Original Research

Diffusion Tensor Imaging of the Median Nerve inHealthy and Carpal Tunnel Syndrome Subjects

Dan Stein, MSc,1 Arnon Neufeld, PhD,2 Ofer Pasternak, PhD,3 Moshe Graif, MD,2

Hagar Patish, MD,4 Etti Schwimmer, MSc,4 Efrat Ziv, MSc,5 and Yaniv Assaf, PhD1,2*

Purpose: To determine if diffusion tensor imaging (DTI) ofthe median nerve could allow identification of patients withcarpal tunnel syndrome (CTS).

Materials and Methods: A total of 13 healthy subjects and9 CTS patients were scanned on a 3T magnetic resonanceimaging (MRI) scanner. The MRI protocol included a DTIsequence from which the fractional anisotropy (FA), appar-ent diffusion coefficient (ADC), and the parallel and radialdiffusivities could be extracted. Those parameters werequantified at different locations along the median nerve(proximal to the carpal tunnel, within the carpal tunnel,and distal to the carpal tunnel).

Results: At the carpal tunnel, the FA, radial diffusivity, andADC differed significantly between healthy subjects andCTS patients (P � 0.0002). This highly significant differencebetween the two groups was due to an opposite trend ofchanges in the DTI indices between the proximal to thecarpal tunnel and within the carpal tunnel locations. Inhealthy subjects the FA increased (�20%, P � 0.001) andthe radial diffusivity and ADC decreased (by �15% and�8%, respectively, P � 0.05) between the proximal to thecarpal tunnel and within the carpal tunnel locations. InCTS subjects the FA decreased (by �21%, P � 0.05) and theradial diffusivity increased (by �23%, P � 0.01) between theproximal to the carpal tunnel and within the carpal tunnellocations.

Conclusion: DTI enables visualization and characteriza-tion of the median nerve in healthy subjects and CTS pa-tients. DTI indices show clear-cut discrimination between

the two groups and in fact enables the of use DTI in thediagnosis of CTS.

Key Words: median nerve; carpal tunnel; carpal tunnelsyndrome; magnetic resonance imaging; diffusion tensorimaging; tractography.J. Magn. Reson. Imaging 2009;29:657–662.© 2009 Wiley-Liss, Inc.

THE MEDIAN NERVE is the largest of the three nervesgoing through the forearm (1,2). Upon entering thepalm the median nerve passes through the carpal tun-nel created by the wrist bones and ligaments (1,2). Acommon pathology encountered in the palms of thehands is carpal tunnel syndrome (CTS) (3–6). Affectingalmost 5% of the population, CTS is the most commonfocal peripheral neuropathy. CTS is characterized bypain, numbness, and paresthesia in the median nervedistribution and weakness in the thenar muscles (3–6).It may be caused by a variety of factors that lead topressure on the median nerve as it passes through thecarpal tunnel at the wrist (3–6).

Diagnosing CTS entails clinical and electrophysiolog-ical examinations before moving on to more invasiveprocedures such as nerve and muscle biopsy (6,7). Thediagnosis of CTS can be straightforward when clinicalsymptoms and electrophysiology are clear, but maysometimes be controversial due to discrepancies in thedifferent measured clinical parameters. Imaging hasthe potential to resolve these discrepancies, as it pro-vides direct measures of the carpal tunnel (6–11). Mag-netic resonance imaging (MRI) enables high-resolutionvisualization of the nerve, through heavily T2-weightedimages, which can confirm CTS in patients with clini-cally suspect symptoms (10,11). However, MRI lacksspecificity and gives inconclusive evidence in cases withmild clinical symptoms where signal changes limit thenoise level.

Diffusion tensor imaging (DTI) is an MRI-based meth-odology in which the image contrast of neuronal fiberregions reflects local tissue architecture (12–17). DTIand its analysis tools were developed and first imple-mented on imaging of central nervous system (CNS)white matter, but theoretically can be applied to anytype of tissue. Common image indices that characterizeDTI are the apparent diffusion coefficient (ADC) (de-

