Cortical structural changes after subcortical stroke: Patterns and correlates

Abstract Subcortical ischemic stroke can lead to persistent structural changes in the cerebral cortex. The evolution of cortical structural changes after subcortical stroke is largely unknown, as are their relations with motor recovery, lesion location, and early impairment of specific subsets of fibers in the corticospinal tract (CST). In this observational study, cortical structural changes were compared between 181 chronic patients with subcortical stroke involving the motor pathway and 113 healthy controls. The impacts of acute lesion location and early impairments of specific CSTs on cortical structural changes were investigated in the patients by combining voxel‐based correlation analysis with an association study that compared CST damage and cortical structural changes. Longitudinal patterns of cortical structural change were explored in a group of 81 patients with subcortical stroke using a linear mixed‐effects model. In the cross‐sectional analyses, patients with partial recovery showed more significant reductions in cortical thickness, surface area, or gray matter volume in the sensorimotor cortex, cingulate gyrus, and gyrus rectus than did patients with complete recovery; however, patients with complete recovery demonstrated more significant increases in the cortical structural measures in frontal, temporal, and occipital regions than did patients with partial recovery. Voxel‐based correlation analysis in these patients showed that acute stroke lesions involving the CST fibers originating from the primary motor cortex were associated with cortical thickness reductions in the ipsilesional motor cortex in the chronic stage. Acute stroke lesions in the putamen were correlated with increased surface area in the temporal pole in the chronic stage. The early impairment of the CST fibers originating from the primary sensory area was associated with increased cortical thickness in the occipital cortex. In the longitudinal analyses, patients with partial recovery showed gradually reduced cortical thickness, surface area, and gray matter volume in brain regions with significant structural damage in the chronic stage. Patients with complete recovery demonstrated gradually increasing cortical thickness, surface area, and gray‐matter volume in the frontal, temporal, and occipital regions. The directions of slow structural changes in the frontal, occipital, and cingulate cortices were completely different between patients with partial and complete recovery. Complex cortical structural changes and their dynamic evolution patterns were different, even contrasting, in patients with partial and complete recovery, and were associated with lesion location and with impairment of specific CST fiber subsets.


