Dynamic Interactions Between Hemispheres Reveal a Compensatory Pathway for Motor Recovery in Moderate-to-Severe Subcortical Stroke

Huaxin Fan; Hewei Wang; Zhengxu Lian; Qiurong Yu; Xinran Wu; Nanyu Kuang; Benjamin Becker; Jianfeng Feng; Mingxia Fan; Lili Song; Limin Sun; Jie Zhang; Craig S. Anderson

doi:10.5853/jos.2025.01725

J Stroke > Volume 28(1); 2026 > Article

Fan, Wang, Lian, Yu, Wu, Kuang, Becker, Feng, Fan, Song, Sun, Zhang, and Anderson: Dynamic Interactions Between Hemispheres Reveal a Compensatory Pathway for Motor Recovery in Moderate-to-Severe Subcortical Stroke

Original Article

Journal of Stroke 2026;28(1):97-114.

Published online: January 2, 2026

DOI: https://doi.org/10.5853/jos.2025.01725

Dynamic Interactions Between Hemispheres Reveal a Compensatory Pathway for Motor Recovery in Moderate-to-Severe Subcortical Stroke

Huaxin Fan^1,^2,^3,^4,^*, Hewei Wang^5,^*, Zhengxu Lian^1,^2,^*, Qiurong Yu^6,^7,^*, Xinran Wu^1,², Nanyu Kuang^1,², Benjamin Becker^8,⁹, Jianfeng Feng^1,², Mingxia Fan¹⁰, Lili Song^11,¹², Limin Sun^5,¹³

, Jie Zhang^1,²

, Craig S. Anderson^11,^12,¹⁴

¹Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China

²Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence, Ministry of Education, Shanghai, China

³Department of Brain Sciences, Imperial College London, London, UK

⁴UK Dementia Research Institute, Imperial College London, London, UK

⁵Department of Rehabilitation Medicine, Huashan Hospital, Fudan University, Shanghai, China

⁶Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China

⁷Faculty of Medical Imaging Technology, College of Health Science and Technology, Shanghai Jiao Tong University School of Medicine, Shanghai, China

⁸State Key Laboratory of Brain and Cognitive Sciences, The University of Hong Kong, Hong Kong, China

⁹Department of Psychology, The University of Hong Kong, Hong Kong, China

¹⁰Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China

¹¹The George Institute for Global Health, Faculty of Medicine, University of New South Wales, Kensington, NSW, Australia

¹²The George Institute for Global Health China, Beijing, China

¹³Department of Rehabilitation Medicine, Binhai Campus, The First Affiliated Hospital of Fujian Medical University (Fujian Hospital of Huashan Hospital, Fudan University), Fuzhou, China

¹⁴Department of Neurology, Royal Prince Alfred Hospital, Sydney, NSW, Australia

Correspondence: Jie Zhang Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China Tel: +86-150-2657-0528 E-mail: jzhang080@gmail.com

Co-correspondence: Limin Sun Department of Rehabilitation Medicine, Huashan Hospital, Fudan University, Shanghai, China Tel: +86-136-1168-9182 E-mail: tracy611@sina.com

^* These authors contributed equally as first author.

Received April 14, 2025 Revised July 23, 2025 Accepted August 8, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background and Purpose

Therapeutic target selection in noninvasive brain stimulation for poststroke motor recovery typically relies on the interhemispheric inhibition model, which is effective for mildly affected patients but offers limited benefits for severely affected individuals. The mechanisms governing recovery from moderate-to-severe stroke remain poorly understood, which hinders the development of targeted interventions.

Methods

We analyzed resting-state functional magnetic resonance imaging data from patients with unilateral subcortical stroke and moderate-to-severe upper limb deficits, both pre- and postintervention, along with data from healthy controls. We developed a novel dynamic lag analysis method for identifying recovery-related homotopic sensorimotor regions with altered interhemispheric interactions. To further uncover the global reorganization pathway, we developed dynamic lateralization approaches to detect large-scale functional connectivity (FC) alterations associated with the identified regions in transient lateralization states.

Results

Dynamic time-lag analysis revealed significantly reduced synchronized states in the homotopic dorsal premotor cortex (PMd) post-intervention compared with pre-intervention, which correlated with motor recovery. Further dynamic lateralization analysis revealed a prolonged segregation state in patients, characterized by weakened interhemispheric and strengthened intrahemispheric interactions. In this state, patients showed decreased FC in the ipsilesional PMd and increased FC in the contralesional PMd with bilateral subcortical networks. These recovery-related alterations were absent in the traditional static analysis.

Conclusions

Dynamic analyses targeting interhemispheric interactions are valuable for understanding neural reorganization after stroke. The diminished interactions between the homotopic PMd indicate a compensatory mechanism. Importantly, a state-dependent compensatory pathway was identified, wherein the contralesional PMd assumes the functions of the ipsilesional PMd through enhanced interactions with subcortical structures, potentially guiding more effective interventions.

Keywords: Moderate-severe stroke; Dynamic functional connectivity; Laterality; Interhemispheric interaction; Noninvasive brain stimulation

Introduction

Stroke is a leading cause of long-term adult disability worldwide, with approximately 13.7 million new cases annually [1]. Motor impairment is the most common issue among stroke survivors, with over 50% experiencing upper limb motor deficit 6 months poststroke [2,3]. Although many individuals show spontaneous recovery within the initial months [4], stroke remains a primary contributor to permanent disability. As the global population ages, the burden of stroke is expected to rise [5], highlighting the need for optimal neurorehabilitative strategies to maximize treatment outcomes. Upper limb motor recovery after stroke varies significantly among patients [6]. While many patients with mild impairment recover up to 70% of their initial function spontaneously, those with severe impairment typically do not [4,7]. Therefore, developing alternative rehabilitation approaches for severely impaired patients is crucial, necessitating further investigation of their neural reorganization patterns.

Functional neuroimaging provides a noninvasive means to gain insights into the neural mechanisms underlying function recovery and brain network reorganization after stroke [4,6,8]. Among these techniques, functional connectivity (FC) analysis based on resting-state functional magnetic resonance imaging (rsfMRI) is crucial [9,10]. It examines the relationship between activity in different brain regions, thus aiding in understanding the patterns of functional organization related to post-stroke motor deficits or motor recovery after intervention [11,12]. Consistent findings indicate that FC between homotopic primary motor cortex (M1) decreases shortly after stroke onset, reflecting the disruption of balanced interhemispheric inhibition [13]. Reduced inhibition from the ipsilesional M1 (iM1) to the contralesional M1 (cM1) leads to increased activation of cM1, which further inhibits iM1, impairing motor output in the affected hemisphere [14]. Over time, FC tends to restore as motor recovery occurs [15,16]. However, these findings primarily concern patients with mild symptoms. Current therapeutic targets for noninvasive brain stimulation often focus on the neural reorganization observed in mildly affected patients, facilitating the M1 in the affected hemisphere or inhibiting M1 in the unaffected hemisphere [14,17]. This approach shows limited efficacy for individuals with moderate-severe motor impairment [18]. In fact, studies have shown that moderately-severely affected patients exhibited a significant decrease in interhemispheric interaction post-intervention [19,20], indicating a distinct neural reorganization pattern. This discrepancy can be explained by the bimodal balance-recovery model, which emphasizes that recovery mode depends on the extent of damage to the affected hemisphere and its structural reserve [21]. Patients with mild motor impairment and sufficient structural reserve require rebalancing of interhemispheric inhibition to promote motor output from residual tissue in the ipsilesional hemisphere. In contrast, patients with severe motor impairment and minimal structural reserve benefit from reduced interhemispheric balancing, which allows the contralesional hemisphere to perform motor functions [21].

Evidence suggests that interhemispheric FC decreases following rehabilitation in moderately severely affected patients, in contrast to the pattern observed in mildly affected patients. However, this reduction in interhemispheric FC is not correlated with motor function recovery [22,23], and the underlying neural pathway for motor recovery remains unclear. This highlights the need for more sensitive brain network biomarkers and approaches to better capture recovery-related changes. Recently, dynamic FC (dFC) analysis has provided new insights into severely impaired patients with stroke, revealing an altered dynamic balance between integration and segregation of resting-state brain networks [24-26]. One study found that patients more frequently transitioned to a cortical-subcortical network segregation state, which was undetectable through static FC analysis [24]. Another study found that damage to white-matter connections between the basal ganglia/thalamus and cortex led to a bias towards an abnormal cortical network integration state [26]. These findings indicate that FC abnormalities in severely affected patients with stroke are state-dependent. Employing dynamic analyses can enhance our understanding of neural reorganization following stroke. Given that state-dependent modulation is a promising avenue for improving stroke treatment efficacy [27], dynamic analysis may help optimize treatment strategies and improve patient outcomes.

Although the abnormal interaction between the ipsilesional and contralesional sensorimotor cortices (SMCs) after stroke has been widely studied, previous dynamic brain network analyses have not explicitly explored these interactions between hemispheres. This gap underscores the need for more sensitive neuroimaging markers that can characterize subtle interhemispheric interactions reflecting motor recovery, especially those with finer temporal resolution. Decreased homotopic FC is often a result of the hemodynamic lag between blood oxygenation level-dependent (BOLD) signals of homotopic regions, reflecting a perfusion deficit after stroke [28-30]. Exploring dynamic changes in the time lag between homotopic regions could be more sensitive to alterations in these interactions than traditional dFC-based approaches [31]. This may provide deeper insights into the mechanisms of stroke recovery.