1Department of Neurobiology, Faculty of Life Sciences, Tel Aviv Univer-sity, Tel Aviv, Israel.2Functional Brain Imaging Unit, Wohl Institute for Advanced Imaging,Tel Aviv Sourasky Medical Center, Tel Aviv, Israel.3Department of Computer Sciences, Sackler Faculty of Exact Sciences,Tel Aviv University, Tel Aviv, Israel.4Department of Physical Therapy, Sackler Faculty of Medicine, Tel AvivUniversity, Tel Aviv, Israel.5Department of Occupational Therapy, Sackler Faculty of Medicine, TelAviv University, Tel Aviv, Israel.Contract grant sponsor: Israel Science Foundation; Grant number:1280/04.*Address reprint requests to: Y.A., Department of Neurobiochemistry,Faculty of Life Sciences, Tel Aviv University, Tel Aviv, 69978, Israel.E-mail: asafyan@zahav.net.ilReceived December 27, 2007; Accepted July 11, 2008.DOI 10.1002/jmri.21553Published online in Wiley InterScience (www.interscience.wiley.com).

JOURNAL OF MAGNETIC RESONANCE IMAGING 29:657–662 (2009)

© 2009 Wiley-Liss, Inc. 657

fined as one-third of the trace of the diffusion tensor),the parallel and radial diffusivities to the fibers (D// andD�), and the fractional anisotropy, FA (12–17). An ad-ditional feature of DTI is the ability to reconstruct in 3Dthe path of large neuronal fiber trajectories, also knownas tractography (18–20). DTI was recently applied toreconstruct peripheral nerves (21–23). In this work weconducted a DTI study on the median nerve in thevicinity of the carpal tunnel in healthy and CTS sub-jects. The main aim of the study was to resolve the 3Dstructure of the nerve with DTI and compare the differ-ent DTI indices between healthy subjects and CTS pa-tients.

MATERIALS AND METHODS

Subjects

Nine CTS subjects, 30–65 years of age (three males, sixfemales; 44 � 13 years, median 38 years), diagnosed byan orthopedic surgeon as having severe CTS, agreed toparticipate in the study. Inclusion criteria includedpathological nerve conduction (from EMG), motor la-tency greater than 4.0 msec, sensory latency greaterthan 3.5 msec, and positive Phalen test (forcible palmarflexion of the wrist that leads to exacerbation of thesymptoms). A second group of 17 healthy subjects,aged 24–34 years (7 males, 10 females; 26 � 4 years,median 27 years), volunteered for the study. None of thehealthy volunteers had any of the CTS clinical symp-toms. Of the 17 healthy subjects, four datasets had tobe omitted because of severe motion along the scan thatprecluded full reconstruction of the nerve fiber. TheMRI protocol was approved by the local InstitutionalReview Board and all subjects signed an informed con-sent.

MRI

MRI was performed on a 3T whole-body Signa HorizonMRI system (GE Medical Systems, Milwaukee WI)equipped with a 40 mT/m gradient coil. The wholeforearm and palm were wrapped in a Fuorinert (FC-77,3M, St. Paul, MN) filled glove, on top of which a generalpurpose coil (5 � 6.5 inches, GE Medical Systems) waswrapped. The Fuorinert glove diminished the interfacebetween the forearm and surrounding air, significantlyreducing the susceptibility-induced artifacts as de-scribed by Neufeld et al (24).

The imaging protocol consisted of T2 and DTI imageseries. The T2-weighted images, scanned for anatomi-cal reference, were acquired using a fast-spin-echo se-quence with the following parameters: TR/TE �3500/85 msec, 36 slices of 3 mm thickness with no gapin between, a field of view (FOV) of 14 cm, and matrix of256 � 256 (in-plane resolution of 0.54 � 0.54 mm2).The number of averages was 2, echo train length was16, and fat suppression was enabled. The T2 seriesacquisition time was �2 minutes. The DTI series wasacquired using diffusion-weighted single-shot echo-planar imaging (DW-EPI) pulse sequence with the fol-lowing parameters: TR/TE � 10,000/81 msec, �/ �25/19 msec, b value of 1000 s/mm2 at six noncollineargradient directions ([1 0 1], [�1 0 1], [0 1 1], [0 1 �1], [1