| INTRODUCTION
Subcortical ischemic stroke can lead to structural damage to brain regions adjacent to the stroke lesion and to specific regions of cerebral cortex remote from the lesions (Conrad et al., 2021;Liu et al., 2015), possibly by the mechanism of axonal degeneration (Yu et al., 2009). In response to the structural damage caused by subcortical stoke, some cerebral cortical regions can reorganize themselves to facilitate recovery of the impaired neurological function (Brodtmann et al., 2012;Liu, Peng, et al., 2020). Structural damage and reorganization in the cerebral cortex can be assessed quantitatively in vivo using structural magnetic resonance imaging (MRI) followed by calculation of cortical thickness, surface area, and gray matter volume (GMV) from the imaging data (Gupta et al., 2019). In the past decade, many studies have reported structural alterations in the cerebral cortex in chronic patients with subcortical stroke Diao et al., 2017;Jones et al., 2016;Zhang et al., 2014). However, only a limited number of chronic cortical structural changes reported by one study can be replicated in other studies. The small sample sizes and the lack of independent replication may be one reason for this inconsistency across studies. Thus, the chronic structural changes reliably observed in the cortex after subcortical stroke should be investigated in large samples and replicated in independent data sets.
The location of stroke lesions is another critical factor accounting for some of the observed variation in chronic cortical changes after stroke; stroke lesions in different locations may impair different brain structures and thus result in different cortical changes (Liu, Peng, et al., 2020). This conclusion is supported by a prior study that showed different patterns of brain structural changes between chronic patients with capsular stroke and those with pontine stroke (Jiang et al., 2017). However, details of the correspondence between subcortical stroke locations and the ensuing chronic structural changes in the cortex remain unclear and would be particularly valuable in elucidating the neural mechanism of long-term behavioral impairment (i.e., motor, cognitive functional) in patients with acute stroke. The technique of voxel-based lesion-symptom mapping (VLSM) has been used to establish the correspondence between lesion locations and specific neurological impairments by comparing neurological scores between patients with and without lesions, on a voxel-by-voxel basis (Bates et al., 2003;Meyer et al., 2016). This method has had successes in identifying the relationships between stroke lesion locations and motor impairment (Lo et al., 2010), cognitive deficit (Munsch et al., 2016), and aphasia (Harvey & Schnur, 2015;Magnusdottir et al., 2013). Liu et al. found that acute stroke lesions in the bilateral primary motor area (M1) and right supplementary motor area (SMA) fibers were associated with chronic motor deficits . Moreover, in patients with subcortical stroke, the presence of an acute stroke lesion in the right caudate nucleus and nearby white matter was correlated with chronic attention deficit (Liu et al., 2018). Theoretically, the VLSM approach can be adapted to study the correspondence between stroke lesion locations and specific chronic structural changes in the cortex by replacing the neurological scores by cortical structural measurements.
However, this approach has not been attempted until now.
The corticospinal tract (CST) is the most important white-matter tract for motor output in the human brain, and the impairment of the CST has been associated with motor outcomes in patients with subcortical stroke (Lin et al., 2019;Stinear et al., 2007). Although the CST mainly originates from M1, it also contains contributions from the premotor cortex (PMC), SMA, and primary somatosensory area (S1) (Schieber, 2007;Welniarz et al., 2017).  have reconstructed a fine-grained map of the CST fibers showing their different cortical origins, and have associated impairments of different subsets of these CST fibers with different brain structural and functional changes in patients with subcortical stroke. However, the correspondence between early impairment of specific CST fibers and subsequent slow trends in cortical structural parameters has not been systematically investigated in patients with subcortical stroke.
Many studies have associated cortical structural changes with neurological outcomes (Khlif et al., 2021;Ueda et al., 2019) or motor outcomes (Ueda et al., 2019) in patients with subcortical stroke. Subcortical stroke patients with different degrees of recovery appear to have different patterns of cortical structural change in the chronic stage and different trajectories of evolving cortical change poststroke.
For example, stroke patients achieving only partial recovery (PR) show more significant reductions in cortical thickness in the primary motor area than do those with complete recovery (CR; Zhang et al., 2014).
However, the differences in the evolution patterns of cortical structural changes between stroke patients with different degrees of recovery remain largely unknown, and such information would be helpful for understanding the mechanisms of neurological recovery after subcortical stroke.
In this study, we aimed to (a) identify chronic cortical structural changes in 181 patients with subcortical stroke with a discoveryreplication design and investigate the differences in cortical structural changes between patients with partial and CR; (b) uncover the correspondence of chronic cortical structural changes with acute stroke lesion location, as well as with early impairments of specific CST fiber subsets; and (c) explore the differences in the evolution patterns of cortical structural change between patients with different degrees of recovery in a longitudinal data set of 81 patients with subcortical stroke.