In this study, we developed dynamic analysis approaches to characterize the interactions between ipsilesional and contralesional SMC, aiming to uncover the neural reorganization patterns in patients with moderate-to-severe stroke who exhibit significant motor recovery (defined as an average change in Fugl-Meyer Assessment Upper Extremity [FMA-UE] score greater than 4.25, which is the minimal clinically important difference [MCID]) [32]. First, we developed a dynamic time lag analysis to explore the longitudinal changes in functional interaction between homotopic SMC during stroke recovery, identifying the region most closely associated with adaptive plasticity. Given that the decreased interaction between homotopic regions of interest (ROI) is related to the increased lateralization of the interaction between the ROI and other regions [33], we also performed a larger-scale dynamic lateralization analysis based on the whole-brain FC profile of the identified ROI to detect distinct lateralization states. This investigation examined the FC patterns of SMC to uncover the underlying circuitry mechanisms responsible for adaptive plasticity after stroke.

Methods

Participants

Forty patients with stroke from the Department of Rehabilitation Medicine at Huashan Hospital, Fudan University and 43 healthy controls were recruited for this study. The participants were previously enrolled in a published study [34]. Five patients were excluded due to significant head motion during scanning, resulting in a final cohort of 35 patients. Inclusion criteria for pre-intervention patients were as follows: (1) aged 18 to 80 years, (2) first-onset unilateral ischemic or hemorrhagic subcortical stroke, (3) at least 3 months post-stroke, (4) moderate-to-severe motor dysfunction in the hemiplegic upper extremity, as assessed by a FMA-UE score ≤37 [35], and (5) being right-handed. Exclusion criteria were as follows: (1) contraindications to magnetic resonance imaging (MRI), (2) bilateral lesions or involvement of the cerebral cortex, (3) Mini-Mental State Examination (MMSE) score <27, (4) presence of severe hand spasticity, neglect, aphasia, or sensory disturbances, (5) history of alcohol, drug abuse, or epilepsy, and (6) participation in any other experimental rehabilitation or drug studies. For healthy controls, the inclusion criteria were as follows: (1) age 18-80 years and (2) being right-handed. The exclusion criteria were as follows: (1) contraindications to MRI, (2) any history of neurological or psychiatric disorders, and (3) MMSE score <27.

The Ethics Committee of East China Normal University approved this retrospective study (HR 017-2020) and waived the requirement for informed consent.

Intervention and clinical outcome measurement

All patients received rehabilitation interventions 5 days per week for 4 weeks. The daily interventions consisted of three hours of conventional rehabilitation intervention (CRI) that encompassed physical and occupational therapy programs that were widely acknowledged, tailored to individual needs, and routinely practiced by experienced therapists in the rehabilitation center. The interventions targeted both the upper and lower extremities to improve skills in basic activities of daily living. They incorporated techniques such as muscle stretching, active/passive mobilization, neuromuscular facilitation, and task-specific training. A subset of patients (21 out of 35) also underwent one of the following additional treatments, with each session lasting between 20 and 40 minutes: motor imagery training (MIT) (n=5: imagery of upper limb basic movements and goal-directed movements, 30 min), transcutaneous electrical acupoint stimulation (TEAS) (n=9: applied to four acupoints—LI11, LI10, TE5, and LI4—at a pulse duration of 200 μs and 100 Hz, 40 min), or transcranial direct current stimulation (tDCS) combined with task-oriented training (n=7: anodal 2 mA stimulation over the iM1, 20 min).

Prior to and after the 4-week intervention, each patient underwent a motor function assessment using the FMA-UE scale administered by an experienced independent physician. The FMA-UE scale, comprising 33 items evaluated on a 3-point ordinal scale (ranging from 0 to 2 points), has been established as a dependable and valid measure of upper-limb impairment in stroke survivors, demonstrating consistency both within and across raters. The maximum achievable FMA-UE score is 66. The motor recovery score was calculated as FMA-UE_{post-intervention} minus FMA-UE_{pre-intervention}.

Acquisition of neuroimaging data and preprocessing

High-resolution structural T1-weighted images and rsfMRI data were collected at the Shanghai Key Laboratory of Magnetic Resonance at East China Normal University, Shanghai, China. The mean framewise displacement value and maximum displacement of each participant were below 0.5 mm and 1 mm, respectively. Detailed information regarding the acquisition of MRI data and preprocessing is provided in the Supplementary Methods.

Lesion map and image flip

For each patient, the lesion was manually delineated on the corresponding high-resolution T1-weighted images using MRIcron (https://www.nitrc.org/projects/mricron). The resulting binary lesion masks were spatially normalized to the Montreal Neurological Institute (MNI) template. Subsequently, the normalized lesion masks were combined and superimposed onto the MNI template. Prior to data analysis, MRI volumes of 17 patients with right hemispheric lesions were flipped along the midsagittal plane [36]. To mitigate the impact of interhemispheric asymmetry, the MRI data of the same proportion of healthy controls (n=20) underwent identical processing procedure [37]. Figure 1 displays the lesion overlap map after the flip with lesions primarily located in the basal ganglia, internal capsule, thalamus, and corona radiata.

Parcellation of the SMC and the whole brain

All individual functional MRI (fMRI) images were normalized to the MNI space (voxel size=3×3×3 mm³) during preprocessing. Several volumetric atlases in the MNI space were subsequently registered to the normalized fMRI data in the volumetric space. Human Motor Area Template (HMAT) is devoted to the parcellation of SMC, including bilateral M1, dorsal premotor cortex (PMd), ventral premotor cortex (PMv), primary somatosensory cortex (S1), supplementary motor area, and presupplementary motor area [38]. Following a previous study [39], we generated the parcellation of cerebral cortex by integrating the HMAT and the well-established atlases of Human Connectome Project multimodal parcellation [40]. Each cortical region in the resulting parcellation was assigned to one of Yeo’s seven functional networks based on the majority of overlapping voxels [41], with the somatomotor network redefined using the HMAT-derived sensorimotor regions. Furthermore, we parceled the subcortical regions associated with motor function (i.e., the putamen, caudate, pallidum, and thalamus) by automatic subcortical segmentation (ASEG) [42]. All atlases were resampled to a uniform voxel resolution of 3×3×3 mm³ using nearest-neighbor interpolation to ensure spatial consistency across analyses. We treated these subcortical regions as a unified network based on evidence from a previous study, which identified a principal subcortical connectivity component encompassing these regions and demonstrated its relevance in characterizing dynamic brain states after stroke [26]. We therefore generated eight large-scale networks encompassing the visual, somatomotor, dorsal attention, ventral attention, limbic, frontoparietal, default, and subcortical networks. To explicitly characterize the interaction between the two hemispheres after a stroke, we split each of these eight networks into two hemispheres. Thus, eight pairs of large-scale networks were generated. To verify the robustness of our results, we included the bilateral cerebellum in the ASEG as an additional pair of intrinsic brain networks: bilateral cerebellum networks.

Time-varying hemodynamic lag analysis

Abnormal hemodynamic lag has recently been identified in patients with stroke through a time-lagged correlation analysis between the BOLD signals of each voxel and the whole brain, which may reflect a perfusion deficit [28,29]. A recent study indicated that a lag between homotopic ROI leads to a decrease in homotopic FC [30], suggesting that hemodynamic lag could serve as a more fundamental measure of stroke. Therefore, we followed the spirit of dynamic FC analysis and extended the static hemodynamic lag analysis to its dynamic counterpart, using the sliding- window approach [24,43] to identify subtle changes in interhemispheric interaction after intervention. We computed the time lag between the BOLD signals of the two homotopic ROIs in successive time windows. The sliding window length (w) and step were set to 16 repetitions (repetition time [TR]) (1 TR=2 s) and 1 TR, respectively. This choice was based on empirical evaluations across a range of window lengths (10-30 TR, i.e., 20-60 s), consistent with prior recommendations that this range offers a balance between temporal sensitivity and estimation reliability [44-47]. In line with Chen et al. [48], we selected the window length that yielded the most statistically representative and robust results in our analyses. The chosen length also aligns with the minimum reliable window length reported in previous methodological studies [47]. The calculation of lags was restricted to [−w2, w2]. The time-lag (τ) in each window for the homotopic BOLD signals of ipsilesional/left ROI x and contralesional/right ROI y was then assigned as

τ=argmaxτ∑n=1wxnyn+τ, 0≤τ≤w2∑n=1wxn−τyn, −w2≤τ<0

where is the lag corresponding to the maximum cross-correlation coefficient between x and y.

Setting the contralesional ROI as the reference, we defined four states of hemodynamic lag between homotopic ROIs of SMC: synchronized state for τ=0, desynchronized state for τ≠0 including delay state for 0<τ<w2, lead state for −w2<τ<0, and uncorrelated state for |τ|=w2 (Figure 2A). To quantify the temporal distribution of these states, we calculated the proportion of sliding windows assigned to each state relative to the total number of windows, which is referred to as the time spent in each state. To further characterize the temporal dynamics, we computed the average delay and average lead as the mean dynamic time-lag values (τ) across all time windows classified as delay and lead states, respectively.

