Microstructural status of ipsilesional and contralesional corticospinal tract correlates with motor skill in chronic stroke patients

Abstract

INTRODUCTION

Stroke commonly causes acute, contralateral loss of motor function. Poststroke recovery of this lost function typically occurs over the subsequent weeks to months, yet is variable among stroke survivors. Motor recovery has been shown to relate to reorganization of activity in the ipsilesional and contralesional sensorimotor cortices [Dijkhuizen et al., 2001; Jaillard et al., 2005; Ward et al., 2003]. Accumulating evidence from studies in animal models of stroke suggest that structural remodeling of white matter of the ipsilesional and contralesional hemispheres also plays a role in motor recovery [Brus‐Ramer et al., 2007; Carmichael and Chesselet, 2002; Dancause et al., 2005; Liu et al., 2008; Stroemer et al., 1995]. Little is known about the relationship between white matter structure of the sensorimotor network and motor recovery in patients with stroke.

The relatively recent development of diffusion tensor imaging (DTI), a noninvasive magnetic resonance imaging (MRI) technology that measures the random motion of water molecules in brain tissue, enables examination of white matter microstructure in vivo. Fractional anisotropy (FA) is a metric derived from DTI data that quantifies the extent to which water diffusion is directionally restricted, and is influenced by a number of factors including axonal myelination, diameter, density, and orientational coherence [Beaulieu, 2002; Takahashi et al., 2002]. FA is currently the most commonly used metric to reflect microstructural status of white matter. During the first few weeks after stroke in patients who go on to recover relatively little motor function, the ipsilesional corticospinal tract (CST) undergoes progressive decline in FA due to loss of axonal integrity that leads to Wallerian degeneration [Moller et al., 2007; Watanabe et al., 2001]. In animals recovering from experimentally‐induced stroke, FA of perilesional white matter declines during the first few days poststroke [van der Zijden et al., 2008], yet rises during subsequent weeks [Ding et al., 2008; Jiang et al., 2006; van der Zijden et al., 2008]. The rise in FA of perilesional white matter has been shown to colocalize with histological evidence of white matter reorganization [Ding et al., 2008; Jiang et al., 2006], increased angiogenesis and local cerebral blood flow [Ding et al., 2008], and enhanced neural connectivity with perilesional sensorimotor cortex based on manganese‐enhanced MRI [van der Zijden et al., 2008]. Treatment of animals poststroke with certain cell‐based or pharmacological agents enhances the rise in perilesional FA in association with improved motor recovery [Ding et al., 2008; Jiang et al., 2006]. Anatomical tracing studies in animals with experimental stroke have also shown that contralesional corticofugal axons sprout to make new connections with nuclei denervated due to stroke, and that this form of axonal remodeling is enhanced by certain poststroke treatments in association with improved motor recovery [Andrews et al., 2008; Carmichael and Chesselet, 2002; Chen et al., 2002; Lee et al., 2004; Liu et al., 2008; Ramic et al., 2006]. In patients with chronic stroke, a progressive increase in FA of normal‐appearing, whole‐brain white matter has been observed (from 3–6 to 24 months poststroke) [Wang et al., 2006]. Together, these findings suggest that structural changes of ipsilesional and contralesional white matter occur after stroke and relate to motor recovery. In this study, we sought to examine the relationship between motor outcome in patients with chronic stroke and microstructural status of brain white matter tracts as measured by FA.

To meet this aim, we employed a recently developed method called Tract‐Based Spatial Statistics (TBSS) that optimally aligns major white matter tracts across subjects, thereby allowing for valid voxelwise analyses of FA of all major white matter tracts [Smith et al., 2006]. In addition, we focused on evaluating FA of the ipsilesional and contralesional CST using tractography. As the majority of research aimed at understanding the neural underpinnings of motor outcome in patients with stroke has focused on reorganization of gray matter activity, this study widens the lens of inquiry by investigating the influence of microstructural status of white matter.

