Disrupted intrinsic functional brain topology in patients with basal ganglia ischemic stroke

Introduction

Ischemic stroke affecting the basal ganglia is a prevalent cerebrovascular disorder primarily localized in the deep brain region. The preservation of normal motor function depends critically on the functional and structural integrity of this area, owing to its critical role as a conduit for the pyramidal tract and a regulator of motor commands. Therefore, patients with ischemic stroke in this region often develop severe contralateral limb movement impairments, leading to a substantial decline in their overall quality of life (1,2). A growing body of research suggests that a link exists between the basal ganglia and language processing (3), cognitive function (including executive function and memory) (4), and emotional regulation- particularly in the context of poststroke depression (PSD) and fatigue (5). Consequently, investigating the motor, language, cognitive function, and other associated neural mechanisms underlying basal ganglia ischemic stroke (BGIS) may provide key insights into research on the functional rehabilitation of individuals affected by BGIS.

Resting-state functional magnetic resonance imaging (rs-fMRI) is a noninvasive neuroimaging technique that can characterize the functional organization of the brain in the absence of external stimuli or tasks. This technology can facilitate investigations into connectivity patterns, functional modules, and interactions within the brain. Moreover, rs-fMRI can be integrated with graph theory analysis (GTA) approaches to clarify the intrinsic functional topological characteristics (i.e., the abstract pattern of connections between brain regions) of the brain, which may in turn help considerably elucidate the mechanisms of brain dysfunction (6).

Previous studies in the field of BGIS have primarily use local metrics (7-9) or seed point-based functional connectivity (FC) analysis (10) to examine the connections within specific brain networks. However, these studies have been limited in their ability to assess changes in the overall topological features of the whole brain (11). By contrast, the application of GTA offers a robust mathematical framework for quantifying the topological properties of brain networks and functional connections between networks. This approach allows for a comprehensive characterization of functional integration (the ability to combine information from distributed brain regions) and segregation (the ability for specialized processing within densely interconnected groups of regions) at a global level throughout the brain, distinguishing GTA from other methods for analyzing brain function (12). Recent research indicates that healthy human brain networks typically display a small-world configuration—a topology characterized by high clustering (segregation) for specialized processing and short path lengths (integration) for efficient communication—which is thought to support efficient information processing (13). Furthermore, human brain networks demonstrate sparse connectivity, striking a balance between order, diversity, and hierarchy (14). In this study, we sought to determine how acute BGIS disrupts this efficient organization.

A comprehensive examination of the related literature suggested that the topological characteristics of brain networks may undergo alterations in individuals affected by stroke (15-17). However, these studies have been largely inconclusive, likely due to variations in sample size, lesion location and volume, the time since stroke onset, and methodology, such as the choice of parcellation atlas and network-density thresholds. For instance, an investigation into the structural brain networks during the initial stages following stroke found a marked decrease in overall network efficiency among individuals with severe strokes as compared to healthy controls (HCs). In contrast, individuals with mild-to-moderate stroke exhibited no significant changes in network efficiency, while all patients with stroke displayed consistent clustering coefficients and modularity (15). A further examination of brain network characteristics in individuals with PSD reported that metrics such as classification, clustering coefficient, feature path length, and local efficiency were higher in patients with PSD than in those without it during task states, whereas overall efficiency was lower in individuals with PSD (17,18). Consequently, we conducted a large-sample study in which the topological properties of stroke-related brain networks were examined and participants categorized based on lesion location.

Due to the intricate pathogenesis of BGIS, further evidence is required to clarify the alterations in the topological properties of brain networks in patients with BGIS and the impact of stroke on the integration and dissociation of brain networks. Therefore, in this study, we used the GTA method and the Dosenbach atlas (19) to examine the modifications in the characteristics of the brain functional network in patients with BGIS and to determine the correlation between these alterations and motor, cognitive, psychological, and other functions. Based on a review of the literature, we hypothesized that patients with acute BGIS exhibit disrupted global network topology and region-specific alterations in nodal centrality. Specifically, we expected to find increased connectivity in motor-related regions and decreased connectivity in cognitive-related regions. We further hypothesized that these alterations are associated with clinical deficits. The primary aims of this study were (I) to compare global and nodal graph theory metrics between patients with BGIS and HCs; and (II) to clarify the relationships between these altered network properties and clinical measures of motor, cognitive, and emotional function. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-317/rc).