To identify the indices of dynamic time lag that were most related to motor recovery, we conducted a two-step feature selection. First, we compared measures of dynamic time lag (including the time spent in each of the four states, average delay, and average lead) between patients before and after the intervention. Variables passed false discovery rate (FDR) correction at q<0.1 during the comparison (considering the six pairs of homotopic regions and six neuroimaging variables for a total of 36 multiple comparisons) were selected. Second, we investigated the relationship between the longitudinal change (post-intervention minus pre-intervention) in the selected indices and the longitudinal change in motor function, as assessed by the FMA-UE before and after the intervention. General linear models were employed to control for age, sex, and head motion in the primary analysis. Given the overall motor improvement observed in the cohort following intervention, we hypothesized that post-intervention changes in selected neuroimaging indices would be directly associated with motor recovery. One-sided statistical tests were used to assess associations. This approach is consistent with prior stroke research that applied one-sided tests to examine the hypothesized directional relationships between interhemispheric FC and motor function [49]. Indices showing significant associations with motor function changes (surviving FDR correction at q<0.05) were considered relevant.

Dynamic lateralization analysis

The altered interaction between homotopic SMC regions may impact the larger-scale functional organization of the brain, which can be observed by the laterality index [33]. In conventional static lateralization analysis, patients with stroke exhibited asymmetric patterns of task-state activation and resting-state FC for ipsilesional and contralesional SMC [50,51]. However, the time-varying properties of lateralization in patients with stroke has not been investigated; these properties may help identify transient states of lateralization that reflect subtle alterations in patients, thus facilitating the understanding of reorganization mechanisms.

Building on our previous work on dynamic lateralization analysis [52], we proposed a novel approach to characterize the time-varying lateralization patterns of homotopic SMC ROIs by utilizing their FC profiles (Figure 2B). We first used the same sliding-window arrangements as those used in the time-varying hemodynamic lag analysis. For the pair of SMC regions (by time-varying hemodynamic lag analysis) and each intrinsic resting-state network (RSN), we define a dynamic laterality index (dLI) at each time window t as

dLItk,l=rtROIk,RSNIl−rtROIk,RSNCl

where ROIk represents the ROI in hemisphere k (1 for ipsilesional/left hemisphere, 2 for contralesional/right hemisphere); RSNs include the visual, somatomotor, dorsal attention, ventral attention, limbic, frontoparietal, default, and subcortical networks; RSNl represents the l-th RSN (where l=1, 2, …, 8); I and C represent the ipsilesional/left and contralesional/right hemisphere, respectively; r(∙) denotes the mean Pearson’s correlation coefficient (FC) between ROIk and all brain regions with the RSNl in one hemisphere. Therefore, for each time window, we obtained a 16-dimensional dLI vector (2 ROIs×8 RSNs) capturing hemisphere-specific asymmetries in FC from both ROI perspectives.

To identify recurring lateralization states associated with homotopic SMC regions over time, we concatenated the dLI vectors across all time windows and participants into a 16×24,295 matrix (16 features per time window, 24,295 total windows). This matrix was then subjected to k-means clustering (using the squared Euclidean distance) following the approach in reference [53]. Each time window was assigned to a specific lateralization state based on the clustering results. Silhouette and Calinski-Harabasz indices were used to identify the optimal number of clusters.

We then calculated the proportion of sliding windows assigned to each lateralization state relative to the total number of windows for each participant, which was referred to as the time spent in each lateralization state. These proportions were compared between patients before and after the intervention, as well as with healthy controls. To gain further insight into the neural circuit related to motor recovery, we compared the FC between the identified homotopic SMC and eight pairs of brain networks, as well as the FC between bilateral subcortical networks in each lateralization state among pre- and post-intervention patients and healthy controls. For each time window, SMC-network FC was computed by averaging the FC between the identified SMC and all regions within the target network, and interhemispheric FC between bilateral subcortical networks was computed by averaging 16 correlations between all pairs of subcortical regions across hemispheres (four regions per hemisphere). The reported SMC network and bilateral subcortical FC values for each lateralization state were obtained by averaging the corresponding time window FCs assigned to that state. Traditional static FC analyses were also performed for comparison using a full time series. The covariates age, sex, and head motion were regressed. FDR at q<0.05 was used for correction of multiple comparisons. Finally, we used Pearson’s correlation coefficient to investigate the relationship between the time spent in a synchronized state (time-varying hemodynamic lag analysis) and the time spent in each lateralization state (dynamic lateralization analysis).

Robustness analyses

In addition to repeating the time-varying hemodynamic lag analysis using a range of window lengths (10 TR to 30 TR) to identify the optimal window and assess the stability of the results, we conducted several additional analyses to evaluate the robustness of our findings.

To test the stability of lateralization clustering, we included the cerebellum, given its known contralateral connections to the cerebral motor areas, and examined whether the identified lateralization states and corresponding FC patterns remained stable.

To address the clinical heterogeneity of our study population, including variations in post-stroke interval, motor recovery level, stroke type, and intervention type, we conducted a series of subgroup analyses. Participants were stratified based on (1) post-stroke interval (late subacute [3-6 months] vs. chronic [≥6 months]), (2) motor recovery level (low recovery [ΔFMA-UE <4.25] vs. high recovery [ΔFMA-UE ≥4.25]), (3) stroke type (ischemic vs. hemorrhagic), and (4) intervention type (MIT, TEAS, CRI, and tDCS). Given the limited sample size within each intervention group, leave-one-out analyses were performed to evaluate whether any single intervention type disproportionately influenced the findings. For each stratification, we compared the baseline motor function scores, motor recovery levels, post-stroke intervals, and longitudinal changes in key neuroimaging features (time spent in a synchronized state and average delay of PMd). Furthermore, we examined the longitudinal changes in neuroimaging features and their association with motor recovery within each subgroup.

To further account for the potential confounding effects of clinical characteristics on the associations between changes in neuroimaging features and motor recovery, we conducted supplementary analyses that additionally regressed out baseline motor function, stroke type, and post-stroke interval.

Results

Participant characteristics

Participants included 35 patients with unilateral subcortical stroke and 43 age-matched healthy controls. The pre-intervention demographic and clinical characteristics of patients and healthy controls are shown in Table 1. There was no significant difference in age between the two groups, whereas sex showed a significant difference (27 males and 8 females for patients; 21 males and 22 females for controls; χ²=5.39, P=0.0203). The poststroke time interval and stroke severity were determined at baseline. The post-stroke time interval for 16 patients was 3-6 months, whereas that for the remaining 19 patients exceeded 6 months. Among them, 4 patients experienced moderate strokes, with FMA-UE scores ranging from 31 to 37, whereas 31 patients suffered severe strokes, with FMA-UE scores below 30. The group mean score of FMA-UE in patients before and after intervention were 16.3±11.1 and 24.5±15.2, respectively, and the increased FMA-UE scores were significantly larger than MCID at the level of 4.25 (t=3.24, P=0.0013).

Static FCs between homotopic SMCs were not associated with motor recovery

As a comparison to dynamic network approaches in characterizing motor recovery-related brain reorganization, static FC between homotopic SMCs was calculated by computing Pearson’s correlation between the corresponding BOLD signals of six pairs of homotopic SMC regions over the entire scan. We found that the FCs associated with the bilateral S1, M1, PMv, and PMd were significantly decreased in patients before treatment compared with those in healthy controls (t=-5.37, P<0.0001; t=-4.59, P<0.0001; t=-4.19, P=0.0001; t=-2.75, P=0.0082, respectively, for S1, M1, PMv, and PMd, FDR q<0.05) (Supplementary Table 1). A decreasing trend was observed in the FCs after treatment. However, this decrease was neither significant nor related to motor recovery, as measured by the alteration of FMA-UE after treatment (Supplementary Tables 2 and 3), suggesting the need for more sensitive neuroimaging markers to comprehensively understand the reorganization of the brain related to recovery.

Decreases in the synchronized state of homotopic PMd were associated with motor recovery

To identify the subtle changes in the functional interactions with the homotopic SMC related to motor recovery, we developed a time-varying hemodynamic lag analysis and identified the transient synchronized (zero lag) and desynchronized (non-zero lag) states associated with homotopic SMC regions. To identify the specific hemodynamic lag measure most related to motor recovery, we calculated the correlation between the motor recovery score and the relevant hemodynamic lag measures that demonstrated significant differences (FDR correction at q<0.1) between pre- and post-intervention in patients. We found that the time spent in the synchronized state and the average delay (in the delay state) of the bilateral PMd were most associated with motor recovery. The detailed statistical results of the pre- and post-intervention comparisons are presented in Supplementary Table 4, and the correlations between changes in hemodynamic lag measures and motor recovery are presented in Supplementary Table 5.

First, we found decreased bilateral PMd interactions in patients (both before and after the intervention) compared with controls. This finding was consistent across both static FC (t=-2.75, P=0.0082; t=-4.01, P=0.0002 for pre- and post-intervention respectively; FDR q<0.05) and dynamic time-lag analysis that measured the time spent in the synchronized state of the PMd (t=-2.11, P=0.0388, uncorrected; t=-4.20, P=0.0001, FDR q<0.05, respectively for pre- and post-intervention) (Figure 3A). However, the time spent in the synchronized state of the PMd was also significantly decreased after treatment (t=-2.88, P= 0.0068, FDR q<0.1) (Figure 3A). Furthermore, the greater the decrease in the time spent in the synchronized state after the intervention, the higher the level of motor recovery (r=-0.39, P_one-sided=0.0097, FDR q<0.05) (Figure 3B), which could not be revealed by the static FC analysis (t=-1.57, P=0.1247 for the pre-post-intervention comparison; t=-1.29, P=0.2044 for the correlation between the longitudinal change in static FC and motor recovery).