1 0], [�1 1 0]). The number of averages was 2 andnumber of repetitions was 5 (sum of 10 acquisitionswith total acquisition time of �12 minutes). Ten acqui-sitions were needed to gain sufficient signal-to-noiseeven in low-signal areas. We chose to split these into 5repetitions of the 2 averages per scan to have betternoise averaging within each acquisition but to remainwith the ability to correct for hand movement during thescan. The slice geometry was the same as the T2 seriesbut with a matrix size of 64 � 64 (interpolated to 128 �128), giving final in-plane resolution of 1.09 � 1.09mm2. The use of anisotropic image voxels (slice resolu-tion is 3 times the in-plane resolution) could create apartial volume artifact in the slice direction, leading topoor fiber-tracking estimation and slightly reduced FAand ADC values. Nevertheless, the median nerve couldbe reconstructed in all subjects, indicating that theeffect of the slice thickness had only a minor effect onthe tractography (probably due to the perpendicularalignment of the fibers with respect to the slice direc-tion).

Preprocessing

Prior to DTI analysis, the five sets of DTI repetitionsunderwent realignment, averaging, and interpolation inorder to reduce motion artifacts and noise. Realignmentwas done using SPM2 (UCL, London, UK) with a 4thdegree b-spline interpolation with no wrap and nomask. Following realignment, images were resized inMatLab (MathWorks, Natick, MA) to matrix dimensionsof 128 � 128 using a bi-cubic interpolation. The five DTIrepetitions were then averaged to reduce noise. Sub-jects who moved their palm significantly during thescan (3 mm as indicated by the realignment processgraphs) were excluded from analysis to avoid additionalartifacts and image processing errors.

Image Analysis

Following successive preprocessing of the data, the im-ages were analyzed with the commonly used routinegiven by Basser et al (16) to produce for each subjectADC, D//, D�, and FA maps. The DTI analysis was alsoused to produce directionally encoded color maps and3D tractography maps of the median nerve. For trac-tography we used a modified FACT (fiber assignment bycontinuous tracking) algorithm (18,25,26). FACT fol-lows the vector’s angle found in each pixel throughstraightforward linear line propagation to its neighbor-ing pixel. A threshold value of FA 0.35 determineswhen the anisotropy is high enough to start the track-ing and a second threshold of FA � 0.2 is used as astopping threshold. FACT is applied here in the “brute-force” approach, which goes through all pixels, finds alltracts originating in each of them, and then choosesonly those going through the selected seed tractographyregion of interest (ROI) (25,26). The seed tractographyROIs were placed �3 cm proximal to the carpal tunneland distal to the carpal tunnel. The seed ROIs werechosen manually, surrounding the nerve area. Themain motivation to use tractrography was to have anobjective selection of the median nerve-related imagepixel with a minimum effect of partial volume effects.

658 Stein et al.

Postprocessing and Statistics

Successful fiber-tracking the nerve path was validatedon the anatomical images. We visually confirmed thatthe reconstructed fiber tract lay within the territory ofthe median nerve and did not deviate to ligament ormuscle areas (which are also directional tissues) (seeFig. 2A, red areas). In all CTS and 13 of the 17 healthysubjects the reconstructed nerve was within the ana-tomical territory of the median. In four healthy subjectsthe path of the nerve deviated from the anatomicalterritory and could not be fully reconstructed due toexcessive motion during the test (see realignmentabove). In addition, we marked negative eigenvaluesthat often appear in DTI, most probably due to noise ormotion artifact, and result in erroneous estimation ofthe diffusion indices along the nerve. In all subjects, nonegative eigenvalues appeared in the path of the recon-structed nerve.

Following successful validation of the tractography,we then extracted the different diffusion indices alongthe tract (ADC, D//, D�, and FA). For this purpose wechose four locations along the path of the nerve: loca-tion 1: �3.0 cm proximal to the carpal tunnel; location2: �1.5 cm proximal to the carpal tunnel; location 3: atthe carpal tunnel; location 4: 1.5 cm distal to the carpaltunnel. For that purpose we visually identified the car-pal tunnel location for each subject. Each of the ana-lyzed nerve segments consisted of five slices of 3 mm fora total of 1.5 cm. The locations of the segments areillustrated in Fig. 2B. A Mann–Whitney U-test was usedto compare the indices of the different locations withineach of the two groups (healthy subjects vs. CTS), or perlocation between the two groups. P � 0.05 was consid-ered significant.