Methods

Participants

Between May 2019 and October 2022, a cohort of patients who experienced first-time acute BGIS and who were admitted to the Department of Neurology, The First Affiliated Hospital of Guangxi Medical University were enrolled in this study, as were HCs consisting of local volunteers, hospital staff, and patients’ families recruited via advertisements. Approval for the study protocol was obtained from the Ethics Committee of The First Affiliated Hospital of Guangxi Medical University {ethics approval No. 2021[KY-E-184]}. Before the initiation of the study, all participants provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Prior to enrolment, all participants were fully informed about the study’s purpose and significance. All study procedures strictly complied with the predefined observational protocol.

Inclusion criteria for the acute BGIS group

The inclusion criteria for the acute BGIS group were as follows: (I) age between 18 and 80 years; (II) MRI examination conducted within 1 week of condition onset; (III) a diagnosis of acute ischemic stroke according to “Chinese guidelines for diagnosis and treatment of acute ischemic stroke 2018” (20) as determined by a neurology resident physician in collaboration with an attending physician or a physician of a higher level through comprehensive approaches including medical history collection, physical examination, and imaging examinations; (IV) first-time occurrence ischemic stroke in the basal ganglia region as confirmed by MRI with no other intracranial lesions present except for the stroke-related lesion; (V) a conscious, voluntary patient consent to undergo both MRI examination and scale assessment; (VI) absence of MRI contraindications; (VII) no history of depression or other mental disorders prior to the onset of BGIS; (VIII) no history of alcohol or drug dependence; and (IX) right-handedness as confirmed by the Edinburgh Handedness Inventory.

Exclusion criteria for the acute BGIS group

The exclusion criteria for the acute BGIS group were as follows: (I) any contraindications to MRI examination, including but not limited to cardiac pacemaker implantation, internal metal implantation, and claustrophobia; (II) special physiological conditions, such as pregnancy or lactation in women; (III) other intracranial abnormalities detected by MRI after the onset of BGIS, such as pre-existing ischemic stroke, tumors, brain injuries, or demyelinating diseases detected by MRI of the current condition were excluded; (IV) severe physical diseases (e.g., heart, liver, or kidney disorders) or a history of head injury, epilepsy, or mental disorders; (V) excessive head movement during MRI examination that could lead to poor-quality image data or incomplete data affecting subsequent analysis, defined as a movement of more than 3.0 mm in any of the x, y, or z directions or a head rotation angle >3° during scanning; (VI) white-matter high signal intensity [according to the modified Fazekas score (>1 point)] as determined by professional imaging physicians; and (VII) neuroimaging informatics technology initiative (NIfTI) images with poor normalization after data preprocessing.

Inclusion criteria for the HC group

The inclusion criteria for the HC group were as follows: (I) age between 18 and 80 years; (II) no contraindications to MRI examination; (III) age and education level comparable to those of the patient group; and (IV) no history of myocardial infarction, stroke, or other serious systemic diseases.

Exclusion criteria for the HC group

The exclusion criteria for the HC group were identical to those of the acute BGIS group.

Neurological scale scores

The clinical profile of patients with acute BGIS was characterized during their hospitalization through a series of standardized scales. Neurological deficit was evaluated with the National Institutes of Health Stroke Scale (NIHSS). Motor function and daily living capabilities were assessed with the Fugl-Meyer assessment (FMA) and the Barthel Index (BI), respectively. Furthermore, the Montreal Cognitive Assessment (MoCA) and the Self-Rating Depression Scale (SDS) were implemented to determine cognitive status and the degree of PSD, respectively.

MRI data collection

MRI data collection was carried out at the Department of Radiology, The First Affiliated Hospital of Guangxi Medical University, via a MAGNETOM Prisma 3.0T MRI scanner (Siemens Healthineers, Erlangen, Germany). During the scanning process, all participants were instructed to maintain a supine position, keep their eyes open, and refrain from any body movement. The T1-weighted imaging (T1WI) sequence was acquired as the conventional MRI sequence.

High-resolution three-dimensional sagittal T1W images of the entire brain were obtained via the three-dimensional brain volume sequence. The scanning parameters were as follows: repetition time (TR), 2,300 ms; echo time (TE), 2.98 ms; inversion time, 900 ms; slice thickness, 1 mm; voxel size 1 mm × 1 mm × 1 mm; interslice gap, 0 mm; field of view (FOV), 256 mm × 256 mm; and scanning duration, 5 minutes and 21 seconds. A total of 176 sagittal images were captured.