The average delay of the PMd (in the delay state) was negatively related to the time spent of the PMd in the synchronized state (r=-0.69, P<0.0001) from the time-varying hemodynamic lag analysis. For patients, it significantly increased after treatment (t=4.25, P=0.0002, FDR q<0.05) (Figure 3C), and the greater increase in the average delay of PMd after intervention, the greater motor recovery for these patients with stroke (r=0.33, P_one-sided=0.0249, FDR q<0.05) (Figure 3D). Together, both the time spent in the synchronized state and the average delay of the bilateral PMd showed a trend in which larger deviations of these features from healthy controls corresponded to higher levels of motor recovery during the intervention for patients.

Increased interhemispheric segregation in patients with stroke

The decrease in the synchronized state of the homotopic PMd in patients with stroke in the time-varying hemodynamic lag analysis suggested a decrease in interhemispheric interactions within certain time windows. To further understand how the reduction of the synchronized state of homotopic PMd may affect their larger-scale functional interactions with the whole brain, we carried out dynamic lateralization analysis using the FC profile of the bilateral PMd, that is, the FC between the PMd and eight intrinsic brain networks.

We identified two distinct states of lateralization regarding the bilateral PMd in terms of their connection profile with the whole brain: the segregation state, in which the ipsilesional/left and contralesional/right PMd have large (positive) FCs with the same hemisphere and small (or even negative) FCs with the opposite hemisphere, and the integration state, in which the bilateral PMd have adequate FCs to both the same and opposite hemispheres (Figure 4A). In the segregation state, the FCs of the bilateral PMd lateralized to the same hemisphere, that is, the time-averaged laterality index, remained positive for the ipsilesional/left PMd and negative for the contralesional/right PMd. In the integration state, however, the FCs of the bilateral PMd did not lateralize to any hemisphere, that is, the laterality index was near zero for both the ipsilesional PMd (iPMd) and contralesional PMd (cPMd).

We further compared the time spent in each state between groups and found that patients with stroke demonstrated an increase in the time spent in the segregation state both before and after intervention (t=3.10, P=0.0034, FDR q<0.05; t=3.38, P=0.0016, FDR q<0.05, respectively, for pre- and post-intervention), while there was a decrease in the time spent in the integration state compared to healthy controls (t=-3.10, P=0.0034, FDR q<0.05; t=-3.38, P=0.0016, FDR q<0.05, respectively, for pre- and post-intervention) (Figure 4B). Regarding how the reduction in the synchronized state of the homotopic PMd may affect the lateralization state of functional interactions with the whole brain, we found that the time spent in the synchronized state was negatively related with the time spent in the segregation state (r=-0.70, P<0.0001) (Figure 4C).

FC of iPMd decreased while FC of cPMd increased within the bilateral subcortical network in the segregation state for patients with stroke

To further understand the neural circuits involved in motor recovery, we investigated the differences in FC in the segregation and integration states between pre- and post-intervention, and between patients and healthy controls. We also performed a traditional static FC analysis among these three groups for comparison with dynamic approaches. We found decreased FC between iPMd and bilateral subcortical networks in patients with stroke, both pre- and post-intervention, compared with healthy controls. This decrease was consistent across segregation, integration, and static FC analyses (FDR q<0.05) (Figure 5A). However, when examining FC in the segregation state, we observed increased FC between the cPMd and bilateral subcortical networks in patients with stroke (both pre- and post-intervention) compared to controls. In this state, increased FC between the bilateral subcortical networks was also observed after the intervention (FDR q<0.05) (Figure 5A). These two results were not observed in the integration-state or static FC analysis. Furthermore, the FC between the PMd and the subcortical network in patients with stroke showed asymmetry across the segregation state, integration state, and static FC analyses. In contrast, healthy controls exhibited asymmetry primarily in the segregation state, with relatively more symmetrical FC in the integration state and static FC analyses (Figure 5A and Supplementary Figure 1).

Robustness analyses

Additional analyses were conducted to evaluate the robustness of the findings. First, we repeated the time-varying hemodynamic lag analysis using a range of window lengths (10-30 TR). The results remained generally consistent across window sizes; both the time spent in a synchronized state and the average delay of the bilateral PMd continued to show the same trend: the larger the deviations of these features from healthy controls, the higher the level of motor recovery for patients during the intervention (Supplementary Table 6). Furthermore, these features demonstrated high correlations across different window lengths (r=0.88-0.96 for time spent in synchronized state; r=0.68-0.84 for average delay), supporting the robustness of the findings with respect to window length selection.

Second, we included the bilateral cerebellar networks in the dynamic lateralization analysis. Similar to the original analysis, we identified the segregation and integration states. Consistent with the anatomical projections, the lateralization direction of the PMd-cerebellum connectivity was opposite to that of the PMd-cortical connectivity. Importantly, the key findings in patients with stroke were decreased FC between the iPMd and bilateral subcortical networks, while the increased FCs between the cPMd and bilateral subcortical networks in the segregation state remained largely unchanged (Supplementary Figure 2).

Third, subgroup analyses based on the post-stroke interval, motor recovery level, stroke type, and intervention type demonstrated that although clinical heterogeneity influenced the strength of statistical significance, the patterns of neuroimaging changes and their associations with motor recovery remained consistent with the primary analyses. Detailed results are provided in the Supplementary Results and Supplementary Tables 7-10.

Fourth, to account for potential clinical confounding factors, we repeated the regression analysis relating changes in neuroimaging features to motor recovery while additionally controlling for baseline motor function, stroke type, and post-stroke interval. While the statistical significance was slightly attenuated (t=-1.96, P_one-sided=0.0299 for time spent in the synchronized state; t=1.47, P_one-sided=0.0758, for average delay), the overall patterns were consistent with those observed in the primary analyses.

Discussion

In this study, we examined the dynamic functional brain organization of patients with unilateral subcortical stroke and moderate-to-severe upper limb motor impairment, both before and after an intervention, to identify distinct neural reorganization patterns for motor recovery. By developing dynamic analyses that explicitly explored interactions across hemispheres, we found that (1) the time spent in the synchronized state of the homotopic PMd, already reduced for pre-intervention patients compared to healthy controls, decreased further post-intervention, and such decrease was correlated with motor recovery; (2) the time spent in the synchronized state of the homotopic PMd was negatively related to that of time spent in the segregation state, characterized by weak interhemispheric but strong intrahemispheric connections between PMd and other brain networks, with an increased time spent in the segregation state in patients; (3) in the segregation state, patients showed increased FC between cPMd and bilateral subcortical networks and decreased FC between iPMd and bilateral subcortical networks. Our results suggest that strengthened interaction between the cPMd and bilateral subcortical networks in the segregation state appears to serve as a compensatory pathway for motor recovery in moderate-to-severe stroke, potentially facilitated by disrupted interhemispheric interaction between the bilateral PMd. This study offers a novel perspective on the dynamic interaction between the ipsilesional and contralesional hemispheres in stroke-affected brains and suggests a potential mechanism underlying recovery in patients with moderate-to-severe motor impairment, which could inform future individualized therapeutic approaches.

Connectivity-based approaches utilizing resting-state fMRI have effectively revealed disruptions in the functional network caused by stroke in both animal models and human subjects [8,16]. Previous studies consistently show that patients with stroke exhibit decreased FC with the homotopic M1 compared to healthy controls, correlating with motor impairment and reflecting the disruption of balanced interhemispheric inhibition [13,15]. Similar findings have been observed in rsfMRI studies on rats recovering from induced stroke [54] and in electroencephalography studies showing a strong correlation between disrupted coherent resting-state oscillations within the α/β-band and motor impairments [55,56]. During rehabilitation, FC with the homotopic M1 gradually increased and eventually normalized to the healthy levels [16], supporting the interhemispheric competition model [13]. This model suggests that stroke causes disrupted balanced inhibition between hemispheres, leading to increased activation of cM1 and further motor impairment due to reduced inhibition from iM1 to cM1 [14]. However, much of the previous research has focused on M1 involvement, while the premotor cortex also plays a crucial role in motor recovery post-stroke. The right PMd is essential for directing inhibition towards the left premotor and motor cortex during unilateral pantomimed grasping movements, indicating its role in top-down movement control across hemispheres [57]. In individuals with chronic stroke and moderate-severe upper limb motor impairment, tDCS applied to the ipsilesional premotor cortex significantly increased interhemispheric FC between bilateral premotor cortices, reducing proximal motor deficits [58]. A pilot study involving five participants with moderate-severe chronic stroke revealed an increase in FC between iM1 and cPMd following a 10-day period of iM1 tDCS combined with rehabilitation [59]. Overall, the premotor cortices, particularly the cPMd, are crucial for facilitating functional recovery and motor output following a stroke, likely through the interplay between the cPMd and bilateral SMCs.

Our study highlights the important role of the contralesional hemisphere and the cPMd in motor recovery in patients with stroke with unilateral subcortical lesions and moderate-to-severe impairment. Previous research suggests that contralesional motor areas, particularly cPMd, may compensate for lost ipsilesional functions when a significant portion of iM1 or its corticospinal projections are affected [60]. PMd is crucial for coordinating and regulating complex hand movements [61]. Functionally, more impaired patients show increased activation of cPMd during hand grip tasks [62]. Additionally, transcranial magnetic stimulation (TMS) to cPMd causes a greater performance disruption in patients with greater impairment during reaction time tasks [63]. Structurally, PMd is connected to the reticulospinal tract and the red nucleus spinal tract, which control the ipsilateral proximal limb [64]. Animal studies have also identified uncrossed corticospinal ventral tracts originating ipsilaterally from the PMd, comprising about 10%-20% of the corticospinal tract (CST).65 These findings suggest a structural basis for the compensatory role of cPMd in taking over lost ipsilesional functions in patients with stroke with moderate to severe motor impairment.