RESULTS

Qualitative Observations

The DTI dataset demonstrated the directional arrange-ment of the median nerve compared with other parts ofthe palm and forearm anatomy (Fig. 1). When the dif-fusion was measured approximately perpendicular tothe nerve fiber, its signal was higher in the diffusion-weighted image compared with the surrounding tissues(Fig. 1A, [1 1 0] and [�1 1 0] directions). The direction-ally color-coded FA maps demonstrate the directionaland fibrous arrangement of the median nerve (Fig. 1C),appearing as a blue line, an indication of fibers crossingparallel to the forearm. In addition, examination of thenerve’s appearance on the FA maps (Fig. 1D) show thatthe FA signal appears brighter (higher FA, cyan arrows)at the area of the carpal tunnel than distal or proximalto it. The 3D reconstruction of the nerve was achievedusing the tractography process as given for a represen-tative healthy subject in Fig. 2A,B along with the 3Dreconstruction of the palm and bones.

Figure 3 shows a typical dataset of a CTS patientdepicting the high-resolution T2, FA, directionally col-or-coded FA, parallel and radial diffusivity, and ADCmaps. Similar to the healthy subjects, the nerve in theCTS patients could be easily separated from surround-ing tissue in the FA, color-coded, and radial diffusivity

maps (Fig. 4B–D). But, unlike in the healthy subjects,the FA index in CTS patients decreased when the nervepassed through the carpal tunnel (denoted by the cyanarrow in Fig. 4C); At this location the radial diffusivityappeared to increase.

Quantitative Analysis

Figure 4 shows the quantitative analysis of the fourlocations along the median nerve in healthy subjects

Figure 1. Diffusion tensor imaging of the median nerve. a:Diffusion-weighted images of the six gradient directions (indi-cated above each image) given at two sections (axial, top; sag-ittal, bottom) of the left hand of a healthy volunteer. Note thatin the approximate perpendicular gradient directions ([1 1 0],[�1 1 0]), the median nerve signal is enhanced (yellow arrows).b: A fat-saturated high-resolution T2-weighted image at thesame slice locations as in (A). c,d: Directionally color-coded FAand FA maps, respectively, at the same slice locations. Thegray bar at the right of (D) refers to (D). In the directionallycolor-coded FA map, blue represent of anisotropic diffusionaligned at the superior–inferior direction, red at left to rightdirection and green at anterior–posterior direction. Note thehomogenous appearance of the median nerve in these twoimages (the blue line in (C) indicated by the yellow arrows). Inaddition, the FA signal appears to increase at the area of thecarpal tunnel (cyan arrows in (D)). The radial diffusivity (e),parallel diffusivity (f), and apparent diffusion coefficient (f)maps of the same slice locations as above, respectively. Thegray bar at the right of (G) refers to (E,F,G).

DTI of the Median Nerve 659

and CTS patients for each of the four DTI indices (FA,radial diffusivity, parallel diffusivity, and ADC).

The DTI indices changed significantly along the pathof the nerve in both the healthy and in the CTS groups,

but in opposite trends. In the healthy group the FAincreased when the fibers passed through the carpaltunnel (�20%, P � 0.001, see Fig. 4A, full gray bars), anincrease accompanied by a decrease in ADC values(�8%, P � 0.05, Fig. 4B). The origin of these changeswas rooted in the radial and parallel diffusivities: whilethe parallel diffusivity showed no change along the dif-ferent nerve locations (Fig. 4D), the radial diffusivitywas significantly reduced when passing through thecarpal tunnel (�15%, P � 0.01, Fig. 4C).

In contrast to the healthy subject, the quantitativeanalysis of the four nerve locations in the CTS grouprevealed that the FA decreased when the nerve passedthrough the carpal tunnel (�21%, P � 0.05, Fig. 4A,bars with diagonal stripes) and the radial diffusivityincreased (23%, P � 0.001, Fig. 4C). The parallel diffu-sivity and ADC showed no significant changes along thenerve for the CTS group.

Due to the opposite trends in changes along the nervebetween the healthy and CTS group, a comparison be-tween the groups yielded the following results: In thefirst location, the most proximal to the forearm, thehealthy and CTS subjects showed similar DTI indices(Fig. 4); When passing through the carpal tunnel (loca-tion 3 in Fig. 4), significant differences were observed:the FA decreased in the CTS subjects by 31% (P �

Figure 2. Tractography of the median nerve. a: Front view of a3D reconstruction of the palm with the palm bones (in light blue)and the reconstructed nerve (red). b: Side view of the same imagein (A) with an indication of the four nerve locations used forquantitative analysis (see Materials and Methods). The data pre-sented in this figure are from the same subject as in Fig. 1.