MATERIALS AND METHODS

Image Analysis

DTI

Diffusion tensor images were corrected for motion and eddy current distortion using FMRIB's (http://www.fmrib.ox.ac.uk/fsl) Linear Image Registration Tool by registering each image to the first acquired T2‐weighted image of the volume using a 12 degree‐of‐freedom affine transformation and mutual information cost function [Jenkinson et al., 2002]. The data were then skull‐stripped using FMRIB's Brain Extraction Tool. The diffusion tensor and associated eigensystem were estimated at each voxel using a linear least squares regression method from which our primary DTI metric of interest, FA, was calculated [Basser, 1995]. The FA metric reflects the fraction of the total magnitude of the diffusion tensor ascribed to anisotropic diffusion, and scales from a value of 0 indicating isotropy to 1 indicating anisotropy. As FA depends on the relative magnitude of the eigenvalues, maps of axial diffusivity and radial diffusivity were also generated to examine the underlying principal diffusivities. Axial diffusivity is the first eigenvalue of the diffusion tensor and reflects the magnitude of diffusivity parallel to the direction of maximal diffusion. In white matter, axial diffusivity is in the direction parallel to axons, and is thought to relate largely to axonal properties [Song et al., 2003]. Radial diffusivity was calculated as the average of the second and third eigenvalues, and reflects the magnitude of diffusivity orthogonal to the direction of maximal diffusion [Song et al., 2002]. In white matter, radial diffusion is in the direction orthogonal to axons and is thought to be affected largely by the axolemma and myelin sheath [Song et al., 2002, 2003].

Voxelwise analysis of white matter tract FA

To perform voxelwise analysis of the DTI data, images from the four patients with a right‐sided stroke (Patient no. 1, 5, 7, 10) were flipped about the midsagittal plane, thereby lateralizing the stroke to the left hemisphere in all patients. Because some voxelwise analyses compared FA in patients to that in controls, the DTI data from the control subjects matched to these patients were also flipped midsagittally.

Statistical analyses of FA were performed using FMRIB's TBSS (v1.0) [Smith et al., 2006]. The FA volume from each patient and control was aligned to the within‐group target volume that was closest to the group mean using a nonlinear registration algorithm (http://www.doc.ic.ac.uk/~dr/software) [Rueckert et al., 1999], followed by a 12 degree‐of‐freedom affine registration from the target FA volume to the Montreal Neurological Institute brain template (MNI152). Visual inspection of each subject's registered FA volume compared to the MNI152 template revealed no noticeable misalignment. A mean FA volume was created and used to generate an FA skeleton that corresponds to the center of major white matter tracts common to the group. Each subject's FA data were mapped to the skeleton by searching perpendicular to the skeleton, finding the highest FA value, and assigning that value to the skeleton.

We distinguished the portion of the FA skeleton in which FA values were mapped from a site of chronic infarction from the remainder of the skeleton. To do this, the T2‐weighted images of the DTI volume were aligned across all patients using the same transformation matrices determined to align their FA volumes. The lesion in each patient was then marked on the spatially normalized T2‐weighted images using MRIcro software [Rorden and Brett, 2000], and the total lesion volume was recorded. Subsequently, the same mapping used to map each patient's FA data to the skeleton was applied to the lesion mask, thereby identifying voxels of the skeleton that contained FA values from lesioned white matter. These voxels, which were 8% of all skeleton voxels, were considered to be within the lesion zone. The remainder of skeleton voxels was considered to be remote from the lesion (see Supporting Information Fig. ​Fig.11).

Open in a separate window

Figure 1

White matter tract regions exhibiting significant (clusterwise P < 0.001) positive correlation between FA and motor skill in patients with chronic stroke. (A) Regions in lesion zone. (B) Regions remote from lesion. Shown are clusters in: 1, ipsilesional precentral gyrus; 2, ipsilesional superior corona radiata; 3, ipsilesional superior/anterior corona radiata; 4, contralesional superior posterior limb of internal capsule; 5, ipsilesional middle posterior limb of internal capsule, cerebral peduncle; 6, ipsilesional superior/anterior corona radiata. Clusters shown overlaying mean FA image from patients with stroke.

The correlation between motor skill of the stroke‐affected hand (first principal component score) and FA in patients was tested voxelwise in the lesion zone and, separately, remote from the lesion. The regression model included age as a nuisance regressor because the patients were of varying age (Table ​(TableI)I) and increasing age has been shown to correlate with widespread reductions in brain white matter FA [Pfefferbaum et al., 2000].

To further probe the relationship between motor skill and microstructural status of white matter tracts, we tested for differences in FA between normal controls and patient subgroups with better and poorer motor skill of the affected hand (based on a median‐split of first principal component scores) using unpaired t‐tests. As there was no significant difference in age among the three groups (controls, better‐recovered patients, poorer‐recovered patients; P = 0.42, analysis of variance (ANOVA)), between‐group comparisons were not adjusted for subject age.