For rs-fMRI, axial scanning was conducted with the echo-planar imaging (EPI) sequence. The scanning parameters were as follows: TR, 2,000; TE, 35 ms; slice thickness, 3 mm; voxel size, 2.6 mm × 2.6 mm × 3 mm; number of slices, 40; matrix size, 64×64; FOV, 240 mm × 240 mm; flip angle (FA), 90°; and scanning duration, 6 minutes and 20 seconds.

All participants were subjected to the identical MRI scanning sequences and parameters.

fMRI data preprocessing

In this study, the preprocessing of rs-fMRI data was executed with the MATLAB R2022a platform (MathWorks, Natick, MA, USA; http://www.mathworks.com; accessed on August 2022). The preprocessing was carried out via Data Processing & Analysis for Brain Imaging (DPABI) version 7.0 (http://rfmri.org/dpabi; accessed on May 2023) (21), which is based on Statistical Parametric Mapping version 12 (SPM12) (http://www.fil.ion.ucl.ac.uk/spm; accessed on August 2022).

A standardized preprocessing pipeline was implemented for the rs-fMRI data. The initial 10 time points were discarded to mitigate the effects of scanner equilibrium and subject acclimatization. Subsequent steps included the following: slice timing correction for the 40-slice interleaved sequence; head motion correction and exclusion based on a threshold of 3.0 mm/3°; spatial normalization to Montreal Neurological Institute (MNI) space via diffeomorphic anatomical registration through exponentiated lie algebra following coregistration and segmentation of structural images; regression of confounding signals from white matter, cerebrospinal fluid, and head motion (22); detrending to eliminate linear drifts; and finally, band-pass filtering within a 0.01- to 0.10-Hz frequency range.

Network construction

In this study, we employed the Dosenbach atlas (19), which demarcates 160 nodes that are distributed across the whole brain. Given that our study was centered on the investigation of brain networks, the cerebellum was excluded from the analysis. After removal of the 18 nodes located in the cerebellum, a total of 142 nodes were retained. For each of the remaining nodes, a sphere with a 5-mm radius was created centered on each node. Based on the blood-oxygen-level-dependent (BOLD) signal, the blood oxygen levels of all voxels within each sphere were extracted. These values were then averaged to obtain the signal value for each node. The interrelationships between nodes, defined as edges, were determined by calculating the Pearson correlation coefficient of the BOLD signal. Subsequently, these correlation coefficients were transformed into Z-scores via the Fisher r-to-z formula. For each individual, a 142×142 matrix was generated. Finally, a single FC matrix was converted into a binarized adjacency matrix {A_ij = [A_ij]} based on a predefined density threshold. In this matrix, A_ij =1 if the absolute value of the Fisher z-transformed correlation coefficient between regions i and j exceeded the threshold; otherwise, A_ij =0. This absolute-value criterion preserved both strongly positive and strongly negative functional connections as valid network edges (23). Density thresholds were carefully selected to identify significant topological characteristics, such as high global and local efficiencies, within the brain networks under study. We computed binary topological parameters for all FC matrices across a sparsity range of 0.10 to 0.34 (step =0.01). This established range helps mitigate the potential bias introduced from the selection of a single threshold (24). Previous research has indicated that this range is less affected by head movements of the participants, thus ensuring more reliable and stable results in the analysis of brain network topology (22).