| Participants
The experimental protocol was approved by the local medical research ethics committee. Written, informed consent was obtained from each participant. This study retrieved cross-sectional and longitudinal patients with subcortical ischemic stroke from four hospitals.
We divided our samples into three independent data sets: the first data set (114 patients with chronic subcortical stroke) was used for discovering chronic cortical structural changes; the second data set (8 independent patients with chronic subcortical stroke) was used for replicating chronic cortical structural changes; and the third data set (81 longitudinal patients with subcortical stroke) was primarily used for observing the longitudinal cortical structural changes. To establish a link between observations of poststroke cortical structural changes, 59 patients with chronic data from data set 3 were also included the replication analysis (8 patients from data set 2 and 59 patients from data set 3).
The inclusion criteria for patients with stroke were as follows: (a) first-onset acute ischemic stroke; (b) a single lesion in the basal ganglia or neighboring regions; and (c) right-handedness before stroke onset (Oldfield, 1971). The exclusion criteria were as follows: (a) recurrent stroke defined by clinical history and MRI evaluation; (b) any other brain abnormalities on MR images; (c) a modified Fazekas score for white matter hyperintensities greater than 1 (Fazekas et al., 1987); and (d) history of any other neurological or psychiatric disorders.
In compiling data set 1 and 2, we reviewed the MRI images and clinical data of inpatients within 5 years, and diffusion weighted imaging (DWI) in the acute stage (≤7 days) were used to confirm stroke location. Patients in the chronic stage after stroke onset (>6 months) were included. Patients meeting the inclusion criteria were prospectively recruited to participate in this study. Patients for data set 1 were recruited from August 1, 2014 to March 30, 2015. A total of 123 patients who satisfied the inclusion criteria agreed to participate in this study. Nine patients were excluded for recurrent stroke (n = 3), severe white matter hyperintensity (n = 3), and history of traumatic brain injury (n = 3). Finally, 114 patients for discovery were included in this study. Patients for data set 2 and 3 were recruited from June 1, 2017 to April 30, 2020. A total of 8 patients with subcortical stroke for replication were included in data set 2. In collection of the data set 3, 95 patients with subcortical stroke were recruited followed a longitudinal design (four time points: ≤7 days, 1 month, 3 months and >6 months). Fourteen patients were excluded for loss of followup after inclusion (n = 9), recurrent stroke (n = 3), and other brain abnormalities (n = 2). Finally, 81 longitudinal patients with subcortical stroke (52 patients with data for four time points and 29 patients with data for at least two time points) were included in this study. Healthy controls for data sets 1 and 2 were prospectively recruited from the population sample with the following criteria: (a) sex-, age-, and education-matched with the patients; (b) free from neurological dysfunction; (c) free from brain structural damage by MRI examination; (d) no history of alcohol or drug dependency; (e) in good physical condition for image acquisition; (f) Fazekas score for white matter hyperintensity ≤1 (Fazekas et al., 1987); and (g) lacunae absent. Finally, 103 cross-sectional healthy controls (data set 1) and 10 longitudinal healthy controls (data set 2) with data for all the four time points (Supplementary Table 1) were included in this study.

| Neurological assessments
The National Institutes of Health Stroke Scale was used to assess global neurological deficits, and the Fugl-Meyer Assessment of the whole extremity (WE_FM, i.e., the total score of the upper-and lower-extremity motor assessment) was used to evaluate motor deficits in patients. These assessments were performed in the chronic stage (>6 months) for patients in cross-sectional data set and at four time points (≤7 days, 1 month, 3 months, and >6 months) for patients in longitudinal data set. For both data sets, patients with subcortical stroke were divided into PR (WE_FM <100) and CR (WE_FM = 100) subgroups according to the WE_FM scores in the chronic stage (>6 months). In 30 patients with PR of longitudinal data set, 13 patients with data for 4 time points and 17 patients with data for at least 2 time points were included. In 51 patients with CR of longitudinal data set, 39 patients with data for 4 time points and 12 patients with data for at least 2 time points were included (Supplementary Table 1).