Each statistical map was corrected to a cluster‐wise significance level of P < 0.001. A significant cluster was defined as a set of contiguous voxels, each having a P‐value of less than 0.05, with a size that occurred by chance less than once in 1,000 times, determined by Monte Carlo simulations (10,000 iterations) on synthesized noise entered into the group analysis. Noise values were uniformly distributed from 0 to 1 to reflect the range of FA values. We report the size of significant clusters as well as the P‐value and MNI coordinates of the voxel within the cluster with the maximum significance. The anatomical location of significant clusters was determined by expert knowledge (author N.M.). To ease visualization of significant clusters of the FA skeleton, they are shown after thickening circumferentially by 2 mm.

For each cluster exhibiting significant between‐group differences in FA, underlying axial and radial diffusivities were examined. To do this, mean axial and radial diffusivities were calculated for each subject of the patient subgroup and normal controls. Between‐group differences in these diffusivities were tested using two‐tailed, unpaired t‐tests.

Tractography‐based analysis of CST FA

A tractography method called Path‐of‐Interest statistics (POIstats v1.4), a tool within the FreeSurfer software library (http://surfer.nmr.mgh.harvard.edu), was used for CST reconstruction using each subject's native space DTI volume. The right and left CST were reconstructed between the precentral gyrus and ipsilateral cerebral peduncle. The hand region of the precentral gyrus was labeled on the single axial T2‐weighted image in which the hand knob region [Yousry et al., 1997] was most noticeable (14–20 mm inferior from the vertex of the brain). Borders of this label were the precentral sulcus (anterior), central sulcus (posterior), line extending posterior from the superior frontal sulcus (medial), and brain surface (lateral). The cerebral peduncle was labeled on the single axial T2‐weighted image in which the region was most prominent.

Path reconstruction used a replica‐exchange Monte Carlo method [Hansmann, 1997; Hukushima and Nemoto, 1996; Swendsen and Wang, 1986] in which multiple path replicas, each with a different value of the “temperature bath” parameter, connected the endpoint labels by a spline curve through the diffusion tensor volume. To remove the influence of voxels containing cerebrospinal fluid on path reconstruction, we masked out voxels with trace value >6 mm²/s that characterizes the high level of total diffusivity of cerebrospinal fluid in ventricles and/or the site of chronic infarct. In search of the path replica with the lowest energy configuration, each path replica was perturbed at control points of the spline curve in a random direction and random, though bounded, distance. The energy of each path replica was determined using the relationship between the direction of each path line segment and the orientation distribution function [Kimmich and Weber, 1993; Tuch et al., 2003] calculated from the underlying diffusion tensor data. Replica exchange was implemented to guard against path replicas becoming trapped in local minima, with the decision to exchange replicas governed by the Metropolis update probability function [Metropolis et al., 1953]. Path replica perturbation and exchange were halted when the mean energy of successive replica sets was stable for 10 iterations. The optimal path among the final set of path replicas was assigned to the replica with the least energy overall. To account for the anatomic fact that white matter tracts have cross‐sectional extent, a probability density volume was constructed around the optimal path such that the value at each voxel was inversely proportionate to the square of the distance from the optimal path. A threshold of 5 × 10⁻⁴ was applied to the probability density volume, and the resulting volume was considered to be the reconstructed CST (see Supporting Information Fig. ​Fig.2).2). The location of the reconstructed CST in the ipsilesional hemisphere of patients was found to be symmetric to that in the contralesional hemisphere, indicating that the presence of the lesion did not interfere with performance of the tractography method. Applying a threshold of 1 × 10⁻³ to the probability density volume to yield the reconstructed CST did not change results of subsequent analyses.

Open in a separate window

Figure 2

White matter tract regions exhibiting significant (clusterwise P < 0.001) differences in FA in patient subgroups compared to normal controls. (A) Regions of reduced FA in patients with poorer motor skill. (B) Regions of elevated FA in patients with better motor skill. Shown are clusters in: 1, lesion zone involving precentral gyrus, superior/anterior corona radiata, genu of internal capsule and anterior limb of internal capsule; 2, remote from lesion in contralesional posterior limb of internal capsule, cerebral peduncle; 3, remote from lesion in ipsilesional posterior limb of internal capsule, cerebral peduncle, genu of internal capsule and anterior limb of internal capsule; 4, remote from lesion in ipsilesional superior/anterior corona radiata; 5, remote from lesion in contralesional PLIC; 6, remote from lesion in ipsilesional posterior limb of internal capsule; 7, remote from lesion in ipsilesional superior corona radiata. Clusters shown overlaying mean FA image from all subjects.

