NeuroImage 47 (2009) 451–458

Contents lists available at ScienceDirect

NeuroImage
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y n i m g

A longitudinal diffusion tensor imaging study on Wallerian degeneration of

corticospinal tract after motor pathway stroke
Chunshui Yu a,⁎, Chaozhe Zhu b, Yujin Zhang b, Hai Chen c, Wen Qin a, Moli Wang c, Kuncheng Li a
a
Department of Radiology, Xuanwu Hospital, Capital Medical University, No. 45, Chang-Chun Street, Xuanwu District, Beijing 100053, PR China
b
State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing 100875, PR China
c
Department of Neurology, Xuanwu Hospital, Capital Medical University, Beijing 100053, PR China

a r t i c l e i n f o a b s t r a c t

Article history: Wallerian degeneration of the corticospinal tract (CST) after motor pathway ischemic stroke can be
Received 5 March 2009 characterized by diffusion tensor imaging (DTI). However, the dynamic evolution of the diffusion indices in
Revised 5 April 2009 the degenerated CST has not previously been completely identified. We investigated this dynamic evolution
Accepted 20 April 2009
and the relationship between early changes of the diffusion indices in the degenerated CST and long-term
Available online 3 May 2009
clinical outcomes. DTI and neurological examinations were performed repeatedly in 9 patients with first-
Keywords:
onset motor pathway subcortical infarction at 5 consecutive time points, i.e. within 1 week, at 2 weeks,
Stroke 1 month, 3 months and 1 year. Using a region of interest method, we analyzed the ratios of the fractional
Wallerian degeneration anisotropy (rFA), mean diffusivity (rMD), primary eigenvalue (rλ1) and transverse eigenvalue (rλ23)
Corticospinal tract between the affected and unaffected sides of the CSTs. We did not find any significant changes in the
Diffusion tensor imaging diffusion indices of the contralesional CSTs across time points. The rFA decreased monotonously during the
Magnetic resonance imaging first 3 months and then stabilized. The rMD increased after 2 weeks and stabilized after the third month. The
rλ1 decreased during the first 2 weeks and then remained unchanged. The rλ23 increased during the first
3 months and then stabilized. We also found that the changes in the rFA between the first 2 time points were
correlated with the NIHSS (P = 0.00003) and the Motricity Indices (P = 0.0004) after 1 year. Our results
suggest that for patients with motor pathway stroke the diffusion indices in the degenerated CST stabilize
within 3 months and that early changes in the rFA of the CST may predict long-term clinical outcomes.
© 2009 Elsevier Inc. All rights reserved.

Introduction
Degeneration of the distal parts of nerves after injury to the
proximal axon or cell body is referred to as Wallerian degeneration
(WD), which occurs in both peripheral and central nervous systems.
WD was first reported by Waller (1850) who found this pattern of
degeneration after cutting the glossopharyngeal and hypoglossal
nerves of the frogs. In the central nervous system, WD is characterized
by a highly stereotypical course, starting with disintegration of the
axonal skeleton and membrane within days after injury, followed by
degradation of the myelin sheath and infiltration by macrophages and
microglia, with subsequent atrophy of the affected fiber tracts
(Johnson et al., 1950; Lampert and Cressman, 1966). WD of the
corticospinal tract (CST) after motor pathway ischemic stroke is a
well-known phenomenon, which can be characterized by diffusion
tensor imaging (DTI).
DTI, a non-invasive MRI technique, measures the random motion
of water molecules and provides information about cellular integrity
and pathology (Le Bihan 2003). Within a highly ordered white
matter tract, water molecules diffuse faster in the direction parallel to
⁎ Corresponding author. Fax: +86 10 8319 8376.
E-mail address: chunshuiyu@yahoo.cn (C. Yu).
the tract than in the perpendicular directions because axonal mem-
branes and myelin sheaths restrict transverse diffusion (Beaulieu
and Allen, 1994; Wimberger et al., 1995). Pathological processes
which change the microstructural environment, such as neuronal
size, extracellular space and tissue integrity, result in altered
diffusion (Anderson et al., 1996; Sevick et al., 1992). Molecular
diffusion can be measured by several indices derived from DTI data.
Mean diffusivity (MD) and fractional anisotropy (FA) are commonly
used indices that reflect the average amplitude and the directionality
of molecular motion, respectively (Pierpaoli and Basser, 1996). They
are calculated from 3 diffusion tensor eigenvalues, which are the
diffusion coefficients along the major, medium and minor axes of the
diffusion ellipsoid. The primary eigenvalue (λ1) is the diffusion
coefficient along the direction of maximum diffusion. The transverse
eigenvalue (λ23) is generated by averaging the medium (λ2) and
minimum eigenvalues (λ3), thus avoiding sorting bias due to similar
magnitudes of λ2 and λ3 (Basser and Pajevic, 2003). This measure
reflects the average diffusivity perpendicular to the direction of
maximum diffusion.
Two popular analytic approaches for detecting changes in the
diffusion indices of a specific fiber tract are the pure region of
interest (ROI) analysis and the tract ROI analysis. In the pure ROI
analysis the ROIs are manually defined on MR images. This method

1053-8119/$ – see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2009.04.066
452 C. Yu et al. / NeuroImage 47 (2009) 451–458