Guided by the interhemispheric competition model, repetitive TMS (rTMS) can facilitate iM1 or inhibit cM1 to enhance interhemispheric balance and improve motor function [66]. However, treatment efficacy is inconsistent, with some patients showing improvements and others not [67], particularly those with greater impairment [18]. This suggests a distinct recovery mechanism for this subgroup, which is not fully explained by the interhemispheric competition model. In our study, we used time-varying hemodynamic lag analysis to identify altered interactions in the homotopic SMC between patients (pre- and post-intervention) and healthy controls. Consistent with previous findings of reduced interhemispheric interactions post-stroke [15], we observed a reduced time spent in the synchronized state of the homotopic PMd in patients both before and after intervention, which may reflect a disruption of balanced interhemispheric inhibition [13]. In contrast to the generally reported increase in homotopic FC during rehabilitation, our study found a significantly reduced time spent in the synchronized state of the homotopic PMd after intervention. Moreover, a greater reduction in synchronized state time correlated with better motor function recovery. These findings suggest that a continued decrease in the interaction between the bilateral PMd or the persistence of interhemispheric imbalance may be linked to motor recovery in patients with subcortical stroke with more severe deficits. Our results also showed a trend in which a larger deviation of this feature in patients, compared to healthy controls, corresponded to higher levels of motor recovery after intervention, which may reflect a compensatory recovery pattern in moderate-to-severe stroke. To ensure robustness, we tested multiple sliding window lengths (10-30 TR) and found that key time lag indices, particularly the time spent in the synchronized state of the PMd, were highly correlated across window lengths. Despite some variability in statistical significance, the core findings across window lengths reinforce their reliability and align with methodological recommendations for balancing temporal resolution and estimation stability [48,68]. As patients in this study received various interventions, the observed decrease in bilateral PMd interactions may represent a common neural mechanism underlying motor recovery across these interventions. This inverse relationship between bilateral PMd interaction and motor recovery raises questions regarding the guiding role of the interhemispheric competition model in noninvasive brain stimulation for moderate-to-severe stroke. Our results support the bimodal balance-recovery model, which suggests that the interhemispheric competition model is relevant for mildly impaired patients with adequate structural reserve, while severely impaired patients benefit from reduced interhemispheric balancing, as the contralesional hemisphere aids functional recovery in those with little structural reserve [21]. Two studies focusing on patients with moderate-to-severe stroke also reported decreased interhemispheric FC during recovery [19,20], aligning with our findings. Therefore, the reduced time spent in the synchronized state in our results may reflect a specific compensatory neural reorganization associated with motor recovery in moderate-to-severe patients compared to that in mild patients. This reorganization may be linked to a disrupted interhemispheric balance, as indicated by the reduced interaction with the homotopic PMd, and could potentially support the compensatory role of the unaffected hemisphere.

To explore how the reduction in the synchronized state of the homotopic PMd influences large-scale interactions between the PMd and other brain networks, we employed dynamic lateralization analysis using the FC profile of the homotopic PMd. This approach allowed us to investigate the compensatory neural pathways involved in motor recovery, particularly those related to the unaffected hemisphere and PMd, within specific time windows where compensatory mechanisms might emerge (i.e., the synchronized state), as highlighted by dynamic time-lag analysis. We identified two distinct states of lateralization that delineated interhemispheric interaction at the macroscopic level: segregation and integration. The segregation state is characterized by the lateralization of the bilateral PMd FC to the same hemisphere, with weak interhemispheric FC and strong intrahemispheric FC. This state has been described but not explicitly defined in previous stroke studies using traditional static approaches, which showed decreased interhemispheric FC and increased intrahemispheric FC [69], and decreased interhemispheric FC was related to increased intrahemispheric FC [70], indicating an abnormal brain state. Our results are consistent with these findings, showing that patients with stroke spent more time in the segregation state than controls. This increased time in the segregation state was associated with a decrease in the time spent in the synchronized state of the homotopic PMd, which could reflect a neural reorganization process that potentially contributes to motor recovery. This pattern suggests that disrupted interaction with the homotopic PMd may be linked to weakened interhemispheric connections and strengthened intrahemispheric connections, which could support the neural remapping of motor functions. We also identified an integration state characterized by no lateralization of bilateral PMd FC to any hemisphere, with comparable interhemispheric and intrahemispheric FC. The time spent in the integration state was reduced in patients, suggesting that dynamic approaches may provide additional insights into the macroscopic brain states that are sensitive to motor impairment in moderate-to-severe stroke. Additionally, the reversed lateralization observed in the PMd-cerebellum connectivity during the segregation state aligns with known contralateral cerebellar- cortical projections, further supporting the anatomical plausibility and robustness of our findings.

To locate the compensatory pathway for motor recovery within the identified lateralization state, we found that in the segregation state, patients exhibited decreased FC between the iPMd and bilateral subcortical networks and increased FC between the cPMd and bilateral subcortical networks compared to the controls. This suggests that the cPMd compensates for the motor control function of the iPMd. Its strengthened interactions with bilateral subcortical networks likely serve as a compensatory pathway for motor recovery in patients with moderate-to-severe stroke, as supported by previous functional and structural studies. Functionally, similar increases in FC in the contralesional hemisphere have been reported in patients with severe unilateral lesions [54,71]. Structurally, rats with unilateral cortical lesions show abnormal increased projections from the contralesional cerebral cortex to the ipsilesional striatum [72]. Additionally, unilateral cortical lesions can lead to the expansion of the CST controlling the affected limb and projections to ipsilesional red nucleus, both originating from the contralesional motor cortex, while projections from the ipsilesional motor cortex to bilateral subcortical regions are reduced [73]. We also observed increased FC between bilateral subcortical networks in the segregation state after intervention, likely driven by the increased interaction between cPMd and these networks. These findings imply that the compensatory pathway that promotes motor recovery may be effective only in the segregation state, suggesting the potential benefits of state-specific interventions.

In summary, our results revealed a distinct pattern of neural reorganization that facilitates motor recovery in patients with moderate-to-severe impairment who experienced unilateral subcortical stroke during the late subacute to chronic phase, differing from that of mildly impaired patients. Specifically, the cPMd assumes the function of the iPMd through increased interaction with bilateral subcortical networks in an interhemispheric disconnection state (segregation state), possibly induced by a reduced interaction with homotopic PMd (synchronized state) (Figure 5B). Notably, this mechanism may apply only to unilateral subcortical lesions, and further research on bilateral or cortical damage is required. Our findings suggest an alternative strategy for noninvasive brain stimulation in patients with stroke with moderate-to-severe impairment: facilitating cPMd to enhance its compensatory role rather than following the traditional approach of facilitating iM1 or inhibiting cM1 to restore interhemispheric balance [14]. Facilitating the unaffected hemisphere has been suggested for severely affected patients with stroke [74], and has been shown to promote forelimb function recovery via ipsilateral pathways in rats [75]. Several randomized clinical trials are currently underway to assess the efficacy of high-frequency rTMS or anodal tDCS on cPMd in patients with stroke with severe motor impairment [76-78]. Notably, a pilot randomized controlled trial (RCT) conducted by our team has confirmed the efficacy of this approach, demonstrating a more significant improvement in FMA-UE scores following high-frequency rTMS on cPMd in patients with stroke in the subacute phase, compared to those receiving traditional low-frequency rTMS on cM1 or sham stimulation [79]. Future studies should further explore whether the neural mechanisms underlying this treatment strategy are consistent with the findings of the current study. Additionally, the patients in this study exhibited a state-dependent rehabilitation mechanism, where the compensatory pathway promoting motor recovery was observed only in the segregation state, highlighting the importance of closed-loop noninvasive brain stimulation for optimizing recovery.

The strengths of our study include the collection of data from patients with stroke both before and after the intervention. These longitudinal neuroimaging and motor function data allowed us to identify motor recovery-related neural reorganization, providing an efficient way to pinpoint compensatory pathways. Methodologically, we introduced two dynamic analysis approaches to characterize the interhemispheric interaction between homotopic SMC regions (local) and the FC profile of the PMd (global) to understand neural reorganization. Traditional static FC methods, which characterize neuronal activity coupling, are influenced by hemodynamic lag, particularly in patients with stroke [29,30]. Studies have attempted to correct this lag by shifting the time series [29,80]. Therefore, the traditional FC-based approach may not accurately depict functional interactions between regions. In this study, we found that the hemodynamic lag-based approach could detect subtle changes in functional interactions with the homotopic SMC in patients with stroke, correlating with motor recovery during intervention.

Our study has several limitations. First, the sample size was small and data were collected from a single site, limiting the generalizability of our findings. Further studies are needed to replicate these results in larger multisite cohorts. Second, the patients in this study exhibited clinical heterogeneity in terms of the post-stroke interval, motor recovery level, stroke type, and intervention type. There were no group differences in the neuroimaging feature changes, and the general patterns of these changes and their associations with motor recovery in the subgroups were consistent with the primary analyses, collectively implying a general neural reorganization pattern in patients with moderate-to-severe stroke, regardless of the specific approaches for subgrouping. Among the four intervention groups, patients who received MIT showed the greatest motor recovery, and the strength of the statistical significance in the association between neuroimaging changes and motor recovery varied across subgroups. Together, these findings suggest that different interventions may influence distinct neuroplastic pathways and that additional mechanisms beyond those captured by the neuroimaging features in the current study may also contribute to recovery. Moreover, the small sample sizes within the subgroups may have contributed to fluctuations in statistical significance. Future research using larger, more homogeneous cohorts and targeted interventions, such as the cPMd facilitation intervention for moderate-to-severe stroke recently tested in an RCT by our team [79], is needed to validate and extend these mechanistic insights. Third, we performed only correlation analyses; the causal relationships between the decreased synchronized state and increased segregation state and between the strengthened interaction of the cPMd with bilateral subcortical networks in the segregation state and motor recovery require further investigation. Fourth, the present study did not include patients with mild motor impairment, which limited our ability to establish whether the observed cPMd-subcortical pathway represents a compensatory mechanism specific to moderate to severe stroke. Future studies should investigate dynamic FC alterations across the whole brain connectome in patients with mild and severe impairments to directly identify severity-specific compensatory pathways that may extend beyond the PMd-subcortical axis. Finally, the small proportion of females in the cohort may have introduced sex-dependent biases in stroke manifestations, necessitating further research to explore the effect of sex on stroke mechanisms.