Two analyses were conducted utilizing FA of the reconstructed CST, each paralleling voxelwise analyses of FA. One analysis tested the correlation between motor skill of the affected hand (first principal component score) and mean FA of the reconstructed CST (separate analyses for ipsilesional and contralesional CST) with adjustment for patient age. The second analysis computed mean FA of each axial slice along the length of a reconstructed CST in patients and controls. The mean FA data were then interpolated to cover 50 z‐positions, bringing the CST data reconstructed in native space into spatial correspondence across subjects. The interpolated FA data from patient subgroups were compared to those from normal controls using a functional data analysis (FDA) approach. FDA is a set of statistical procedures that considers data points as samples from a continuous curve in a function space [Ramsay and Dalzell, 1991; Ramsay and Silverman, 1997]. FDA is in contrast to conventional multivariate statistical methods that consider data points on a curve as multivariate data in a finite‐dimensional space, which presents challenging issues related to multiple comparisons of spatially correlated data. Considering the interpolated CST FA data as a curve accounts for the underlying continuity and spatial correlation of the data.

Specifically, using the FDA toolbox in the R statistical analysis package (http://www.r-project.org), the interpolated CST FA data from each subject and hemisphere were fit using B‐splines with each curve smoothed to achieve 15 degrees of freedom. This level of smoothing was chosen empirically to maintain local features of the interpolated FA data (see Supporting Information Fig. ​Fig.3).3). Functional unpaired t‐tests were performed to compare curves of CST FA from patient subgroups with poorer and better motor skill to those from controls. Curves were considered to be significantly different if they differed over any interval at the two‐tailed 0.05 level. A difference between curves at a level greater than 0.05 and less than 0.10 was considered a nonsignificant trend. We report intervals of the curves that showed a significant, or trending toward a significant, difference between groups.

Open in a separate window

Figure 3

FA curves of the ipsilesional CST (A) and contralesional CST (B) in patients with chronic stroke with better (light gray line) and poorer (dark gray line) motor skill of the affected hand and normal control subjects (black line). The tracts were reconstructed between the precentral gyrus (PRG) and cerebral peduncle (CP), extending through the posterior limb of internal capsule (PLIC). FA curves are smoothed versions of interpolated FA data. Asterisks indicate the interval of the curve at which the patient subgroup differed from controls at P < 0.05 based on FDA. Dots indicate the interval of the curve at which the patient subgroup differed from controls at 0.05 < P < 0.10. Confidence intervals are not shown for ease of visualization.

White matter hyperintensity volume

The severity of white matter disease affects functional outcome after stroke [Arsava et al., in press; Kissela et al., 2009] and relates to microstructural status of white matter as measured by FA [Jones et al., 1999]. Accordingly, we questioned whether the differences in white matter tract FA we observed between patient subgroups and controls (see Results) could relate to between‐group differences in white matter disease. Therefore, we quantified the volume of white matter hyperintensity in the subjects based on the high‐resolution T2‐weighted images according to the semiautomated method described previously [Gurol et al., 2006] using MRIcro software [Rorden and Brett, 2000]. To avoid potential confound due to stroke, the volume of signal hyperintensity was measured in the contralesional hemisphere of the patients, and in the comparable hemisphere of the matched control subjects. White matter hyperintensity volumes were normalized by intracranial size measured using the high‐resolution T1‐weighted images. Normalized volumes of white matter hyperintensity among better‐recovered patients, poorer‐recovered patients, and controls were compared using one‐way ANOVA.