is easy to perform, but the definition of the ROI must be carefully
designed to improve the reproducibility, because the boundary of a
fiber tract is often ill-defined. In tract ROI analysis defining the ROI
of a fiber tract is based on tractography, which can either be
performed by considering a fiber tract as an entire ROI or by
defining a segment of the fiber tract as a ROI (Berman et al., 2005;
Lin et al., 2006; Pagani et al., 2005; Partridge et al., 2005; Yu et al.,
2007, 2008). In this study, we analyzed the CST using both
methods.
Using a non-invasive approach to characterize and stage WD in
the CST is clinically important for patients with motor pathway
stroke. A few previous studies have delineated the process of WD
using DTI (Pierpaoli et al., 2001; Thomalla et al., 2004, 2005). At the
chronic stage (more than 1 year) after stroke, the degenerated CST
showed a sharp decrease in FA, a slight increase in MD, a decrease in
λ1, and an increase in λ23 (Pierpaoli et al., 2001). At the early stage
(within 2 weeks), the degenerated CST demonstrated a decrease in
FA and λ1, an unchanged MD, and an increased λ23 (Thomalla et al.,
2004). A single pioneering study on 2 stroke patients who were
examined at three time points (TP), is the only one to date that has
attempted to describe the time course of WD (Thomalla et al., 2005).
Thus, the detailed time course of WD after stroke still needs to be
investigated.
In previous DTI studies of WD of the CST, the ratios of the diffusion
indices (rFA, rMD, rλ1 and rλ23) between the affected and unaffected
sides of the CSTs have commonly been used based on the hypothesis
that diffusion indices of the contralesional CST are unchanged after
stroke. However, a recent study has reported that the FA of the
normal-appearing white matter of the whole brain was increased
within 2 years after stroke, apparently suggesting a long-term
improvement in apparent white matter integrity following ischemic
stroke (Wang et al., 2006). Therefore, it is critically important to study
the possible plasticity in the contralesional CST before using the ratios
(rFA, rMD, rλ1 and rλ23) to study the dynamic changes of the
degenerated CST.
Impairment of motor function is one of the most serious disabling
sequelae of ischemic stroke. Thus an early prediction of the long-term
outcome of motor function is critically important for stroke treatment
and rehabilitation. Two previous studies (Maeda et al., 2005;
Thomalla et al., 2004) attempted to predict the motor outcome at
3 months after stroke using the ratios of the diffusion indices of the
CSTs from an earlier time point (TP). However, this method may be
confounded by individual side differences in the diffusion indices of
the CSTs before the stroke. Thus we propose that it seems to be more
appropriate to use the changes in the ratios of the diffusion indices
between 2 earlier TPs (i.e. within 1 week and at 2 weeks) to predict
the long-term motor outcomes.
Here, we aimed to investigate (1) whether the contralesional CST
shows plastic changes following motor pathway ischemic stroke; (2)
the dynamic evolution of the diffusion indices in the degenerated CST;
and (3) whether the clinical outcomes could be predicted by the early
changes in the diffusion indices.
Subjects and methods
Subjects
Nine right-handed (Oldfield, 1971) patients (all males; age 48 ±
5 years, range 41–53 years) with motor pathway subcortical infarction
were recruited from the inpatient services at the Xuanwu Hospital of
Capital Medical University. All patients were first-onset stroke and
manifested motor deficits. None of them had a history of any other
neurological or psychiatric disorders. Conventional MR images did not
find any abnormalities except for the infarct in patients. Neurological
examinations included the National Institutes of Health Stroke Scale
(NIHSS) (Brott et al., 1989), and the Motricity Index (MI) of the
affected side (Demeurisse et al., 1980). They were scanned and
clinically assessed at 5 consecutive TPs, i.e. within 1 week, at 2 weeks,
at 1 month, at 3 months and at 1 year. The demographical and clinical
characteristics of stroke patients were summarized in Table 1. The
location of the lesion in each patient was shown in Fig. 1. We enrolled
9 control subjects with no history of stroke and a normal neurological
examination, and well matched to the stroke patients with regard to
age (48 ± 5 years, range 41–53 years), gender (9 males) and hand
dominance (all right-handed). All aspects of the study received
approval from the Local Ethical Committee and all subjects provided
full written informed consent about all aspects of the study before the
MR examinations.
MRI examination
We examined all subjects with a 3.0 T MR scanner (Trio system;
Siemens Magnetom scanner, Erlangen, Germany). The DTI scheme
included the collection of 12 images with non-collinear diffusion
gradients (b = 1000 s/mm2) and 1 non-diffusion-weighted image
(b = 0 s/mm2), employing a single shot echo planar imaging
sequence. Using an integrated parallel acquisition technique (iPAT)
with an acceleration factor of 2 reduced the acquisition time and
allowed us to obtain images with less distortion from susceptibility
artifacts. We collected 45 slices from each participant. The field of
view was 256 × 256 mm, the acquisition matrix was 128 × 128 and was
zero filled into a 256 × 256 matrix, the number of averages was 3, and
the slice thickness was 3 mm without a gap, which resulted in voxel-
dimensions of 1 mm × 1 mm × 3 mm. The echo time and repetition
time were 87 ms and 6000 ms, respectively. To maintain the
consistency of the slices among the 5 scans, we consistently
positioned the slices of each scan parallel to a line that joined the
most infero-anterior and infero-posterior parts of the corpus
callosum, and attempted to keep the central slices of the last 4 scans
consistent with that of the first scan.

Table 1
Demographic, clinical and radiological data from ischemic stroke patients.

Case Sex Age Lesion Time points (days after stroke) NIHSS Motricity index
(years) Location Volume (ml) 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
1 M 49 R IC, BG, CR 12.24 6 20 33 90 244 2 2 2 2 2 72 94 94 99 99
2 M 42 L CR, BG 24.77 5 14 32 145 355 10 3 2 0 0 17 44 65 95 95
3 M 48 L IC, BG, CR 23.64 2 13 35 88 298 14 11 10 8 5 0 7 10 41 48
4 M 53 L IC, BG, CR 12.08 4 18 35 97 352 8 6 3 2 2 7 29 44 57 57
5 M 52 L CR, BG 9.95 1 14 30 90 370 5 2 2 0 0 71 92 99 99 99
6 M 51 L IC, BG, CR 3.54 7 14 30 95 425 10 8 8 5 2 7 19 24 44 58
7 M 42 R CR, BG 7.04 2 13 31 98 323 18 12 8 9 7 0 0 14 17 29
8 M 41 L IC, BG, CR 31.86 6 15 30 110 376 15 13 13 6 6 0 7 17 39 42
9 M 52 R IC 1.73 5 16 25 92 353 9 8 6 3 3 14 24 34 42 42

M, male; L, left; R, right; IC, internal capsule; BG, basal ganglia; CR, corona radiate; NIHSS, National Institutes of Health Stroke Scale.
 C. Yu et al. / NeuroImage 47 (2009) 451–458 453

Fig. 1. Lesion location in each stroke patient. Lesion of each stroke patient is shown on a T2 weighted image with the largest area of the lesion. Arabic numbers denote the case
numbers of the stroke patients.