Network analysis

The DPABINet tool within the DPABI toolkit was used to calculate both the global and local network metrics of the brain at diverse sparse thresholds. For the global network topology characteristics, the selected metrics included local efficiency (Eloc), global efficiency (Eglob), clustering coefficient (Cp), shortest path length (Lp), normalized clustering coefficient (Gamma), normalized characteristic path length (Lambda), small-worldness (Sigma), assortativity, and modularity. The local network topology attributes included degree centrality (DC), nodal efficiency (NE), nodal clustering coefficient, subgraph centrality, betweenness centrality, and eigenvector centrality. Ultimately, by calculating the area under the curve (AUC) within the sparsity range, we obtained nine global network parameters and six local node attributes. This comprehensive quantification of network characteristics provided a more in-depth understanding of the brain’s topological organization, which was crucial for clarifying the underlying neural mechanisms and relationships among various physiological or pathological conditions (25). To aid interpretation, key graph theory metrics are briefly explained here intuitively. Among the metrics included, the clustering coefficient reflects the extent of local interconnectedness or “cliquishness” of a node’s neighbors, modularity quantifies the degree to which a network can be subdivided into distinct, tightly interconnected groups of nodes (modules), and DC is a node’s number of connections. Betweenness centrality measures how often a node acts as a critical “bridge” along the shortest paths between other nodes. Eigenvector centrality identifies nodes that are connected to other well-connected nodes, suggesting influence within the network. The small-worldness parameter Sigma (σ) quantifies the optimal balance between segregated (high clustering) and integrated (short path length) information processing, a hallmark of efficient brain networks. The Yeo map classifies the human cortex into seven distinct networks. In this study, six of these networks were employed: the somatic motor network [SMN; 29 regions of interest (ROIs) located in the precentral and postcentral gyrus and auditory cortex], ventral attention network (VAN; 16 ROIs located in the supramarginal gyrus, insula, middle frontal gyrus, supplementary motor area), visual network (VN; 22 ROIs located in the occipital lobe and posterior fusiform gyrus), dorsal attention network (DAN; 14 ROIs located in the temporo-occipital cortex, angular gyrus, superior parietal lobule, and premotor cortex), default mode network (DMN; 33 ROIs located in the inferior parietal lobule, posterior cingulate cortex, lateral temporal cortex, and ventral and medial prefrontal cortex), and frontoparietal network (FPN; 21 ROIs located in the superior parietal lobule, precuneus, lateral frontal cortex, and dorsal cingulate cortex). The limbic network identified by Yeo et al. was excluded from the analysis because none of the 142 Dosenbach nodes were present within this network. Drawing on the methods of previous studies, a subcortical network (SCN; seven ROIs located in the putamen and thalamus) was defined to include these nodes. This approach allowed for a more comprehensive consideration of the neural connections relevant to our research, ensuring that all significant nodes were accounted for in the analysis of the brain’s network architecture (26,27).

Statistical analysis

Statistical analyses for demographic comparisons were performed with SPSS version 26.0 (IBM Corp., Armonk, NY, USA; https://www.ibm.com/products/spss-statistics; accessed on August 2022), with the normality of all continuous variables (including demographic data, clinical scales, and global/nodal network metrics) formally assessed via the Shapiro-Wilk test at α =0.05 prior to analysis. Based on normality test outcomes, normally distributed variables were analyzed with parametric tests including two-sample t-tests for group comparisons of age, education years, and network metrics and the Pearson correlation coefficient for determining relationships between normally distributed variables; meanwhile, nonnormally distributed variables were analyzed with nonparametric Mann-Whitney tests for group comparisons of clinical scales and the Spearman rank correlation coefficient for determining the relationships involving nonnormal variables. To control for multiple comparisons, false-discovery rate (FDR) correction (Q <0.05) was implemented for all nodal-level analyses across 142 nodes, while global network metrics (9 metrics) were analyzed without FDR given the limited comparison set. Gender differences were assessed with chi-squared tests, with age, gender, and education level included as covariates in the network metric group comparisons. The two-sample t-test within DPABINet compared area-under-curve values of global and nodal network metrics between groups. Exploratory correlation analyses (with Pearson or Spearman coefficients based on data distribution) and visualizations were conducted with R version 3.5.3 (The R Project for Statistical Computing; https://www.r-project.org; accessed on May 2023) without multiple comparison correction, with all tests performed as two-tailed analyses at a significance threshold of P<0.05.

Results

Demographic information

From the 192 initially eligible participants, four BGIS cases were excluded for incomplete data and 23 cases for excessive head motion, resulting in a final analytical cohort of 165 participants. The BGIS group comprised 82 patients with balanced lesion laterality (left: 46.3%; right: 53.7%), and the HC group comprised 83 HCs. A significant gender difference was detected between the two groups (P<0.05) (Table 1). It was observed that the variables of NIHSS, BI, FMA, and SDS did not follow a normal distribution, while age and MoCA were found to be normally distributed.

Comparisons between patients with BGIS and HCs

Small-world attribute analysis of the functional brain network

In the BGIS group and HC group, Sigma >1 in all the sparsity degrees (Figure 1), indicating that the functional brain network of the BGIS group and HC group had typical small-world attributes.