Conclusions

Our study developed dynamic analyses of interhemispheric interaction, offering a novel framework for understanding neural reorganization in patients with moderate-to-severe upper limb deficits following unilateral subcortical stroke during the late subacute to chronic phase. We suggest that the strengthened interaction between the cPMd and bilateral subcortical networks in the interhemispheric segregation state may serve as a compensatory pathway for motor recovery, potentially advancing noninvasive brain stimulation strategies for these patients.

Supplementary materials

Supplementary materials related to this article can be found online at https://doi.org/10.5853/jos.2025.01725.

Supplementary Methods, Supplementary Results

jos-2025-01725-Supplementary-Methods,Results.pdf

Supplementary Table 1.

Comparisons of static interhemispheric functional connectivity between pre-intervention patients and healthy controls

jos-2025-01725-Supplementary-Table-1-4.pdf

Supplementary Table 2.

Comparisons of static interhemispheric functional connectivity between patients before and after intervention

jos-2025-01725-Supplementary-Table-1-4.pdf

Supplementary Table 3.

Correlations between change of motor function and change of static interhemispheric functional connectivity in patients

jos-2025-01725-Supplementary-Table-1-4.pdf

Supplementary Table 4.

Comparisons of variables in time-varying hemodynamic lag analyses between patients before and after intervention

jos-2025-01725-Supplementary-Table-1-4.pdf

Supplementary Table 5.

Correlations between change of motor function and change of variables identified in time-varying hemodynamic lag analyses

jos-2025-01725-Supplementary-Table-5,6.pdf

Supplementary Table 6.

Robustness analyses of the time-varying hemodynamic lag analysis when window length ranged from 10 TR to 30 TR

jos-2025-01725-Supplementary-Table-5,6.pdf

Supplementary Table 7.

Clinical and neuroimaging comparisons between heterogeneity subgroups

jos-2025-01725-Supplementary-Table-7-9.pdf

Supplementary Table 8.

ANOVA of clinical and neuroimaging variables across intervention subgroups

jos-2025-01725-Supplementary-Table-7-9.pdf

Supplementary Table 9.

Post hoc analysis of significant ANOVA effects across intervention subgroups

jos-2025-01725-Supplementary-Table-7-9.pdf

Supplementary Table 10.

Robustness analyses of the time-varying hemodynamic lag analysis across clinically heterogeneous subgroups

jos-2025-01725-Supplementary-Table-10.pdf

Supplementary Figure 1.

FC alterations of patients in integration state and static FC analyses. (A) The difference of FC between bilateral PMd and subcortical networks in integration state among patients of pre- and post-intervention and healthy controls. (B) The difference of FC between bilateral PMd and subcortical networks in static FC analysis among patients of pre- and post-intervention and healthy controls. pre, pre-intervention; post, post-intervention; HC, healthy controls; FC, functional connectivity; PMd, dorsal premotor cortex; I, ipsilesional; Subc, subcortical network; C, contralesional. **P<0.01; ***P<0.001; ****P<0.0001.

jos-2025-01725-Supplementary-Fig-1,2.pdf

Supplementary Figure 2.

Robustness analyses of the dynamic lateralization analysis when included bilateral cerebellum networks. (A) The average dynamic laterality in the segregation state and integration state. (B) The difference of FC between bilateral PMd and subcortical networks in segregation state among patients of pre- and post-intervention and healthy controls. LI, average dynamic laterality index; VN, visual network; SMN, somatomotor network; DAN, dorsal attention network; VAN, ventral attention network; LN, limbic network; FPN, frontoparietal network; DMN, default network; SN, subcortical network; CN, cerebellum network; PMd, dorsal premotor cortex; I, ipsilesional; C, contralesional; pre, pre-intervention; post, post-intervention; HC, healthy controls; FC, functional connectivity; Subc, subcortical network. *P<0.5; **P<0.01; ***P<0.001; ****P<0.0001.

jos-2025-01725-Supplementary-Fig-1,2.pdf

Notes

Funding statement

This study was supported by STI2030-Major Project (2021ZD0200204 to Jie Zhang), the Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence (Fudan University), Ministry of Education, China, the National Natural Science Foundation of China (81974356 to Limin Sun, 82102665 to Hewei Wang, 81471651 to Mingxia Fan), Key Supporting Discipline Construction Project of the Shanghai Health System (2023ZDFC0304 to Limin Sun), Shanghai Natural Science Foundation (23ZR1408500 to Limin Sun), the Natural Science Foundation of Fujian Province (2025J01724 to Limin Sun), Joint Funds for the Innovation of Science and Technology, Fujian Province (2021Y9130 to Limin Sun), Shanghai Sailing Program (21YF1404600 to Hewei Wang), and China Scholarship Council (202406100214 to Huaxin Fan).

Conflicts of interest

The authors have no financial conflicts of interest.

Author contribution

Conceptualization: Huaxin Fan, Jie Zhang. Study design: Huaxin Fan, Jie Zhang. Methodology: Huaxin Fan, Xinran Wu, Jie Zhang. Data collection: Hewei Wang, Qiurong Yu, Mingxia Fan, Limin Sun. Investigation: Hewei Wang, Qiurong Yu. Statistical analysis: Huaxin Fan, Zhengxu Lian, Xinran Wu, Nanyu Kuang. Writing—original draft: Huaxin Fan, Hewei Wang, Qiurong Yu. Writing—review & editing: Huaxin Fan, Hewei Wang, Qiurong Yu, Xinran Wu, Benjamin Becker, Jianfeng Feng, Mingxia Fan, Lili Song, Craig S. Anderson, Limin Sun, Jie Zhang. Funding acquisition: Jie Zhang, Limin Sun, Hewei Wang, Mingxia Fan, Huaxin Fan. Approval of final manuscript: all authors.

Figure 1.

Overlapped lesion map of 35 patients. The color bar represents the number of patients with lesions in a specific voxel. Left side indicates the left hemisphere. Right hemisphere lesions were flipped across the midline to appear on the left hemisphere.

Figure 2.

Framework of dynamic analyses for interhemispheric interaction. (A) Time-varying hemodynamic lag analysis. Setting the contralesional ROI of the SMC as the reference, four states of hemodynamic lag (τ) between homotopic ROIs were identified: synchronized state for τ=0, desynchronized state for τ≠0 including delay state for τ>0, lead state for τ<0, and uncorrelated state for |τ|>w2 with w being the length of sliding window. (B) Dynamic lateralization analysis. With the same assignment of sliding windows as that in the time-varying hemodynamic lag analysis, a 16-dimension vector representing the lateralization of FC between homotopic SMC ROIs and eight intrinsic brain networks was calculated at each time window. Then, the concatenated vectors across time windows and participants were clustered in time to identify the reoccurring lateralization states. SMC, sensorimotor cortex (the subarea of sensorimotor cortex identified in time-varying hemodynamic lag analysis that related with motor recovery); BOLD, blood oxygenation level-dependent; dLI, dynamic laterality index; Ipsi., ipsilesional; Cont., contralesional; ROI, regions of interest; FC, functional connectivity.

Figure 3.

Motor recovery-related synchronized state of bilateral PMd. (A) Difference in the time spent in synchronized state of bilateral PMd among patients at pre- and post-intervention and healthy controls. (B) Correlation between the change in the time spent in the synchronized state of the bilateral PMd and changes in FMA-UE scores during intervention. (C) Difference in the average delay of bilateral PMd among patients at pre- and post-intervention and healthy controls. (D) Correlation between the change in the average delay of bilateral PMd and the changes in FMA-UE scores during intervention. In (C) and (D), every point indicates a patient. PMd, dorsal premotor cortex; pre, pre-intervention; post, post-intervention; HC, healthy controls; FMA-UE, Fugl-Meyer Assessment Upper Extremity (max score=66). **P<0.01; ***P<0.001; ****P<0.0001.

Figure 4.

Two states of lateralization of the bilateral PMd. (A) The upper part shows the segregation state characterized by strong (positive) intrahemispheric while weak (or even negative) interhemispheric connections between PMd and other brain networks; the bottom part demonstrates the integration state characterized by both adequate interhemispheric and intrahemispheric connections between PMd and other brain networks. (B) Difference in the time spent in the segregation and integration states among patients at pre- and post-intervention and healthy controls. (C) Correlation between the time spent in the segregation and synchronized states of the PMd, with each point indicating a patient at pre- or post-intervention. PMd, dorsal premotor cortex; LI, average dynamic laterality index; VN, visual network; SMN, somatomotor network; DAN, dorsal attention network; VAN, ventral attention network; LN, limbic network; FPN, frontoparietal network; DMN, default network; SN, subcortical network; FC, functional connectivity; pre, pre-intervention; post, post-intervention; HC, healthy controls. **P<0.01.