Data preprocessing
DTI data were preprocessed by the following procedures: (1) all
diffusion-weighted images were visually inspected by 2 radiologists
for apparent artifacts due to subject motion and instrument malfunc-
tion; (2) for each examination of each subject, the diffusion-weighted
images were registered to the corresponding b = 0 images with an
affine transformation to correct for their eddy-current distortion and
head motion; (3) for each subject, each of the follow-up b = 0 images
was aligned to the corresponding b = 0 image at the first TP using an
affine transformation and the derived transformation matrix was
applied to the corresponding parametric maps, such as the FA, MD, λ1,
andλ23 maps; (4) DTI data were interpolated into 1 mm × 1 mm ×
1 mm voxels; (5) the diffusion tensor of each voxel was calculated by a

Fig. 2. Examples for segments of the corticospinal tract and the corpus callosum and region of interest in the corticospinal tract. (A) The segment of the corticospinal tract is a part of
the corticospinal tract from the uppermost slice of the cerebral peduncle to the lowest slice of the pons. (B) The segment of the corpus callosum is defined as the part of the corpus
callosum connecting the bilateral occipital lobes. (C) The white circles show the region of interest of the corticospinal tract.
454 C. Yu et al. / NeuroImage 47 (2009) 451–458

Table 2
Normalized diffusion indices of the contralesional sCSTs in stroke patients and healthy controls.

Diffusion Controls Stroke patients (n = 9)

indices (n = 18) 1 2 3 4 5
nFA 1.04 ± 0.17 0.97 ± 0.22 0.96 ± 0.19 0.96 ± 0.18 0.94 ± 0.22 0.96 ± 0.19
nMD 0.96 ± 0.17 0.98 ± 0.17 0.98 ± 0.15 0.99 ± 0.16 1.02 ± 0.20 1.04 ± 0.20
nλ1 0.94 ± 0.16 0.94 ± 0.16 0.94 ± 0.15 0.94 ± 0.15 0.95 ± 0.17 0.96 ± 0.18
nλ1 0.98 ± 0.24 1.03 ± 0.23 1.03 ± 0.20 1.05 ± 0.20 1.10 ± 0.27 1.13 ± 0.25

sCST, segment of the corticospinal tract; nFA, normalized fractional anisotropy; nMD, normalized mean diffusivity; nλ1, normalized primary eigenvalue; nλ23, normalized transverse
eigenvalue. Diffusion indices of the contralesional sCSTs were normalized by the diffusion indices of the segment of the corpus callosum.

linear least-square fitting algorithm (Basser et al., 1994). After
diagonalization of the diffusion tensor, the diffusion tensor eigenva-
lues (λ1, λ2 and λ3) were obtained; the λ23 was generated by
averaging λ2 and λ3. The MD and FA were derived for each voxel
according to the following equations:
We projected the sCST and sCC that were obtained in the above
manner to the coregistered parameter maps of the other 4 TPs.
Quantitative analysis was performed by defining all voxels penetrated
by a tract as the tract ROI. Then we calculated the FA, MD, λ1 and λ23
for these tracts for each TP of each participant.

λ1 + λ2 + λ3 Dynamic changes in the diffusion indices in the degenerated CST

MD =
3
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðλ1 −λ2 Þ2 + ðλ1 − λ3 Þ2 + ðλ2 −λ3 Þ2 Tract ROI analysis cannot precisely characterize the dynamic
FA = : changes in the degenerated CST because of the effects of atrophy of
2ðλ1 + λ2 + λ3 Þ2
the affected CST in the late TPs. Thus a pure ROI method was adopted
Here MD measures the average magnitude of the molecular to study the dynamic changes in the diffusion indices in the
motion and FA measures its directionality. All the preprocessing degenerated CST. We established the ROI on the FA images at the
procedures were implemented using FMRIB's free software FSL fifth TP and then copied this ROI to the other TPs. This ensured that we
(Oxford Centre for Functional MRI of the Brain, UK). always located the ROI within the CST in any TP. Specifically, we placed
the ROI at the middle slice of the pons on the axial plane (Fig. 2C).
Plasticity of contralesional CST Finally, we used the ratios (rFA, rMD, rλ1 and rλ23) between the
affected and unaffected CST-ROIs to study the dynamic changes of the
To investigate whether the contralesional CST undergoes plastic diffusion indices in the affected CST.
changes after motor pathway ischemic stroke, we adopted a normal-
ized tract ROI analysis using the ratios of the diffusion indices between Reproducibility analysis
a segment of the CST (sCST, Fig. 2A) from the uppermost slice of the
cerebral peduncle to the lowest slice of the pons and a segment of the Since a single rater manually defined the ROIs for the pure ROI
corpus callosum (sCC, Fig. 2B) connecting the bilateral occipital lobes. analysis, we determined the intra-rater reproducibility for the
The reasons that we studied the sCST are as follows: (1) the fibers of positioning of the ROIs by re-defining these ROIs and re-measuring
the sCST are highly oriented, which is suitable for the assessment by the FA, MD, λ1 and λ23 of each ROI at the fifth TP in all patients on 2
DTI; (2) to establish the comparability across subjects; (3) the sCST separate occasions (separated by at least 3 months).
was not directly involved by the lesions of stroke patients. We
normalized the diffusion indices of the sCST by the sCC because this Statistical analysis
part of the corpus callosum was relatively reproducible and was not
affected by the lesions. This method may help to reduce the influence We analyzed the data with SPSS 13.0 for Windows (SPSS Inc, Chicago,
of the slight instability of the MR scanner. Ill). All data were first examined for normality to conform to the
Tractography for the CC and CST was implemented using DTI assumptions of the parametric statistics used. We tested the side
Studio (free software from the Radiology Department, Johns Hopkins differences in the diffusion indices of the sCST in healthy controls using
University, USA), in which the fiber assignment by means of an independent-samples t test, but found no significant differences
continuous tracking (Mori et al., 1999; Mori and Van Zijl 2002) was (P N 0.05). Thus we merged them to increase statistical power. To
used with a FA threshold of 0.2 and angle transition threshold of investigate possible plasticity in the contralesional CST in stroke
45°during tracking. To reconstruct the tracts of interest, we used a patients, we performed a one-way analysis of variance (ANOVA) on
multiple-ROI approach (Catani et al., 2002; Wakana et al., 2004) based the normalized diffusion indices of the contralesional sCST (sCST/sCC)
on existing anatomic knowledge about tract trajectories. We
performed tracking from all voxels inside the brain, and assigned
Table 3
the results that penetrated the manually defined ROIs to the specific Comparisons of diffusion indices between the affected and unaffected CSTs at the first
tracts associated with these ROIs. time point in stroke patients using a paired samples t test.
We used DTI data from the first TP to trace the fibers since the
Diffusion indices Affected side Unaffected side t value P
diffusion indices of the degenerated CST had not yet changed signi-
Tract ROI analysis
ficantly at this stage, and placed the seed ROI for tracing the CST in the
FA 0.42 ± 0.09 0.44 ± 0.06 0.662 0.526
cerebral peduncle. Then we used a filter ROI, which was placed in the MD (× 10− 4 mm2s− 1) 9.15 ± 1.24 8.81 ± 1.34 0.445 0.668
anterior part of the pons, to exclude non-CST fibers. Other non-CST fibers λ1 (× 10− 4 mm2s− 1) 13.43 ± 1.14 13.16 ± 1.59 0.324 0.754
were excluded using the “not” function of the software. The sCST was a λ23 (× 10− 4 mm2s− 1) 7.01 ± 1.48 6.63 ± 1.38 0.491 0.637
Pure ROI analysis
segment of the traced CST from the uppermost slice of the cerebral pe-
FA 0.53 ± 0.10 0.55 ± 0.08 − 1.171 0.275
duncle to the lowest slice of the pons (Fig. 2A). To trace the sCC we used MD (× 10−4 mm2s− 1) 7.22 ± 0.78 7.49 ± 1.21 − 1.095 0.306
the posterior part of the CC on the midsagittal FA image as the first seed λ1 (× 10−4 mm2s− 1) 11.87 ± 0.56 12.53 ± 1.51 − 1.505 0.171
ROI. Then we excluded other fibers using 2 filter ROIs, which were placed λ23 (× 10− 4 mm2s− 1) 4.89 ± 1.10 4.96 ± 1.26 − 0.294 0.776
in the bilateral occipital lobes on the parasagittal planes. Only fibers CST, corticospinal tract; ROI, region of interest; FA, fractional anisotropy; MD, mean
passing through all 3 ROIs were defined as the sCC (Fig. 2B). diffusivity; λ1, primary eigenvalue; λ23, transverse eigenvalue.
 C. Yu et al. / NeuroImage 47 (2009) 451–458 455

