T. W. DEACON,in “Book of Abstracts. Society of Magnetic Resonance in Medicine, 10th Annual Meeting, 1991,” p. 985.

2. P. GHOSH,X. ZHOU,W. LIN,A. S. FENG,E. GROMAN, A N D P. C. LAUTERBUR, in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 9th Annual Meeting, 1990,” p. 1042. 3. M. E. MOSELEY,J. KUCHARCZYK, H. S. ASGARI,AND D. NORMAN,Magn. Reson. Med. 19, 321 (1991). 4. D. LE BIHAN,Mugn. Reson Q. 7, 1 (1991 ). 5. D. LE BIHAN,R. TURNER, A N D P. DOUEK,in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 9th Annual Meeting, 1990,” p. 377. 6. C. T. W. MOONEN,M. H. M. DE VLEESCHOUWER, D. DESPRES,P. C. M. VANZIJL,and J. PEKAR,in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 9th Annual Meeting, 1990,” p. 1121.

7. P. DOUEK,R. TURNER, N. PATRONAS, AND D. LE BIHAN,in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 10th Annual Meeting, 1991,” p. 919. 8. M. E. MOSELEYAND M. F. WENDLAND, in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 10th Annual Meeting, 1991,” p. 108. 9. L. GARRIDO,U. M. SPENCER,K. K. KWONG,VANJ. WEDEEN,AND H. L. KANTOR,in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 10th Annual Meeting, 1991,” p. 352. 10. J. GUY,J. FITZSIMMONS, E. A. ELLIS,B. BECK,AND A. MANCUSO, in “Book of Abstracts, Society of Magnetic Resonance in Medicine, 10th Annual Meeting, 1991,” p. 9 13. 11. A. HAASE,J. FRAHM,W. HAENICKE, AND D. MATHEI, Phy:i. Med. Bid. 30,341 ( 1988). 12. C. T. W. MOONENA N D P. C. M. VANZIJL,J. Mugn. Resow. 88,28 ( 1990). 13. M. NEEMAN, J. P. FREYER,AND L. 0. SILLERUD, J. Magn. Reson. 90,303 (1990). 14. M. D. KING,N. VANBRUGGEN, R. G. AHER,J. E. CREMER, J. V. HAVNAL, S. R. WILLIAMS, AND M. DORAN,Magn. Reson. Med. 20, 158 ( 1991 ). 15. LIMINLI AND C. H. SOTAK,J. Magn. Reson. 96,501 (1992).

Magnetic resonance neurography.

We have made cross-sectional image "neurograms" in which peripheral nerve has a greater signal intensity than that of other tissue. Neurographic image...
809KB Sizes 0 Downloads 0 Views