Figure 5.

FC alterations of patients in the segregation state. (A) Differences in FC between bilateral PMd and subcortical networks in the segregation state among patients at pre- and post-intervention and healthy controls. (B) Schematic diagram illustrating the compensatory pathway for motor recovery of patients with moderate-to-severe stroke. pre, pre-intervention; post, post-intervention; HC, healthy controls; FC, functional connectivity; PMd, dorsal premotor cortex; I, ipsilesional; Subc, subcortical network; C, contralesional. *P<0.5; **P<0.01; ***P<0.001; ****P<0.0001.

Table 1.

Participant characteristics

	Patients (n=35)	HC (n=43)	t/χ²	P
Age (yr)	52.7±10.2	56.3±9.6	t_70.8=-1.55	0.1248
Sex			χ²=5.39	0.0203
Male	27	21
Female	8	22
Lesion side, left/right	19/16
Stoke type
Ischemic/hemorrhagic	20/15
Post-stroke time interval (month)	10.8±8.1
FMA-UE	16.3±11.1

Values are presented as mean±standard deviation or n.

HC, healthy controls; FMA-UE, Fugl-Meyer Assessment Upper Extremity (max score=66).

References

1. Saini V, Guada L, Yavagal DR. Global epidemiology of stroke and access to acute ischemic stroke interventions. Neurology 2021;97(20 Suppl 2):S6-S16.

2. Lee KB, Lim SH, Kim KH, Kim KJ, Kim YR, Chang WN, et al. Six-month functional recovery of stroke patients: a multi-time-point study. Int J Rehabil Res 2015;38:173-180.

3. Kwakkel G, Kollen BJ, van der Grond J, Prevo AJ. Probability of regaining dexterity in the flaccid upper limb: impact of severity of paresis and time since onset in acute stroke. Stroke 2003;34:2181-2186.

4. Koch PJ, Park CH, Girard G, Beanato E, Egger P, Evangelista GG, et al. The structural connectome and motor recovery after stroke: predicting natural recovery. Brain 2021;144:2107-2119.

5. Fontana L, Kennedy BK, Longo VD, Seals D, Melov S. Medical research: treat ageing. Nature 2014;511:405-407.

6. Stinear CM. Prediction of motor recovery after stroke: advances in biomarkers. Lancet Neurol 2017;16:826-836.

7. Ward NS. Restoring brain function after stroke - bridging the gap between animals and humans. Nat Rev Neurol 2017;13:244-255.

8. Grefkes C, Fink GR. Connectivity-based approaches in stroke and recovery of function. Lancet Neurol 2014;13:206-216.

9. Kim YH, Cho AH, Kim D, Kim SM, Lim HT, Kwon SU, et al. Early functional connectivity predicts recovery from visual field defects after stroke. J Stroke 2019;21:207-216.

10. Lim JS, Kang DW. Stroke connectome and its implications for cognitive and behavioral sequela of stroke. J Stroke 2015;17:256-267.

11. Guo Z, Jin Y, Bai X, Jiang B, He L, McClure MA, et al. Distinction of high- and low-frequency repetitive transcranial magnetic stimulation on the functional reorganization of the motor network in stroke patients. Neural Plast 2021;2021:8873221.

12. Puig J, Blasco G, Alberich-Bayarri A, Schlaug G, Deco G, Biarnes C, et al. Resting-state functional connectivity magnetic resonance imaging and outcome after acute stroke. Stroke 2018;49:2353-2360.

13. Lee J, Park E, Lee A, Chang WH, Kim DS, Kim YH. Alteration and role of interhemispheric and intrahemispheric connectivity in motor network after stroke. Brain Topogr 2018;31:708-719.

14. Boddington LJ, Reynolds JNJ. Targeting interhemispheric inhibition with neuromodulation to enhance stroke rehabilitation. Brain Stimul 2017;10:214-222.

15. Rehme AK, Grefkes C. Cerebral network disorders after stroke: evidence from imaging-based connectivity analyses of active and resting brain states in humans. J Physiol 2013;591:17-31.

16. Cassidy JM, Mark JI, Cramer SC. Functional connectivity drives stroke recovery: shifting the paradigm from correlation to causation. Brain 2022;145:1211-1228.

17. Lefaucheur JP, Aleman A, Baeken C, Benninger DH, Brunelin J, Di Lazzaro V, et al. Evidence-based guidelines on the therapeutic use of repetitive transcranial magnetic stimulation (rTMS): an update (2014-2018). Clin Neurophysiol 2020;131:474-528.

18. Cunningham DA, Potter-Baker KA, Knutson JS, Sankarasubramanian V, Machado AG, Plow EB. Tailoring brain stimulation to the nature of rehabilitative therapies in stroke: a conceptual framework based on their unique mechanisms of recovery. Phys Med Rehabil Clin N Am 2015;26:759-774.

19. Humphries JB, Mattos DJ, Rutlin J, Daniel AG, Rybczynski K, Notestine T, et al. Motor network reorganization induced in chronic stroke patients with the use of a contralesionally-controlled brain computer interface. Brain Comput Interfaces 2022;9:179-192.

20. Lee J, Chang WH, Chung JW, Kim SJ, Kim SK, Lee JS, et al. Efficacy of intravenous mesenchymal stem cells for motor recovery after ischemic stroke: a neuroimaging study. Stroke 2022;53:20-28.

21. Di Pino G, Pellegrino G, Assenza G, Capone F, Ferreri F, Formica D, et al. Modulation of brain plasticity in stroke: a novel model for neurorehabilitation. Nat Rev Neurol 2014;10:597-608.

22. Carter AR, Patel KR, Astafiev SV, Snyder AZ, Rengachary J, Strube MJ, et al. Upstream dysfunction of somatomotor functional connectivity after corticospinal damage in stroke. Neurorehabil Neural Repair 2012;26:7-19.

23. van Meer MP, Otte WM, van der Marel K, Nijboer CH, Kavelaars A, van der Sprenkel JW, et al. Extent of bilateral neuronal network reorganization and functional recovery in relation to stroke severity. J Neurosci 2012;32:4495-4507.

24. Bonkhoff AK, Espinoza FA, Gazula H, Vergara VM, Hensel L, Michely J, et al. Acute ischaemic stroke alters the brain’s preference for distinct dynamic connectivity states. Brain 2020;143:1525-1540.

25. Bonkhoff AK, Schirmer MD, Bretzner M, Etherton M, Donahue K, Tuozzo C, et al. Abnormal dynamic functional connectivity is linked to recovery after acute ischemic stroke. Hum Brain Mapp 2021;42:2278-2291.

26. Favaretto C, Allegra M, Deco G, Metcalf NV, Griffis JC, Shulman GL, et al. Subcortical-cortical dynamical states of the human brain and their breakdown in stroke. Nat Commun 2022;13:5069.

27. Mrachacz-Kersting N, Stevenson AJT, Jørgensen HRM, Severinsen KE, Aliakbaryhosseinabadi S, Jiang N, et al. Brain state-dependent stimulation boosts functional recovery following stroke. Ann Neurol 2019;85:84-95.

28. Lv Y, Margulies DS, Cameron Craddock R, Long X, Winter B, Gierhake D, et al. Identifying the perfusion deficit in acute stroke with resting-state functional magnetic resonance imaging. Ann Neurol 2013;73:136-140.

29. Siegel JS, Snyder AZ, Ramsey L, Shulman GL, Corbetta M. The effects of hemodynamic lag on functional connectivity and behavior after stroke. J Cereb Blood Flow Metab 2016;36:2162-2176.

30. Siegel JS, Shulman GL, Corbetta M. Measuring functional connectivity in stroke: approaches and considerations. J Cereb Blood Flow Metab 2017;37:2665-2678.

31. Raatikainen V, Korhonen V, Borchardt V, Huotari N, Helakari H, Kananen J, et al. Dynamic lag analysis reveals atypical brain information flow in autism spectrum disorder. Autism Res 2020;13:244-258.

32. Page SJ, Fulk GD, Boyne P. Clinically important differences for the upper-extremity Fugl-Meyer scale in people with minimal to moderate impairment due to chronic stroke. Phys Ther 2012;92:791-798.

33. Gracia-Tabuenca Z, Moreno MB, Barrios FA, Alcauter S. Hemispheric asymmetry and homotopy of resting state functional connectivity correlate with visuospatial abilities in school-age children. Neuroimage 2018;174:441-448.

34. Yu Q, Yin D, Kaiser M, Xu G, Guo M, Liu F, et al. Pathway-specific mediation effect between structure, function, and motor impairment after subcortical stroke. Neurology 2023;100:e616-e626.

35. Woytowicz EJ, Rietschel JC, Goodman RN, Conroy SS, Sorkin JD, Whitall J, et al. Determining levels of upper extremity movement impairment by applying a cluster analysis to the Fugl-Meyer assessment of the upper extremity in chronic stroke. Arch Phys Med Rehabil 2017;98:456-462.

36. Grefkes C, Nowak DA, Eickhoff SB, Dafotakis M, Küst J, Karbe H, et al. Cortical connectivity after subcortical stroke assessed with functional magnetic resonance imaging. Ann Neurol 2008;63:236-246.

37. Nijboer TCW, Buma FE, Winters C, Vansteensel MJ, Kwakkel G, Ramsey NF, et al. No changes in functional connectivity during motor recovery beyond 5 weeks after stroke; a longitudinal resting-state fMRI study. PLoS One 2017;12:e0178017.