in the stroke patients at different TPs and the indices in healthy controls. Table 5
To test whether the diffusion indices of the degenerated CSTs were Posthoc analyses (LSD) after repeated measures ANOVA in degenerated CSTs of the
stroke patients.
abnormal at the first TP, we used a paired samples t test to compare the
diffusion indices of the sCSTs and the ROIs of the CSTs between the P values 2 3 4 5
affected and unaffected sides in stroke patients at the first TP. To study 1 (rFA) b0.001 b0.001 b0.001 b0.001
the dynamic changes in the diffusion indices of the affected CST, we 2 (rFA) 0.005 b0.001 0.001
3 (rFA) 0.034 0.081
performed a repeated measures ANOVA on the rFA, rMD, rλ1 and rλ23 of
4 (rFA) 0.114
the ROIs of the CSTs in stroke patients at different TPs. Then we used a 1 (rMD) 0.189 0.009 0.001 b0.001
post hoc (least significant difference, LSD) test to examine the 2 (rMD) 0.024 0.001 b0.001
differences in the diffusion indices between every 2 TPs. On the data 3 (rMD) 0.005 b0.001
of stroke patients, we performed a Spearman correlation to test for the 4 (rMD) 0.663
1 (rλ1) 0.005 0.005 0.005 0.006
possible correlations between changes in the rFA, rMD, rλ1 and rλ23 of 2 (rλ1) 0.783 0.183 0.111
the ROIs of the CSTs at the first 2 TPs and the clinical measures after 3 (rλ1) 0.251 0.169
1 year. We used intraclass correlation coefficients (ICC) calculated by a 4 (rλ1) 0.864
one-way random effects model to test the intra-rater reproducibility 1 (rλ23) 0.001 b0.001 b0.001 b0.001
2 (rλ23) 0.005 b0.001 0.001
(McGraw and Wong 1996). The significance level was defined as
3 (rλ23) 0.001 0.003
P b 0.05, 2-tailed, for all statistical procedures. 4 (rλ23) 0.121