| Calculation of cerebral cortical measures
All structural images (3D-T1WI) were visually inspected by two radiologists for apparent artifacts due to subject motion and instrument malfunction. Then, the FreeSurfer (FS) V.6.0.0 application (Fischl, 2012) (http://surfer.nmr.mgh.harvard.edu/) was used to preprocess brain structural images and calculate maps of cortical thickness, surface area, and GMV for each participant. The preprocessing procedures were performed using the automated surface-based pipeline with default parameters, which mainly included segmentation, surface reconstruction, and surface-based spatial registration. Specifically, structural images were registered to the Talairach atlas, and intensity variation in the white matter was removed by intensity normalization. The skull data were stripped using a deformable template model (Segonne et al., 2004), and white matter was segmented based on intensity and neighbor constraints. The automated analysis pipeline can fail in the presence of the severe structural abnormalities common in stroke (Fischl, 2012). Therefore, to satisfy our predetermined criteria, quality control was performed using Freeview (a visualization tool packaged with FS), involving visual inspection of the segmentation processes by two radiologists with more than 9 years of experience. Then, we manually corrected inaccurate segmentation to improve the FS segmentation results. The gray-white matter surface was obtained by tessellating the gray-white matter boundary and topology correction, then the pial surface was generated by nudging the gray-white matter surface along the T1 intensity gradients to reach the boundary between gray matter and cerebrospinal fluid. Both surfaces were represented by vertices. For each vertex, the distance between the gray-white matter surface and its corresponding pial surface was defined as the cortical thickness (Fischl & Dale, 2000).
Then, the cortical thickness, surface area, and GMV at each vertex were obtained for each participant. The resulting maps of cortical measures of each participant were transformed into an average surface space (fsaverage template, provided in the FS package) using a spherical registration method (Fischl et al., 1999), then a 10-mm fullwidth-at-half-maximum Gaussian spatial smoothing kernel was applied to the surface to improve the signal-to-noise ratio. These maps were used for vertex-based comparisons of cerebral cortical measures between patients with subcortical stroke and healthy controls.

| Cortical structural changes in patients with chronic subcortical stroke
Cross-sectional data set was used to identify cortical structural changes in patients with chronic subcortical stroke. In the discovery sample (114 patients, and 103 healthy controls of data set 1), patients were further divided into a left-hemispheric lesion group (n = 60) and a right-hemispheric lesion group (n = 54), which were separately compared with the control group (n = 103). For each patient group, a general linear model was used to perform vertex-based comparisons between patients and controls for cortical thickness, surface area, and GMV within a cortical mask. Age, sex, scanner variables, mean cortical thickness (only for cortical thickness analysis), total surface area (only for surface area analysis), and total GMV (only for GMV analysis) of the entire cerebral cortex were entered as nuisance covariates. Multiple comparisons were corrected using a Monte Carlo simulation with a voxel threshold of p < .01 and a correction threshold of p < .05 (1000 simulations, full-width-at-half-maximum = 10 mm). Vertices with significant intergroup differences in cortical measures were extracted as seed masks.
In the replication sample (67 patients, and 103 healthy controls, same as to the discovery sample), patients were also divided into lefthemispheric (n = 35) and right-hemispheric (n = 32) lesion groups, which were used for lesion-side-specific validation of the findings obtained from the discovery sample. Within each seed mask, we performed lesion-side-specific vertex-based intergroup comparisons (p < .05) in the replication sample to identify clusters with reliable cortical structural changes in chronic patients with subcortical stroke. The same statistical model and covariates as in the discovery analysis were used in replication. Clusters significant in both discovery and replication samples were defined as regions of interest (ROIs). Oneway analysis of covariance was used to compare differences in cortical measures in each ROI among the PR, CR, and control groups at p < .05 with the same covariates. Cohen's d was used to describe the effect size (ES) of the intergroup differences.

| Correlations between cortical structural changes and motor outcomes
In the left-or right-hemispheric lesion group (cross-sectional data set), we investigated the cortical structural changes by region that correlated with motor deficits. Specifically, we performed partial correlation analysis between cortical structural measures (cortical thickness, surface area, and GMV, each ROI) and the WE_FM scores while controlling for age, sex, and scanner variables. We performed correlation analyses a total of 10 times between cortical structural changes and motor outcomes. We used the Benjamini-Yekutieli (BY) false discovery rate (FDR) method (Benjamini & Yekutieli, 2001) (p < .05) to correct for multiple comparisons (https://warwick.ac.uk/fac/sci/ statistics/staff/academic-research/nichols/software/fdr). The raw p values were BY-FDR corrected in MATLAB (MATrix LABoratory).
The positive dependence correction causes these BY-FDRs to be closer to 1, or more conservative (Murray & Blume, 2021).