Figure 3. Diffusion tensor imaging of the median nerve of aCTS patient. a: A fat-saturated high-resolution T2-weightedimage at the sagittal section of the right hand of a CTS patientdepicting an enlarged nerve (see arrow). b,c: Color-coded FAand FA maps, respectively, at the same slice locations as in (A).The gray bar at the right of (C) refers to (C). In the directionallycolor-coded FA map, blue represents anisotropic diffusionaligned at the superior–inferior direction, red at left to rightdirection, and green at anterior–posterior direction. Note thatthe FA signal appears to decrease when approaching the loca-tion of the carpal tunnel (cyan arrows). d–f: The radial diffu-sivity, parallel diffusivity, and apparent diffusion coefficientmaps, respectively, of the same slice locations as above. Thegray bar at the right of (F) refers to (D–F).

Figure 4. Quantitative analysis of the DTI indices along themedian nerve of healthy and CTS patients. (a) FA, (b) ADC, (c),radial diffusivity, and (d) parallel diffusivity along the fourlocations. Healthy subject are represented by the full gray barsand CTS subjects by diagonal striped bars. Error bars repre-sent the standard deviation. Asterisk represent the between-group comparison for each location where: *P � 0.05, **P �0.001, and ***P � 0.0001. The within-group statistics are notgiven in the graph. In the healthy group the FA at locations 2,3, and 4 were significantly different from location 1 (P � 0.001)as well as the ADC (P � 0.05) and radial diffusivity (P � 0.001).In the CTS group the FA and radial diffusivities at the carpaltunnel (location 3) were significantly different from all otherlocations (P � 0.05).

660 Stein et al.

0.0001) while the perpendicular diffusivity increased bymore than 55% (P � 0.0001) compared with the healthysubjects; The ADC also differed between the twogroups, increasing by 19% (P � 0.0001) in the CTSgroup while the parallel diffusivity showed no changebetween the two groups.

DISCUSSION

The MRI role in evaluation of CTS is limited due to lowaccuracy, especially in mild CTS where signal changeslimit the noise levels. With the findings of the presentedstudy, diffusion MRI turned out to be sensitive to CTS.This opens the possibility of using DTI for single-sub-ject diagnosis of CTS, and thus increases the role of MRIin the evaluation of suspected CTS patients. The find-ings of this study indicate that patients with character-istic CTS clinical symptoms can be effectively differen-tiated from healthy subjects using DTI with themethods described above. The most sensitive parame-ters are the FA and radial diffusivity.

The changes in the different diffusion parameters inhealthy volunteers along the nerve—a significant in-crease in FA and a significant reduction in the perpen-dicular diffusivity and ADC—fit a model in line with theforearm’s known anatomy. As the nerve goes from afree, lucid form proximal to the carpal tunnel, it grad-ually compresses in the tunnel (1,2). It is likely that thiscompression does not change the number of fibers, butmost probably reduces the amount of extracellularspace (ie, increases fiber density), limiting the space inwhich diffusion of water molecules can take place in.While parallel diffusivity remains almost undisturbedin this model, perpendicular diffusivity, now limited tothe axon’s diameters with small margins around them,is much more restricted.

The DTI-based parameter values of the CTS patientgroup differ significantly from those of the healthy vol-unteers group at the location of the carpal tunnel. Asthe nerve approaches the carpal tunnel, the perpendic-ular diffusivity significantly increases, evidenced by thedecreasing FA and increasing ADC, indicating that dif-fusion in the tissue has become more isotropic. Oneexplanation for such an observation is edema, in whichan excess amount of fluid is trapped in the tissue.Edema also enlarges the extracellular matrix, creatingan isotropic environment that is manifest in a rise inperpendicular diffusivity and a reduction in FA. Con-ventional imaging, either with ultrasound or MRI, oftenshows that the nerve is enlarged (higher cross-sectionalarea), supporting the edema explanation (27,28). In ad-dition, it was found that DTI indices correlate with ex-tracellular space volume fraction, supporting the aboveexplanation (29,30).