Results
Reproducibility of ROI placement
The ICCs of the 2 measurements were 0.98 for the FA, 0.96 for the
MD, 0.91 for the λ1 and 0.89 for the λ23. These findings indicated an
acceptable reproducibility in defining the ROIs.
Plasticity in the contralesional CST in stroke patients
We performed a tract ROI analysis to investigate whether the
contralesional CST underwent plastic changes. See Table 2 for the
normalized diffusion indices of the contralesional sCSTs (sCST/sCC) in
the stroke patients at the 5 TPs, as well as those in the healthy controls.
A one-way ANOVA did not show any significant differences in the
normalized FA (F = 0.48, P = 0.79), MD (F = 0.28, P = 0.92), λ1
(F = 0.03, P = 1.00) and λ23 (F = 0.62, P = 0.68) of the healthy sCSTs
across the 6 groups. This suggests that there was no detectable
plasticity in the contralesional CSTs in the stroke patients. Then we
adopted a more optimized method for normalization by using the
ratios (rFA, rMD, rλ1 and rλ23) between the affected CST and the
unaffected CST to study the dynamic changes in the diffusion indices
in the affected CST.
Differences in the diffusion indices between the affected and unaffected
CSTs at the first TP
To test whether the diffusion indices of the degenerated CSTs were
abnormal at the first TP, we analyzed the differences in the diffusion
indices between the affected and unaffected CSTs at the first TP in the
stroke patients using a paired samples t test. For the first TP we did not
find any significant differences between the affected and unaffected
sides in the diffusion indices of either the segment or the ROIs of the
CSTs in the stroke patients (Table 3).
Table 4
Diffusion indices of the ROI in the degenerated CST at the five time points for the stroke
patients.
LSD, least significant difference; CST, corticospinal tract; FA, fractional anisotropy; MD,
mean diffusivity; λ1, primary eigenvalue; λ23, transverse eigenvalue; rFA, rMD, rλ1 and
rλ23 denote ratios of the FA, MD, λ1 and λ23 between the affected and unaffected sides.
Arabic numbers (1, 2, 3, 4, and 5) represent the time points post stroke, i.e. within
1 week, at 2 weeks, at 1 month, at 3 month, and at 1 year.
Dynamic changes of diffusion indices in the degenerated CST
Table 4 showed the diffusion indices of the ROI of the degenerated
CST at the 5 TPs. The repeated measures ANOVA showed significant
differences in the rFA (F = 44.34, P b 0.001), rMD (F = 22.45,
P b 0.001), rλ1 (F = 8.13, P b 0.001) and rλ23 (F = 31.66, P b 0.001).
After posthoc (LSD) analyses (Table 5), the ROI of the degenerated
CST exhibited differences in the rFA between TP1 and TP2-5, TP2
and TP3-5, TP3 and TP4; in the rMD between TP1 and TP3-5, TP2 and
TP3-5, TP3 and TP4, 5; in the rλ1 between TP1 and TP2-5; and in the
rλ23 between TP1 and TP2-5, TP2 and TP3-5, TP3 and TP4, 5. The
dynamic changes in the rFA were sharply decreased within the first
month, slowly decreased from 1 month to 3 months, and then
remained relatively stable (Fig. 3A). The rMD did not change
significantly within the first 2 weeks, increased from 2 weeks to
3 months, and then reached a relatively stable state (Fig. 3B). The rλ1
decreased within 2 weeks and then reached a stable state or slightly
increased (Fig. 3C). The rλ23 increased over the first 3 months, and
then did not change significantly (Fig. 3D).
Correlations between early changes in the diffusion indices in the
degenerated CST and clinical outcomes after 1 year
It is clinically important to test whether early changes in the
diffusion indices of the degenerated CSTs can predict the clinical
outcomes after 1 year in stroke patients. Spearman correlation was
used to test whether changes in the diffusion indices within the first
2 weeks post stroke can be used to predict the clinical variables after
1 year. We found that the changes in the rFA of the CSTs between the
first 2 TPs were correlated with the MI (rs = − 0.924, P = 0.0004;
Fig. 4A) and NIHSS (rs = 0.962, P = 0.00003; Fig. 4B) after 1 year in
the stroke patients. However, we did not find any significant
correlation (P N 0.05) between the rFA, rMD, rλ1 or rλ23 of the CSTs
at either the first or second TPs and the clinical scales after 1 year.

Diffusion Within 1 week 2 weeks 1 month 3 months 1 year Discussion

indices (n = 9) (n = 9) (n = 9) (n = 9) (n = 9)
rFA 0.96 ± 0.11 0.73 ± 0.12 0.63 ± 0.14 0.55 ± 0.14 0.57 ± 0.13
In this study, we did not find any significant differences in the
rMD 0.97 ± 0.09 1.01 ± 0.08 1.06 ± 0.07 1.16 ± 0.12 1.17 ± 0.10
rλ1 0.96 ± 0.10 0.86 ± 0.05 0.85 ± 0.08 0.88 ± 0.08 0.88 ± 0.06 diffusion indices of the contralesional CSTs of the stroke patients
rλ23 1.00 ± 0.12 1.20 ± 0.17 1.33 ± 0.14 1.53 ± 0.21 1.65 ± 0.28 among any of the 5 TPs and those of the healthy controls, which
CST, corticospinal tract; FA, fractional anisotropy; MD, mean diffusivity; λ1, primary
suggests no detectable plasticity in the contralesional CSTs in these
eigenvalue; λ23, transverse eigenvalue; rFA, rMD, rλ1 and rλ23 denote ratios of the FA, patients. We showed the dynamic changes in the degenerated CST in
MD, λ1 and λ23 between the affected and unaffected sides. stroke patients: (1) the rFA of the degenerated CST monotonously
456 C. Yu et al. / NeuroImage 47 (2009) 451–458

Fig. 3. Dynamic changes in the ratios of the fractional anisotropy (rFA, A), mean diffusivity (rMD, B), primary eigenvalue (rλ1, C) and transverse eigenvalue (rλ23, D) between the
affected and unaffected sides of the corticospinal tracts (CST) in individual stroke patients.

decreased during the first 3 months, and then remained relatively and then did not change significantly; and (4) the rλ23 increased
unchanged; (2) the rMD maintained relatively stable during the first during the first 3 months and then maintained relative stability.
2 weeks, and gradually increased until 3 months, and then reached a Finally, we found that changes in the rFA of the CST within the first
relatively stable level; (3) the rλ1 decreased during the first 2 weeks 2 weeks were correlated with the NIHSS and MI after 1 year, which