| Correlations between cortical structural changes and lesion locations
In the left-or right-hemispheric lesion group (cross-sectional data set), VLSM (Bates et al., 2003) was used to assess the relation between acute stroke lesion locations and chronic cortical structural changes.
DWI has good sensitivity for acute cerebral ischemia, and can accurately describe the location, morphology, and size of stroke lesions.
Therefore, stroke lesions were manually delineated on the normalized DWI acquired in the acute stage. First, individual DWI data were spatially normalized to the EPI template in Montreal Neurological Institute (MNI) space and resampled into 1-mm 3 voxels. Stroke lesions were independently outlined on the normalized DWI using the MRIcron tool (https://www.nitrc.org/projects/mricron) by three radiologists with more than 9 years of experience. The intraclass correlation coefficient for lesion volume was 0.98, and the result of the most senior radiologist was selected as the final lesion contour.
VLSM was performed on the lesion maps against the cortical structural measures of the ROIs in the chronic stage of stroke, with age, sex, lesion volume, and scanner variables as covariates. In this analysis, we only included stroke lesion voxels that were damaged in more than 10% of the patients (Timpert et al., 2015) and only reported stroke lesion clusters that showed significant correlation in more than 10 lesion voxels (McDonald et al., 2017). Correction for multiple comparisons was achieved using a voxel-level FDR method (p < .05, FDR corrected) (Raphaely-Beer et al., 2020).

| Correlations between cortical structural changes and early impairment of CST fibers
In the left-or right-hemispheric lesion group (cross-sectional data set), we performed correlations between early impairment of CST fibers with different cortical origins and chronic cortical structural changes to identify which CST fibers influenced chronic cortical structural changes. Imaging data on acute stroke lesions and the fine-grained map of the CST fibers published by  were used to calculate the percent impairment of each CST fiber class for each patient in the acute stage. For each axial slice showing an overlap between the stroke lesion and a given CST fiber subset, an impairment percentage was calculated as the ratio of the area of the tract in the overlap region to the total area of the CST fiber subset. The largest percentage in these slices was defined as the impairment percentage of the CST fiber in this patient. We then investigated the correlations between early impairment of each CST fiber subset and cortical structural changes in the chronic stage, while controlling for age, sex, scanner variables, and the impairment percentages of other CST subsets. We performed 40 times correlation analyses between cortical structural changes and early CST impairment. To reduce the number of false-positive findings, we used the BY-FDR method (p < .05) to correct for multiple comparisons.

| Evolution of cortical structural changes after stroke
For each ROI with significant cortical structural change in the chronic stage, we used a linear mixed-effects model to investigate the evolution patterns of the cortical structural changes in PR patients, CR patients, and healthy controls (longitudinal data set, 81 patients and 10 healthy controls). The random intercept term accounts for the correlation due to repeated measurements within a single patient (Gibbons et al., 1988). This model allows us to make maximum use of all available data from each patient, even if some time points are missing. All patients are assumed to have a common slope (fixed effect) and only the intercepts are allowed to vary (random effect). The model parameters were estimated by the restricted maximum likelihood method and considered significant if the p values were less than .05. In healthy controls, we characterized the trajectories of these cortical structural changes to establish references to identify strokeinduced changes. In each patient group, we identified significant longitudinal cortical structural changes by assessing the significance of the slopes (p < .05, BY-FDR corrected). For each ROI with chronic cortical structural changes, we investigated the differences in the evolution patterns by comparing the slopes between every pair selected from the PR, CR, and control groups (p < .05, BY-FDR corrected).
To show the multistep nature of the analyses, we provide a flow diagram in Figure 1.