Although it is theoretically feasible to study periph-eral nerve morphometry by DTI, there are very few suchstudies (21–23) for a number of reasons: First, the ge-ometry of limbs combined with fast MRI acquisitiontechniques (EPI) leads to severe image distortions thatmay exclude the ability to reliably define the nerve.Second, the small diameter of the nerve causes partialvolume effects where pixels identified as nerve mayhave contributions from surrounding tissue; Third, DTI

has known pitfalls, especially noise sensitivity andnegative diffusivity artifacts. In this work we showthat one can overcome these issues: The image dis-tortions can be minimized using a specially designedapparatus that reduces susceptibility-induced arti-facts (24). The partial volume effect can be minimizedusing the tractography analysis identifying pixelsthat very likely represent the nerve and not surround-ing tissue. Lastly, although noise will always be aproblem in peripheral nervous system (PNS) DTI,careful analysis of the data can help in evaluating thenegative diffusivities and avoid their contribution tothe quantitative analysis.

An additional source of bias in the results of thepresent study may be related to the different mean agesof the healthy and CTS groups. Although normal agingmay influence the appearance of the nerve, the fact thatproximal to the nerve, in all measured indices, similarvalues were extracted for both groups (Fig. 4) dimin-ishes the age bias effect.

Determining the sensitivity of DTI to CTS requires amore comprehensive study, one that takes into accounta much larger number of CTS patients with a higherdegree of variability of clinical signs. Clinically detect-able edema does not always accompany CTS, so carefulstudy must be made of CTS patients to determinewhether and to what degree abnormal DTI values arerelated to edema.

In addition to diagnosis of CTS, DTI of the forearmcan also play an important role in other PNS patholo-gies. These include trauma injuries; assessment of thesituation of a damaged nerve on which a tumor isfound; evaluating the loss of myelin in Guillain–Barresyndrome; and tracking the recovery of the nerve after ithas been placed in a graft.

DTI of the PNS can also enhance our understandingof the underlying process of diffusion in neuronal tissuein general, and neuronal fibers in particular. For exam-ple, the manner in which myelin affects diffusivity ofwater molecules from the axons it envelops is still notfully known. The relatively simple structure of the PNScompared to the CNS offers a unique opportunity toinvestigate the complex behavior of water diffusionthrough myelin.

In most cases, CTS diagnosis is straightforward, andtherapeutic approaches and patient management canbe easily decided. Nevertheless, occasionally the clini-cal, electrophiological, and imaging measures leads toinconclusive estimation of the severity of CTS, if any.The highly significant differences found with DTI be-tween healthy and CTS groups (P � 0.0001) in thisstudy demonstrate that this method can be used as anadditional criterion for evaluation of nerve condition inCTS. Although only definite, severe cases of CTS wereevaluated in this study, the large differences betweenthe groups implies that this methodology is highly sen-sitive to nerve pathophysiological condition and thusmight help to evaluate less severe or controversialcases. Our results point out that the FA and radialdiffusivity indices are the most sensitive ones for eval-uation of CTS.

DTI of the Median Nerve 661

REFERENCES1. Seeley RR, Stephens TD, Tate P. Anatomy & physiology, 7th ed.

Boston: McGraw Hill, 2006.2. Gray H, Standring S. Gray’s anatomy: the anatomical basis of

clinical practice, 39th ed. Edinburgh: Elsevier Churchill Living-stone, 2005.

3. Sternbach G. The carpal tunnel syndrome. J Emerg Med 1999;17:519–523.

4. Michelsen H, Posner MA. Medical history of carpal tunnel syn-drome. Hand Clin 2002;18:257–268.

5. Kotevoglu N, Gulbahce-Saglam S. Ultrasound imaging in the diag-nosis of carpal tunnel syndrome and its relevance to clinical eval-uation. Joint Bone Spine 2005;72:142–145.

6. Atroshi I, Gummesson C, Johnsson R, Ornstein E, Ranstam J,Rosen I. Prevalence of carpal tunnel syndrome in a general popu-lation. JAMA 1999;282:153–158.

7. Lee D, van Holsbeeck MT, Janevski PK, Ganos DL, Ditmars DM,Darian VB. Diagnosis of carpal tunnel syndrome. Ultrasound ver-sus electromyography. Radiol Clin North Am 1999;37:859–872, x.