Fig. 4. Plots of relationships between the changes in the ratios of the fractional anisotropy (rFA) of the corticospinal tracts within the first 2 weeks and the Motricity Index (MI) of the
affected side (rs = − 0.924, P = 0.0004, (A) and the National Institutes of Health Stroke Scale (NIHSS) (rs = 0.962, P = 0.00003, (B) after 1 year in stroke patients. The rs is the
Spearman correlation coefficient.
 C. Yu et al. / NeuroImage 47 (2009) 451–458
suggests that early changes of the rFA of the CSTs could be used to
predict clinical outcomes in stroke patients.
Plasticity of the contralesional CST
During the first several months after stroke, spontaneous recovery
of motor function is a common phenomenon (Duncan et al., 2000).
Such recovery has been attributed to cortical reorganization, including
the recruitment of the ipsilesional non-motor areas (Calautti and
Baron 2003; Schaechter et al., 2006; Seitz et al., 1998; Weiller et al.,
1993), contralesional motor areas (Chollet et al., 1991; Cramer et al.,
1997; Nelles et al., 1999), and bilateral non-primary motor areas
(Fridman et al., 2004; Johansen-Berg et al., 2002; Ward et al., 2003).
An earlier study reported that the FA of the normal-appearing white
matter of the whole brain increased within 2 years after stroke, which
suggests a long-term improvement in apparent white matter integrity
following ischemic stroke (Wang et al., 2006). Therefore, it is critically
important to study the possible plasticity in the contralesional CST
before using the ratios (rFA, rMD, rλ1 and rλ23) between the affected
and unaffected sides to study the dynamic changes of degenerated
CST. In the present study, we did not find any obvious plasticity in the
contralesional CST in the stroke patients. This supports the use of the
rFA, rMD, rλ1 and rλ23 as measures to assess the severity of WD in the
CST in stroke patients, which has been adopted in many previous
studies (Pierpaoli et al., 2001; Thomalla et al., 2004, 2005).
Dynamic changes in the diffusion indices of the degenerated CST
In the central nervous system WD is characterized by a series of
simultaneous events. From several hours to days after the time of the
lesion, the cytoskeleton of the axons disintegrates into axonal
fragments (George and Griffin, 1994a; Kerschensteiner et al., 2005).
Simultaneously, the myelin sheaths that surround the axons become
less tightly wrapped and eventually break apart forming ovoids
(George and Griffin, 1994b). Although an early transient period of
microglial activation in the degenerating fibers occurs, the appearance
of round macrophages is delayed until days 18–21. The activated
microglial cells digest the axonal and myelin debris. Both axonal
debris and myelin debris are almost completely cleared from the CNS
in 90 days (George and Griffin, 1994b), whereas axonal regeneration
and myelin formation are almost non-existent due to the apoptosis of
oligodendrocytes in the first few weeks following injury (Crowe et al.,
1997). This process can be well characterized by changes in the
diffusion tensor eigenvalues.
WD occurs several hours to days after the insult, but how early can
DTI detect the changes that result from WD in the human brain? To
answer this question, we studied the possible differences in diffusion
indices between the affected and unaffected CSTs within a week after
stroke. We did not find any significant differences using either tract
ROI or pure ROI method. This finding suggests that the pathologic
changes of the affected CST are not apparent enough to be detected by
the technique of DTI within 1 week after stroke. Thus we can regard
the rFA, rMD, rλ1 and rλ23 at the first TP as baseline values, and the
changes in the diffusion indices resulted from WD at different TPs can
be assessed by comparing with the baseline. Additionally, we can use
the changes in rFA, rMD, rλ1 and rλ23 between TP2 and TP1 as the
early changes in the diffusion indices of the degenerated CST to study
whether early changes in the diffusion indices of the degenerated CST
can predict long-term clinical outcomes after 1 year.
At the second week, we found these changes in the diffusion
indices: a sharply decreased FA, a reduced λ1, an increased λ23, and an
unchanged MD. The decrease in the λ1 can be explained by the
fragmentation of axons creating barriers to the longitudinal displace-
ment of water molecules (Kerschensteiner et al., 2005). The increase
in λ23 corresponds to the degradation of the myelin sheaths, leading to
water molecules becoming more mobile perpendicular to the axons.
457
These findings agreed with an animal study (Song et al., 2003), in
which the authors performed serial diffusion measurements of the
optic nerve to describe the process of WD in a mouse model of retinal
ischemia. They found a decrease in the λ1 in the first days of
degeneration corresponding to the disintegration of the axonal micro-
structure, during which time the myelin remained intact. 5 days after
the initial injury λ23 increased, corresponding to the degradation of
the myelin sheaths. The increase in FA and the unchanged MD are the
result of the decrease in λ1 and the increase in λ23.
From 2 weeks to 1 month after stroke, the λ1 did not decrease
further, a finding which can be explained by the complete fragmenta-
tion of the affected axons. However, the λ23 increased further,
corresponding to further degradation of the myelin sheaths and the
beginning of the clearance of the axonal and myelin debris. The
changes in the diffusion tensor eigenvalues may lead to the decrease
in FA and the increase in MD.
From 1 month to 3 months after stroke, most of the axonal frag-
ments are cleared, allowing the water molecules to once again diffuse
in the longitudinal direction, resulting in the slight increase of λ1. The
clearance of the axonal and myelin debris with no apparent axonal
regeneration and new myelin formation can lead to a further increase
in the λ23. The changes in the diffusion tensor eigenvalues may lead to
the decrease in FA and the increase in MD.
From the fourth month to 1 year after stroke, none of the diffusion
indices showed significant changes. These results suggest that the pro-
cess of WD had reached a relatively stable state by 3 months post stroke.
Early prediction of clinical outcomes
We did not find any significant correlation between the diffusion
indices at the first TP (4 ± 2 days, 1–7 days from onset) or the second
TP (15 ± 2 days, 13–20 days from onset) and the clinical variables at
1 year after stroke. These findings differ from a previous DTI study of 9