RESULTS

DISCUSSION

The main finding of this study is that motor skill of the stroke‐affected hand significantly and positively correlated with microstructural status, indexed by FA, of the ipsilesional CST and the contralesional CST in patients with chronic stroke. A correlation between FA of bilateral CST and behavior was detected using a voxelwise approach that was unbiased with respect to location of major white matter tracts and confirmed using a tractography approach applied to the CST. The finding of a relationship between microstructural status of the ipsilesional CST and motor skill is consistent with the established role of the CST in fine motor control [Hepp‐Reymond and Wiesendanger, 1972; Wiesendanger, 1984] and prior studies showing that greater sparing of the ipsilesional CST is associated with better motor outcome of the paretic limb of patients with stroke [Pendlebury et al., 1999; Pineiro et al., 2000; Schaechter et al., 2008]. The finding that the FA‐behavior correlation occurred at regions along the contralesional CST supports prior suggestions that the contralesional CST may play a role in motor recovery after unilateral stroke [Ago et al., 2003; Brus‐Ramer et al., 2007; Caramia et al., 2000; Fisher, 1992; Jankowska and Edgley, 2006; Johansen‐Berg et al., 2002; Liu et al., 2008]. Given the close proximity of the CST with other corticofugal fibers (e.g., corticopontine, corticorubral) in the cerebral hemispheres, FA of these other descending tracts may have contributed to the correlation between CST FA and motor skill. We interpret the observed FA‐behavior correlation as suggesting that biophysical properties that affect white matter FA, such as axonal density, diameter, myelination, and orientation coherence [Beaulieu, 2002; Takahashi et al., 2002], relate to the efficacy of communication along the CST, which in turn relate to the level of motor skill in chronic stroke patients. Prior studies in healthy adults have shown a correlation between performance on perceptual‐motor tasks and FA along functionally‐relevant white matter tracts [Bohr et al., 2007; Bucur et al., 2008; Tuch et al., 2005]. To our knowledge, this is the first report demonstrating that FA of bilateral CST correlates with motor performance of the affected hand in stroke patients.

The CST FA‐behavior correlation was associated with reduced FA of the ipsilesional CST in patients with poorer motor skill compared to normal controls. The finding of reduced FA of the ipsilesional CST in this patient subgroup is consistent with prior reports, and likely represents primary (in the lesion zone) and secondary (remote from the lesion) axonal degeneration and gliosis [Pierpaoli et al., 2001; Werring et al., 2000]. Reduced FA of the ipsilesional CST was found to be associated with elevated radial diffusivity, suggesting that loss in integrity of the axolemma and/or myelin sheath contributed to the reduced anisotropy. Changes in axial diffusivity have been shown to occur at regions of primary degeneration (i.e., elevation) and secondary degeneration (i.e., reduction) of the ipsilesional CST [Pierpaoli et al., 2001]. That we observed little or no significant elevation in axial diffusivity in the lesion zone may have been due to our operational definition of this FA skeleton region (see Materials and Methods section); reductions in axial diffusivity due to mapping from voxels of chronic infarction may have been diluted by mapping from voxels of undamaged white matter. The lack of significant reduction in axial diffusivity of the ipsilesional CST remote from the lesion may indicate relatively little axonal loss in these patients.

The CST FA‐behavior correlation was also associated, rather unexpectedly, with reduced FA of the contralesional CST in patients with poorer motor skill compared to normal controls. Contralesional CST regions of reduced FA had elevated radial diffusivity, suggesting a similar loss in integrity of the axolemma and/or myelin sheath as along the ipsilesional CST. The cross‐sectional design of our study precludes definitive conclusions about the antecedent of these between‐group differences. However, several findings, in sum, suggest that loss in microstructural status of the contralesional CST more likely evolved poststroke than was present prestroke. The control subjects in our study were well matched to the patients for age, handedness, and gender, excluding these factors as responsible for possible prestroke differences in contralesional CST FA and radial diffusivity. We also found no between‐group differences in the severity of white matter disease, which could theoretically affect susceptibility of the CST, both ipsilesional and contralesional, to poststroke degenerative changes. Further, a previous study showed progressive loss in FA of contralesional white matter over the first 6 months poststroke that paralleled the loss of ipsilesional CST FA [Buffon et al., 2005], supporting the possibility that our findings reflect poststroke change in microstructural status of the contralesional CST in the patients with poorer motor skill.