| Demographic and clinical information
The demographic and clinical data of the participants are listed in Table 1. Cross-sectional data set included 181 patients with subcortical stroke (131 men; mean age, 55.5 ± 8.1 years; 114 patients for discovery and 67 patients for replication) and 103 healthy controls (58 men; mean age, 56.1 ± 7.3 years). We collected DWI and conventional MRI data in the acute stage as well as 3D-T1WI and WE_FM data in the chronic stage for these patients. Longitudinal data set included 81 patients with subcortical stroke (61 men; mean age, 53.9 ± 8.9 years) and 10 healthy controls (3 men; mean age, 55.9 ± 5.2 years). We collected DWI, 3D-T1WI, and WE_FM data

| Cortical structural changes in patients with chronic subcortical stroke
In the discovery sample, we found 10 cortical regions with significant structural differences (4 in cortical thickness, 3 in surface area, and 3 in GMV) between chronic stroke patients and healthy controls (red color in Figure 3; Table 2). All cortical structural differences were validated in the replication sample (light blue color in Figure 3; Table 2).

| Correlations between cortical structural changes and motor outcomes
In patients with a left-hemispheric lesion, the WE_FM scores were positively correlated with the cortical thickness of the left postcentral gyrus and frontal pole and with the surface area of the left cingulate gyrus (p < .05). In patients with a right-hemispheric lesion, the WE_FM scores were positively correlated with the cortical thicknesses of the right precentral and lingual gyri, and with the GMV of the left rectus, right cuneus, and lingual gyrus (p < .05). However, these correlations did not survive the BY-FDR correction for multiple comparisons.

| Correlations between cortical structural changes and lesion locations
In patients with lesions in the left hemisphere (cross-sectional data set), acute-stroke lesions in the left putamen (center MNI coordinates: À23, 5, 11; peak MNI coordinates: À19, 2, 19; cluster volume: 965 ml) were correlated with increased surface area in the left temporal pole in the chronic stage (p < .05, FDR correction; Figure 4a). In patients with lesions in the right hemisphere (cross-sectional data set), acute-stroke lesions in the right CST fibers originating from the  Figure 4b,c). Other brain structural changes were not significantly correlated with lesion locations.

| Correlations between cortical structural changes and early CST impairment
For 181 patients with chronic subcortical stroke, Table 4

| DISCUSSION
In this study, we investigated chronic cortical structural changes and their evolution patterns in subcortical stroke patients with different degrees of motor recovery, elucidating the correspondence of these changes with the locations of lesions and the early impairment of specific CST fiber subsets. We found that chronic cortical structural changes and their evolution patterns were largely different between partially and completely recovered patients. We linked some chronic cortical structural changes to specific lesion locations and to the early impairment of specific CST fiber subsets. These findings indicate that early brain impairments suggested by lesion locations are predictive of chronic cortical structural changes, and that the evolution patterns of certain cortical structural changes are predictive of long-term motor recovery in these patients.

| Cortical structural measures can provide complementary information
In contrast to most previous studies, which characterized cortical structural changes using either cortical thickness or GMV (Jiang et al., 2017;Ueda et al., 2019), we used cortical thickness, surface area, and GMV to characterized cortical structural changes after subcortical stroke and found that different cortical structural measures were changed in different cortical regions. Our findings suggest that these structural measures can provide complementary information about cortical structural changes (Fornito et al., 2008) and should be used in combination to characterize such changes (Buechler et al., 2020).
GMV is a composite indicator, being affected by both cortical thickness and surface area. For example, a simultaneous significant cortical thickness reduction and surface area increase may result in a nonsignificant GMV change; however, a slightly reduced cortical thickness with a slightly reduced surface area may lead to a significant GMV reduction. Cortical thickness measures distances between the gray/white matter interface and the pial surface (the surface between gray matter and cerebrospinal fluid). Cortical area measures the area of the gray-matter surface. Cortical thickness and surface area are two independent dimensions that reflect the microstructural characteristics of the cortex, and multidimensional measurements are thus preferred for noninvasively capturing microstructural changes after stroke. However, the developmental and reorganizational trajectories of cortical thickness and surface area differ (Wierenga et al., 2014), differing in both onset and timing and by anatomical region (Brodtmann et al., 2012). This situation may contribute to the result that one measure may show differences after stroke while another may not.