stroke patients (9 ± 4 days, 2–16 days from onset) (Thomalla et al.,
2004), which reported significant correlations between a decrease in
the rFA in the cerebral peduncle and motor scales at the outcome
examination 90 days after stroke. Although we cannot provide a
definite reason, the differences in recruitment criteria and examina-
tion times may account for the differences. In the present study, we
found strong correlations between early changes in the rFA of the CSTs
and the MI (rs = −0.924, P = 0.0004; Fig. 4A) and NIHSS (rs = 0.962,
P = 0.00003; Fig. 4B) at 1 year after stroke. We think that the early
changes in the rFA between the first 2 TPs could be superior to the
measurements at a single TP since the latter is more likely to be
influenced by individual side differences that existed before the stroke
onset. These results are partly consistent with previous studies that
found a worse outcome in patients if WD was evident on conventional
MRI (Fukui et al., 1994; Sawlani et al., 1997). In another study,
anisotropy contrast images derived from diffusion-weighted imaging
along 3 directions at 3 weeks after ischemic stroke or intracerebral
hemorrhage showed early WD in all patients with poor recovery, but
WD was not found in patients who got a good recovery (Watanabe
et al., 2001). In line with these data, the results of our study indicate
that the extent of early WD at sites remote from the initial infarct
appears to be a predictor of motor outcomes.
Methodological consideration
In this study, we used an affine transformation to align the images
from later time points to the first time point. This procedure could
cause a misalignment problem as a result of volume changes (edema
during the early stage and atrophy in the later course), such that some
voxels containing fibers from the CST at the first time point might not
contain CST fibers at later time points following volume changes due
to atrophy. This could affect the results of our ROI analysis of the
diffusion indices of the degenerated CSTs. In the pure ROI analysis of
458 C. Yu et al. / NeuroImage 47 (2009) 451–458
the present study, we tried to avoid the problem by drawing the ROI of
the CST on the FA image of the last time point. Our goal was to ensure
that the ROIs would be located within the CST at all time points
because the atrophy effect of the ipsilesional CST is more prominent at
later stages. Moreover, when we drew the ROIs, we also carefully
checked the location of the ROI at each time point in each patient to
ensure that the ROIs were within the CSTs.
In conclusion, we did not find apparent plasticity in the contrale-
sional CST in patients with motor pathway ischemic stroke, which
suggests that plasticity of the contralesional CST is not a critical
mechanism for spontaneous recovery of motor function in stroke
patients. We observed the dynamic evolution of WD in vivo using DTI,
which may help in understanding the process of WD in other
neurological diseases. We propose that early changes in the rFA of the
CST may be useful for predicting the long-term clinical outcomes in
stroke patients. This predictive ability may contribute to determining
optimal strategies for stroke treatment and rehabilitation at an early
post stroke stage.
Acknowledgments
This study was partly supported by the Natural Science Foundation
of China (Nos. 30670601, 30870694), and the Program for New Century
Excellent Talents in University (NCET-07-0568). We appreciate the
English language assistance of Drs. Rhoda E. and Edmund F. Perozzi.
References
Anderson, A.W., Zhong, J., Petroff, O.A., Szafer, A., Ransom, B.R., Prichard, J.W., Gore, J.C.,
1996. Effects of osmotically driven cell volume changes on diffusion-weighted
imaging of the rat optic nerve. Magn. Reson. Med. 35, 162–167.
Basser, P.J., Pajevic, S., 2003. A normal distribution for tensor-valued random variables:
applications to diffusion tensor MRI. IEEE. Trans. Med. Imaging. 22, 785–794.
Basser, P.J., Mattiello, J., Le Bihan, D., 1994. Estimation of the effective self-diffusion
tensor from the NMR spin-echo. J. Magn. Reson. B. 103, 247–254.
Beaulieu, C., Allen, P.S., 1994. Water diffusion in the giant axon of the squid: implications
for diffusion-weighted MRI of the nervous system. Magn. Reson. Med. 32, 579–583.
Berman, J.I., Mukherjee, P., Partridge, S.C., Miller, S.P., Ferriero, D.M., Barkovich, A.J.,
Vigneron, D.B., Henry, R.G., 2005. Quantitative diffusion tensor MRI fiber
tractography of sensorimotor white matter development in premature infants.
NeuroImage 27, 862–871.
Brott, T., Adams, H.P., Olinger, C.P., Marler, J.R., Barsan, W.G., Biller, J., Spilker, J., Holleran,
R., Eberle, R., Hertzberg, V., Rorick, M., Moomaw, C.J., Walker, M., 1989.
Measurements of acute cerebral infarction: a clinical examination scale. Stroke
20, 864–870.
Calautti, C., Baron, J.C., 2003. Functional neuroimaging studies of motor recovery after
stroke in adults: a review. Stroke 34, 1553–1566.
Catani, M., Howard, R.J., Pajevic, S., Jones, D.K., 2002. Virtual in vivo interactive
dissection of white matter fasciculi in the human brain. NeuroImage 17, 77–94.
Chollet, F., DiPiero, V., Wise, R.J., Brooks, D.J., Dolan, R.J., Frackowiak, R.S., 1991. The
functional anatomy of motor recovery after stroke in humans: a study with positron
emission tomography. Ann. Neurol. 29, 63–71.
Cramer, S.C., Nelles, G., Benson, R.R., Kaplan, J.D., Parker, R.A., Kwong, K.K., Kennedy,
D.N., Finklestein, S.P., Rosen, B.R., 1997. A functional MRI study of subjects recovered
from hemiparetic stroke. Stroke 28, 2518–2527.
Crowe, M.J., Bresnahan, J.C., Shuman, S.L., Masters, J.N., Beattie, M.S., 1997. Apoptosis and
delayed degeneration after spinal cord injury in rats and monkeys. Nat. Med. 3, 73–76.
Demeurisse, G., Demol, O., Robaye, E., 1980. Motor evaluation in vascular hemiplegia.
Eur. Neurol. 19, 382–389.
Duncan, P.W., Lai, S.M., Keighley, J., 2000. Defining post-stroke recovery: implications
for design and interpretation of drug trials. Neuropharmacology 39, 835–841.
Fridman, E.A., Hanakawa, T., Chung, M., Hummel, F., Leiguarda, R.C., Cohen, L.G., 2004.
Reorganization of the human ipsilesional premotor cortex after stroke. Brain 127,
747–758.
Fukui, K., Iguchi, I., Kito, A., Watanabe, Y., Sugita, K., 1994. Extent of pontine pyramidal
tract Wallerian degeneration and outcome after supratentorial hemorrhagic stroke.
Stroke 25, 1207–1210.
George, R., Griffin, J.W., 1994a. The proximo-distal spread of axonal degeneration in the
dorsal columns of the rat. J. Neurocytol. 23, 657–667.