38. Mayka MA, Corcos DM, Leurgans SE, Vaillancourt DE. Three-dimensional locations and boundaries of motor and premotor cortices as defined by functional brain imaging: a meta-analysis. Neuroimage 2006;31:1453-1474.

39. Archer DB, Coombes SA, McFarland NR, DeKosky ST, Vaillancourt DE. Development of a transcallosal tractography template and its application to dementia. Neuroimage 2019;200:302-312.

40. Glasser MF, Coalson TS, Robinson EC, Hacker CD, Harwell J, Yacoub E, et al. A multi-modal parcellation of human cerebral cortex. Nature 2016;536:171-178.

41. Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 2011;106:1125-1165.

42. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 2002;33:341-355.

43. Mitra A, Snyder AZ, Blazey T, Raichle ME. Lag threads organize the brain’s intrinsic activity. Proc Natl Acad Sci U S A 2015;112:E2235-E2244.

44. Preti MG, Bolton TA, Van De Ville D. The dynamic functional connectome: state-of-the-art and perspectives. Neuroimage 2017;160:41-54.

45. Zhang J, Cheng W, Liu Z, Zhang K, Lei X, Yao Y, et al. Neural, electrophysiological and anatomical basis of brain-network variability and its characteristic changes in mental disorders. Brain 2016;139(Pt 8):2307-2321.

46. Shirer WR, Ryali S, Rykhlevskaia E, Menon V, Greicius MD. Decoding subject-driven cognitive states with whole-brain connectivity patterns. Cereb Cortex 2012;22:158-165.

47. Jones DT, Vemuri P, Murphy MC, Gunter JL, Senjem ML, Machulda MM, et al. Non-stationarity in the “resting brain’s” modular architecture. PLoS One 2012;7:e39731.

48. Chen J, Sun D, Shi Y, Jin W, Wang Y, Xi Q, et al. Alterations of static functional connectivity and dynamic functional connectivity in motor execution regions after stroke. Neurosci Lett 2018;686:112-121.

49. Chen JL, Schlaug G. Resting state interhemispheric motor connectivity and white matter integrity correlate with motor impairment in chronic stroke. Front Neurol 2013;4:178.

50. Saunders J, Carlson HL, Cortese F, Goodyear BG, Kirton A. Imaging functional motor connectivity in hemiparetic children with perinatal stroke. Hum Brain Mapp 2019;40:1632-1642.

51. Marshall RS, Perera GM, Lazar RM, Krakauer JW, Constantine RC, DeLaPaz RL. Evolution of cortical activation during recovery from corticospinal tract infarction. Stroke 2000;31:656-661.

52. Wu X, Kong X, Vatansever D, Liu Z, Zhang K, Sahakian BJ, et al. Dynamic changes in brain lateralization correlate with human cognitive performance. PLoS Biol 2022;20:e3001560.

53. Wang Y, Wang C, Miao P, Liu J, Wei Y, Wu L, et al. An imbalance between functional segregation and integration in patients with pontine stroke: a dynamic functional network connectivity study. Neuroimage Clin 2020;28:102507.

54. van Meer MP, van der Marel K, Wang K, Otte WM, El Bouazati S, Roeling TA, et al. Recovery of sensorimotor function after experimental stroke correlates with restoration of restingstate interhemispheric functional connectivity. J Neurosci 2010;30:3964-3972.

55. Hordacre B, Goldsworthy MR, Welsby E, Graetz L, Ballinger S, Hillier S. Resting state functional connectivity is associated with motor pathway integrity and upper-limb behavior in chronic stroke. Neurorehabil Neural Repair 2020;34:547-557.

56. Dubovik S, Pignat JM, Ptak R, Aboulafia T, Allet L, Gillabert N, et al. The behavioral significance of coherent resting-state oscillations after stroke. Neuroimage 2012;61:249-257.

57. Bencivenga F, Tullo MG, Sulpizio V, Galati G. Interhemispheric interplay between the left and right premotor cortex during grasping as assessed by dynamic causal modelling. Sci Rep 2023;13:4958.

58. Unger RH, Lowe MJ, Beall EB, Bethoux F, Jones SE, Machado AG, et al. Stimulation of the premotor cortex enhances interhemispheric functional connectivity in association with upper limb motor recovery in moderate-to-severe chronic stroke. Brain Connect 2023;13:453-463.

59. Chen JL, Schlaug G. Increased resting state connectivity between ipsilesional motor cortex and contralesional premotor cortex after transcranial direct current stimulation with physical therapy. Sci Rep 2016;6:23271.

60. Kantak SS, Stinear JW, Buch ER, Cohen LG. Rewiring the brain: potential role of the premotor cortex in motor control, learning, and recovery of function following brain injury. Neurorehabil Neural Repair 2012;26:282-292.

61. Kurata K. Hierarchical organization within the ventral premotor cortex of the macaque monkey. Neuroscience 2018;382:127-143.

62. Bestmann S, Swayne O, Blankenburg F, Ruff CC, Teo J, Weiskopf N, et al. The role of contralesional dorsal premotor cortex after stroke as studied with concurrent TMS-fMRI. J Neurosci 2010;30:11926-11937.

63. Johansen-Berg H, Rushworth MF, Bogdanovic MD, Kischka U, Wimalaratna S, Matthews PM. The role of ipsilateral premotor cortex in hand movement after stroke. Proc Natl Acad Sci U S A 2002;99:14518-14523.

64. Zaaimi B, Edgley SA, Soteropoulos DS, Baker SN. Changes in descending motor pathway connectivity after corticospinal tract lesion in macaque monkey. Brain 2012;135(Pt 7):2277-2289.

65. American Physiological Society. Handbook of physiology: the nervous system Bethesda: American Physiological Society; 1984.

66. Smith MC, Stinear CM. Transcranial magnetic stimulation (TMS) in stroke: ready for clinical practice? J Clin Neurosci 2016;31:10-14.

67. Grefkes C, Fink GR. Disruption of motor network connectivity post-stroke and its noninvasive neuromodulation. Curr Opin Neurol 2012;25:670-675.

68. Allen EA, Damaraju E, Plis SM, Erhardt EB, Eichele T, Calhoun VD. Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 2014;24:663-676.

69. Griffis JC, Metcalf NV, Corbetta M, Shulman GL. Structural disconnections explain brain network dysfunction after stroke. Cell Rep 2019;28:2527-2540.e9.

70. Siegel JS, Ramsey LE, Snyder AZ, Metcalf NV, Chacko RV, Weinberger K, et al. Disruptions of network connectivity predict impairment in multiple behavioral domains after stroke. Proc Natl Acad Sci U S A 2016;113:E4367-E4376.

71. Hong W, Lin Q, Cui Z, Liu F, Xu R, Tang C. Diverse functional connectivity patterns of resting-state brain networks associated with good and poor hand outcomes following stroke. Neuroimage Clin 2019;24:102065.

72. Napieralski JA, Butler AK, Chesselet MF. Anatomical and functional evidence for lesion-specific sprouting of corticostriatal input in the adult rat. J Comp Neurol 1996;373:484-497.

73. Payne BR, Lomber SG. Reconstructing functional systems after lesions of cerebral cortex. Nat Rev Neurosci 2001;2:911-919.

74. Brancaccio A, Tabarelli D, Belardinelli P. A new framework to interpret individual inter-hemispheric compensatory communication after stroke. J Pers Med 2022;12:59.

75. Carmel JB, Kimura H, Martin JH. Electrical stimulation of motor cortex in the uninjured hemisphere after chronic unilateral injury promotes recovery of skilled locomotion through ipsilateral control. J Neurosci 2014;34:462-466.

76. Li J, Wang H, Yuan Y, Fan Y, Liu F, Zhu J, et al. Effects of high frequency rTMS of contralesional dorsal premotor cortex in severe subcortical chronic stroke: protocol of a randomized controlled trial with multimodal neuroimaging assessments. BMC Neurol 2022;22:125.

77. Lee SH, Kim WS, Park J, Kim J, Paik NJ. Effects of anodal transcranial direct current stimulation over the contralesional hemisphere on motor recovery in subacute stroke patients with severe upper extremity hemiparesis: study protocol for a randomized controlled trial. Medicine (Baltimore) 2020;99:e19495.

78. Li X, Lin YL, Cunningham DA, Wolf SL, Sakaie K, Conforto AB, et al. Repetitive transcranial magnetic stimulation of the contralesional dorsal premotor cortex for upper extremity motor improvement in severe stroke: study protocol for a pilot randomized clinical trial. Cerebrovasc Dis 2022;51:557-564.

79. Lu S, Wang H, Yuan Y, Ma T, Liu Z, Li Q, et al. Effects of repetitive transcranial magnetic stimulation on contralesional dorsal premotor cortex in subcortical stroke patients with moderate to severe upper limb motor impairment: a pilot randomized controlled trial. medRxiv [preprint]. 2025. Available from: https://doi.org/10.1101/2025.03.05.25323209.

80. Christen T, Jahanian H, Ni WW, Qiu D, Moseley ME, Zaharchuk G. Noncontrast mapping of arterial delay and functional connectivity using resting-state functional MRI: a study in Moyamoya patients. J Magn Reson Imaging 2015;41:424-430.

	Editorial Office Department of Neurology, Asan Medical Center,Ulsan University College of Medicine 88, Olympic-ro 43-gil, Songpa-gu, Seoul 05505, Korea Submission, status and progress, etc ⟫ E-mail: editor@j-stroke.org Website and system ⟫ E-mail: journal@m2community.co.kr Publishing company ⟫ E-mail: ka72sus@smileml.com
	Copyright © 2026 by Korean Stroke Society.