8. Wilder-Smith EP, Seet RC, Lim EC. Diagnosing carpal tunnel syn-drome—clinical criteria and ancillary tests. Nat Clin Pract Neurol2006;2:366–374.

9. Wiesler ER, Chloros GD, Cartwright MS, Smith BP, Rushing J,Walker FO. The use of diagnostic ultrasound in carpal tunnel syn-drome. J Hand Surg [Am] 2006;31:726–732.

10. Howe FA, Filler AG, Bell BA, Griffiths JR. Magnetic resonanceneurography. Magn Reson Med 1992;28:328–338.

11. Cudlip SA, Howe FA, Clifton A, Schwartz MS, Bell BA. Magneticresonance neurography studies of the median nerve before andafter carpal tunnel decompression. J Neurosurg 2002;96:1046–1051.

12. Le Bihan D, Mangin JF, Poupon C, et al. Diffusion tensor imaging:concepts and applications. J Magn Reson Imaging 2001;13:534–546.

13. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectros-copy and imaging. Biophys J 1994;66:259–267.

14. Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffu-sion anisotropy. Magn Reson Med 1996;36:893–906.

15. Basser PJ, Jones DK. Diffusion-tensor MRI: theory, experimentaldesign and data analysis — a technical review. NMR Biomed 2002;15:456–467.

16. Basser PJ, Pierpaoli C. A simplified method to measure the diffu-sion tensor from seven MR images. Magn Reson Med 1998;39:928–934.

17. Pajevic S, Pierpaoli C. Color schemes to represent the orientation ofanisotropic tissues from diffusion tensor data: application to whitematter fiber tract mapping in the human brain. Magn Reson Med2000;43:921.

18. Mori S, van Zijl PC. Fiber tracking: principles and strategies — atechnical review. NMR Biomed 2002;15:468–480.

19. Basser PJ, Pajevic S, Pierpaoli C, Duda J, Aldroubi A. In vivo fibertractography using DT-MRI data. Magn Reson Med 2000;44:625–632.

20. Catani M, Howard RJ, Pajevic S, Jones DK. Virtual in vivo interac-tive dissection of white matter fasciculi in the human brain. Neu-roimage 2002;17:77–94.

21. Hiltunen J, Suortti T, Arvela S, Seppa M, Joensuu R, Hari R.Diffusion tensor imaging and tractography of distal peripheralnerves at 3 T. Clin Neurophysiol 2005;116:2315–2323.

22. Skorpil M, Karlsson M, Nordell A. Peripheral nerve diffusion tensorimaging. Magn Reson Imaging 2004;22:743–745.

23. Meek MF, Stenekes MW, Hoogduin HM, Nicolai JP. In vivo three-dimensional reconstruction of human median nerves by diffusiontensor imaging. Exp Neurol 2006;198:479–482.

24. Neufeld A, Assaf Y, Graif M, Hendler T, Navon G. Susceptibility-matched envelope for the correction of EPI artifacts. Magn ResonImaging 2005;23:947–951.

25. Jiang H, van Zijl PC, Kim J, Pearlson GD, Mori S. DtiStudio:resource program for diffusion tensor computation and fiberbundle tracking. Comput Methods Programs Biomed 2006;81:106–116.

26. Huang H, Zhang J, van Zijl PC, Mori S. Analysis of noise effects onDTI-based tractography using the brute-force and multi-ROI ap-proach. Magn Reson Med 2004;52:559–565.

27. Mesgarzadeh M, Schneck CD, Bonakdarpour A, Mitra A, ConawayD. Carpal tunnel: MR imaging. Part II. Carpal tunnel syndrome.Radiology 1989;171:749–754.

28. Seyfert S, Boegner F, Hamm B, Kleindienst A, Klatt C. The value ofmagnetic resonance imaging in carpal tunnel syndrome. J Neurol1994;242:41–46.

29. Assaf Y, Freidlin RZ, Rohde GK, Basser PJ. New modeling andexperimental framework to characterize hindered and restrictedwater diffusion in brain white matter. Magn Reson Med 2004;52:965–978.

30. Schwartz ED, Cooper ET, Fan Y, et al. MRI diffusion coefficients inspinal cord correlate with axon morphometry. Neuroreport 2005;16:73–76.

662 Stein et al.