The CST FA‐behavior correlation was associated with elevated FA of bilateral CST in patients with better motor skill compared to controls. Accumulating evidence, largely from studies in animal models of stroke, suggests that white matter remodeling occurs in the ipsilesional and contralesional hemispheres during the recovery period after stroke. In the ipsilesional hemisphere of animals with experimental stroke, increased FA of white matter in the lesion borderzone has been shown to correspond spatially to histological evidence of white matter reorganization [Ding et al., 2008; Jiang et al., 2006] and MRI evidence of neural connectivity with perilesional sensorimotor cortex [van der Zijden et al., 2008]. In the contralesional hemisphere, CST axons have been demonstrated to sprout and form new connections with denervated motor nuclei in the brainstem and spinal cord spontaneously after experimental stroke, and this structural remodeling is enhanced by certain post‐stroke treatments (i.e., cell‐based, electrical stimulation, pharmacological paired with rehabilitation) that improve motor recovery [Andrews et al., 2008; Brus‐Ramer et al., 2007; Chen et al., 2002; Lee et al., 2004; Liu et al., 2008; Ramic et al., 2006]. Immunohistochemical studies suggest increases in oligodendrocyte myelinating activity in ipsilesional and contralesional white matter after experimental stroke [Gregersen et al., 2001; Ishiguro et al., 1993; Tanaka et al., 2003]. Our finding that the regions of bilateral CST exhibiting elevated FA had reduced radial diffusivity is consistent with the possibility that increased myelination underlies CST axonal remodeling in the stroke patients with better motor skill. DTI studies in patient populations other than stroke have similarly provided evidence of structural remodeling of functionally‐relevant white matter tracts. An increase in the number of tractography‐reconstructed fibers of the contralesional corticobulbar tract and an increase in FA at regions of the contralesional CST have been reported in patients with congenital hemiparesis due to unilateral periventricular white matter injury [Thomas et al., 2005]. Patients with multiple sclerosis early after clinical onset who had white matter damage in the working memory network yet little or no working memory impairment, showed an increase in the number of tractography‐reconstructed fibers between the right and left thalami [Audoin et al., 2007]. These reports, together with our findings, suggest that structural remodeling of functionally‐relevant white matter tracts may be an adaptive response that compensates for injury to the human brain.

Voxelwise regression analysis revealed that FA of association tracts of the ipsilesional hemisphere, primarily the SLF, correlated with motor skill in the patients with chronic stroke. The SLF, a multicomponent fiber bundle [Makris et al., 2005], includes parieto‐frontal fibers that are critical to exchanging information about the perception of the body in space for the planning, initiation, and updating of goal‐directed movement [Fuster, 2004]. The functional role of this SLF component is consistent with the observed FA‐behavior correlation. However, the functional role of the other association tracts exhibiting the FA‐behavior correlation (e.g., inferior longitudinal fasciculus, extreme capsule) is related to high‐level cognition and emotion and not linked tightly to motor function. Accordingly, the correlation between motor skill and FA of these association tracts may have been epiphenomenal, resulting from stroke‐induced reduction in FA of similar magnitude as that occurring along the neighboring ipsilesional CST. The patients in this study did not undergo testing of high‐level cognition and emotion. Such testing might have allowed us to determine the relative strength of correlation among various functions (i.e., motor, cognitive, emotional) and white matter tract FA.

Voxelwise regression analysis also revealed that FA of the contralesional orbitofrontal region correlated with motor skill in the patients. Interestingly, the orbitofrontal gyrus is a region of the adult human brain that has one of the highest levels of GAP‐43, a presynaptic protein strongly implicated in axonal remodeling after injury [Ng et al., 1988]. We speculate that the observed correlation between FA of the orbitofrontal region and motor skill among the patients may relate to the propensity for axonal remodeling in response to brain injury, and not be specifically tied to motor function.

A technical aspect of this study deserves comment. The voxelwise TBSS method involves automated alignment of each subject's FA volume to the template FA volume. Stroke‐induced tissue necrosis and cavitation can result in structural and signal intensity abnormalities that could cause poor alignment to the template brain and lead to inaccurate results of voxelwise analyses. However, visual inspection of each brain after spatial alignment revealed no noticeable errors, indicating that our results from voxelwise analyses were not due to misalignment of the FA volumes. Further, with regard to the CST, results of analyses of FA of the CST reconstructed using tractography, an independent approach conducted without spatial alignment of brain image volumes, were qualitatively similar to those using the TBSS approach, supporting the veracity of results of the voxelwise analyses.

CONCLUSION

The findings of this DTI study demonstrate that microstructural status of the ipsilesional and contralesional CST correlates with motor skill of the affected hand in chronic stroke patients. The relative strength of poststroke degeneration of bilateral CST tending to limit motor recovery and poststroke remodeling of bilateral CST tending to improve motor recovery may underlie the FA‐behavior relationship we observed at the chronic time‐point. The determinants of the relative strength of degenerative versus restorative changes of the CST are unknown, yet could involve genetic variations that affect responses to ischemic brain damage. A prospective serial study that examines the evolution of poststroke changes in CST microstructure is required to better understand the underpinnings of the FA‐behavior correlation observed in this study. The results of the current study broaden our view of the factors that may play a role in motor recovery after stroke in patients, from one focused on reorganization of activity of the sensorimotor cortices to including the contribution of changes in microstructural status of bilateral CST.

Supporting information

Acknowledgements

REFERENCES