George, R., Griffin, J.W., 1994b. Delayed macrophage responses and myelin clearance
during Wallerian degeneration in the central nervous system: the dorsal radicu-
lotomy model. Exp. Neurol. 129, 225–236.
Johansen-Berg, H., Dawes, H., Guy, C., Smith, S.M., Wade, D.T., Matthews, P.M., 2002.
Correlation between motor improvements and altered fMRI activity after
rehabilitative therapy. Brain 125, 2731–2742.
Johnson, A.C., McNabb, A.R., Rossiter, R.J., 1950. Chemistry of Wallerian degeneration.
A review of recent studies. Arch. Neurol. Psychiatry 64, 105–121.
Kerschensteiner, M., Schwab, M.E., Lichtman, J.W., Misgeld, T., 2005. In vivo imaging of
axonal degeneration and regeneration in the injured spinal cord. Nat. Med. 11,
572–577.
Lampert, P.W., Cressman, M.R., 1966. Fine-structural changes of myelin sheaths after
axonal degeneration in the spinal cord of rats. Am. J. Pathol. 49, 1139–1155.
Le Bihan, D., 2003. Looking into the functional architecture of the brain with diffusion
MRI. Nat. Rev. Neurosci. 4, 469–480.
Lin, F., Yu, C., Jiang, T., Li, K., Li, X., Qin, W., Sun, H., Chan, P., 2006. Quantitative analysis
along the pyramidal tract by length-normalized parameterization based on
diffusion tensor tractography: application to patients with relapsing neuromyelitis
optica. NeuroImage 33, 154–160.
Maeda, T., Ishizaki, K., Yura, S., 2005. Can diffusion tensor imaging predict the functional
outcome of supra-tentorial stroke? No To Shinkei 57, 27–32.
Mori, S., Van Zijl, P.C., 2002. Fiber tracking: principles and strategies—a technical review.
NMR Biomed. 15, 468–480.
Mori, S., Crain, B.J., Chacko, V.P., van Zijl, P.C.M., 1999. Three dimensional tracking of
axonal projections in the brain by magnetic resonance imaging. Ann. Neurol. 45,
265–269.
McGraw, K.O., Wong, S.P., 1996. Forming inferences about some intraclass correlations.
Psychol. Methods 1, 30–46.
Nelles, G., Spiekramann, G., Jueptner, M., Leonhardt, G., Müller, S., Gerhard, H., Diener,
H.C., 1999. Evolution of functional reorganization in hemiplegic stroke: a serial
positron emission tomographic activation study. Ann. Neurol. 46, 901–909.
Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory.
Neuropsychologia 9, 97–113.
Pagani, E., Fillipi, M., Rocca, M.A., Horsfield, M.A., 2005. A method for obtaining tract-
specific diffusion tensor MRI measurements in the presence of disease: applica-
tion to patients with clinically isolated syndromes suggestive of multiple sclerosis.
NeuroImage 26, 258–265.
Partridge, S.C., Mukherjee, P., Berman, J.I., Henry, R.G., Miller, S.P., Lu, Y., Glenn, O.A.,
Ferriero, D.M., Barkovich, A.J., Vigneron, D.B., 2005. Tractography-based quantifica-
tion of diffusion tensor imaging parameters in white matter tracts of preterm
newborns. J. Magn. Reson. Imaging 22, 467–474.
Pierpaoli, C., Basser, P.J., 1996. Toward a quantitative assessment of diffusion anisotropy.
Magn. Reson. Med. 36, 893–906.
Pierpaoli, C., Barnett, A., Pajevic, S., Chen, R., Penix, L., Virta, A., Basser, P.J., 2001. Water
diffusion changes in Wallerian degeneration and their dependence on white matter
architecture. NeuroImage 13, 1174–1185.
Sawlani, V., Gupta, R.K., Singh, M.K., Kohli, A., 1997. MRI demonstration of Wallerian
degeneration in various intracranial lesions and its clinical implications. J. Neurol.
Sci. 146, 103–108.
Schaechter, J.D., Moore, C.I., Connell, B.D., Rosen, B.R., Dijkhuizen, R.M., 2006. Structural
and functional plasticity in the somatosensory cortex of chronic stroke patients.
Brain 129, 2722–2733.
Seitz, R.J., Höflich, P., Binkofski, F., Tellmann, L., Herzog, H., Freund, H.J., 1998. Role of the
premotor cortex in recovery from middle cerebral artery infarction. Arch. Neurol.
55, 1081–1088.
Sevick, R.J., Kanda, F., Mintorovitch, J., Arieff, A.I., Kucharczyk, J., Tsuruda, J.S., Norman,
D., Moseley, M.E., 1992. Cytotoxic brain edema: assessment with diffusion-
weighted MR imaging. Radiology 185, 687–690.
Song, S.K., Sun, S.W., Ju, W.K., Lin, S.J., Cross, A.H., Neufeld, A.H., 2003. Diffusion tensor
imaging detects and differentiates axon and myelin degeneration in mouse optic
nerve after retinal ischemia. NeuroImage 20, 1714–1722.
Thomalla, G., Glauche, V., Koch, M.A., Beaulieu, C., Weiller, C., Röther, J., 2004. Diffusion
tensor imaging detects early Wallerian degeneration of the pyramidal tract after
ischemic stroke. NeuroImage 22, 1767–1774.
Thomalla, G., Glauche, V., Weiller, C., Röther, J., 2005. Time course of Wallerian
degeneration after ischaemic stroke revealed by diffusion tensor imaging. J. Neurol.
Neurosurg. Psychiatry 76, 266–268.
Wakana, S., Jiang, H., Nagae-Poetscher, L.M., van Zijl, P.C., Mori, S., 2004. Fiber
tract-based atlas of human white matter anatomy. Radiology 230, 77–87.
Waller, A., 1850. Experiments on the section of the glossopharyngeal and hypoglossal
nerves of the frog and observations of the alternatives produced thereby in the
structure of their primitive fibres. Philos. Trans. R Soc. Lond. 140, 423–429.
Ward, N.S., Brown, M.M., Thompson, A.J., Frackowiak, R.S., 2003. Neural correlates of
motor recovery after stroke: a longitudinal fMRI study. Brain 126, 2476–2496.
Wang, C., Stebbins, G.T., Nyenhuis, D.L., deToledo-Morrell, L., Freels, S., Gencheva, E.,
Sripathirathan, K., Moseley, M.E., Turner, D.A., Gabrieli, J.D., Gorelick, P.B., 2006.
Longitudinal changes in white matter following ischemic stroke: a three-year
follow-up study. Neurobiol. Aging 27, 1827–1833.
Watanabe, T., Honda, Y., Fujii, Y., Koyama, M., Matsuzawa, H., Tanaka, R., 2001. Three-
dimensional anisotropy contrast magnetic resonance axonography to predict the
prognosis for motor function in patients suffering from stroke. J. Neurosurg. 94,
955–960.
Weiller, C., Ramsay, S.C., Wise, R.J., Friston, K.J., Frackowiak, R.S., 1993. Individual
patterns of functional reorganization in the human cerebral cortex after capsular
infarction. Ann. Neurol. 33, 181–189.
Wimberger, D.M., Roberts, T.P., Barkovich, A.J., Prayer, L.M., Moseley, M.E., Kucharczyk, J.,
1995. Identification of “premyelination” by diffusion-weighted MRI. J. Comput.
Assist. Tomogr. 19, 28–33.
Yu, C., Shu, N., Li, J., Qin, W., Jiang, T., Li, K., 2007. Plasticity of the corticospinal tract in
early blindness revealed by quantitative analysis of fractional anisotropy based on
diffusion tensor tractography. NeuroImage 36, 411–417.
Yu, C., Lin, F., Li, K., Jiang, T., Qin, W., Sun, H., Chan, P., 2008. Pathogenesis of normal-
appearing white matter damage in neuromyelitis optica: diffusion-tensor MR imaging.
Radiology 246, 222–228.