| Subcortical stroke can result in both cortical structural damage and reorganization
Consistent with previous studies (Jones et al., 2016;Liu, Peng, et al., 2020;Zhang et al., 2014), chronic patients with subcortical  (Jones et al., 2016;Zhang et al., 2014). In this study, we also found cortical thinning in the right precentral gyrus in patients with right subcortical lesions and in the left postcentral gyrus in patients with left subcortical lesions. The structural damage in the ipsilesional sensorimotor cortex may be explained by antegrade and/or retrograde axonal degeneration (Yu et al., 2009), since the subcortical stroke lesions can directly impair the output or input fibers of the sensorimotor cortex or impair brain regions that are connected with the sensorimotor cortex (Duering et al., 2012;Duering et al., 2015). However, the structural damage observed in the left rectus in patients with right subcortical lesions cannot be easily explained by the mechanism of axonal degeneration and requires further clarification.
The structural reorganization of the cerebral cortex was scattered through frontal, occipital, and temporal lobes remote from the stroke lesions. The structural reorganization in the prefrontal cortex found in previous studies (Fan et al., 2013;Jiang et al., 2017;Liu, Peng, et al., 2020) and in this study (frontal pole and orbitofrontal gyrus) indicate that cognitive-related cortical regions may be involved in motor recovery and that cognitive strategy may play a beneficial role in motor recovery in patients with subcortical stroke (McEwen et al., 2009). In line with prior studies (Al Harrach et al., 2019;Fan et al., 2013), we also found structural reorganization in several occipital cortical regions. Although the functional significance and neural mechanism of occipital cortical reorganization after subcortical stroke remain unknown, occipital reorganization may be related to motor recovery since these cortical changes were marginally correlated here with motor outcomes (p < .05, uncorrected).

| Cortical structural changes depend on lesion location in subcortical stroke
In this study, stroke patients with lesions in the left and right hemispheres showed structural changes in different cortical regions, which is consistent with the great variation in chronic cortical structural changes following subcortical stroke reported across studies (Diao et al., 2017;Jiang et al., 2017;Liu, Peng, et al., 2020). VLSM showed that right lesions involving CST fibers originating from the primary motor area resulted in cortical thinning in the right precentral gyrus, indicating that retrograde degeneration is the neural mechanism of structural damage in the ipsilesional precentral gyrus (Duering et al., 2015;Yu et al., 2009). VLSM also demonstrated that stroke lesions in the left putamen were associated with increased sur- specific subcortical region can lead to structural reorganization in a specific cortical region. We used a fine-grained map of the cortical origins of CST fibers to determine the degrees of impairment in CST fiber subsets, and uncovered the impacts of early impairments in specific CST subsets on cortical structural changes. Using the CST damage-cortical change association study, we found that early impairment of the CST fibers originating from the primary somatosensory cortex was associated with structural reorganization in the occipital cortex. This may represent a large-scale compensatory mechanism in which the visual system reorganizes itself to compensate for impairments in the somatosensory system (Pundik et al., 2018 3.1 Â 10 À5 / 1.5 Â 10 À3 * 4.5 Â 10 À3 /7.0 Â 10 À2 0.7/1.0 6.4 Â 10 À2 /0.6 3.1 Â 10 À2 /0.3 4.7 Â 10 À4 / 1.2 Â 10 À2 * IL lingual gyrus 2.9 Â 10 À2 /0.3 2.4 Â 10 À2 /0.3 0.8/1.0 2.3 Â 10 À3 / 4.0 Â 10 À2 * 0.3/1.0 7.5 Â 10 À2 /0.7
patients and is an indicator of good recovery.

| CONCLUSION
Both chronic cortical structural changes and their evolution patterns were studied in patients with subcortical stroke. The cortical changes and their evolution differed between partially and completely recovered stroke patients. This effect may be used to screen imaging biomarkers for predictors of motor outcomes. We established the correspondence of some cortical structural changes with the location of the acute stroke lesion and with early impairment of CST fibers with specific cortical origins. These findings are useful for developing approaches for the early prediction of motor outcomes, which is helpful in the design of individualized rehabilitation plans.