Figure 1 Group differences in network topological properties between patients with BGIS and HCs. (A) Violin plots illustrating the AUC parameters of the Sigma, Gamma, and modularity for patients with BGIS and HCs. Means and standard deviations are depicted. (B) Sigma, Gamma, and modularity across a density range between 10% and 34%. Each point and error bar denote the mean and standard deviation at each density level, respectively. Asterisks indicate a significant difference at this density threshold. *, P<0.05. AUC, area under the curve; BGIS, basal ganglia ischemic stroke; HCs, healthy controls.

Comparison of global network attributes

Although both sets of brain functional networks exhibited small-world attributes, statistical analysis revealed significant disparities in small-world parameters between the two groups. Specifically, a two-sample t-test demonstrated significantly lower values in patients with BGIS than in HCs for the clustering coefficient (Gamma t =−2.424; P=0.016), small-world index (Sigma t =−2.337; P=0.021), and modularity (t =−2.599; P=0.025) (Figure 1). Conversely, no significant differences were observed for Eglob, Eloc, Cp, Lp, Lambda, or assortativity.

Comparison of local network attributes

In the comparison of local network attributes between the BGIS and HC groups, it was found that the DC of SMN (the right precentral gyrus) was significantly higher in the BGIS group than in the HC group, while the DC of the FPN [the left ventral prefrontal cortex (vPFC)] and DMN [right dorsolateral superior frontal gyrus (sup frontal)] was significantly lower in the BGIS group than in the HC group (FDR-corrected Q <0.05).

The betweenness centrality in the SMN (right precentral gyrus) was significantly higher in the BGIS group than in the HC group, whereas the betweenness centrality in SMN (the right superior parietal gyrus), FPN (left vPFC), and VN [right dorsal anterior cingulate cortex (dACC)] was significantly lower in the BGIS group than in the HC group (FDR-corrected Q <0.05).

Additionally, the eigenvector centrality in SMN (right precentral gyrus) was higher in the BGIS group compared than in HC group, while the eigenvector centrality in the FPN (left vPFC) and SMN (right superior parietal gyrus) was lower in the BGIS group than in the HC group (FDR-corrected Q <0.05).

There were no significant differences in node efficiency, node clustering coefficient, or subgraph centrality (FDR-corrected Q <0.05), as illustrated in Figure 2.

Figure 2 Group differences in network topological properties between patients with BGIS and HCs. Group differences in degree, nodal efficient, betweenness, and eigenvector at the nodal level. Nonsignificant nodes are shown as green spheres, whereas blue (BGIS < HC) and red (BGIS > HC) spheres denote significant differences after FDR correction (Q <0.05). BGIS, basal ganglia ischemic stroke; dACC, dorsal anterior cingulate cortex; FDR, false-discovery rate; HCs, healthy controls; L, left; R, right; sup frontal, superior frontal gyrus; sup parietal, superior parietal gyrus; vPFC, ventral prefrontal cortex.

Correlation between clinical scales and network topological features

We also compared the correlation between clinical scales and network topology measures. Pearson correlation analysis was used to determine the relationships between normally distributed variables, while Spearman correlation analysis was used to determine the relationships between topological characteristics and nonnormally distributed clinical outcomes. However, after several comparison corrections, no significant results were found. Despite significant network alterations being observed, no correlations persisted after rigorous multiple comparisons correction was applied. This lack of association between network-level changes and acute clinical scores underscores the complexity of poststroke recovery.

Post hoc analysis

A post hoc analysis stratified by lesion laterality (left lesion: n=38; right lesion: n=44) was conducted and revealed no significant differences between the groups (all P values >0.05).

Discussion

This study used rs-fMRI to assess the temporal correlation of BOLD fMRI signals across various brain regions. The ROIs identified in the Dosenbach atlas (19) served as the nodes within the brain network, with functional connections between regions serving as edges to establish the brain functional network. The topological characteristics of brain networks in individuals with BGIS were examined in terms of global attributes, small-world attributes, local node attributes, and alterations in FC strength.

Moreover, the topological architecture of functional brain networks was investigated in a well-characterized cohort of acute patients with BGIS and matched HCs through identical GTA protocols. We found disrupted topological integration in patients with BGIS, particularly lower small-worldness (Sigma) and modularity compared with HCs. However, these global impairments manifested alongside enhanced nodal centrality in motor processing networks. Notably, at the nodal level, we identified variability in centrality patterns across functional networks: higher connectivity hubs in the SMN contrasted with lower hubs in the FPN and DMN. These alterations suggested a network-level tradeoff between motor compensation and cognitive integration.

Retention of overall small-world topology

This reduction in small-worldness (σ) indicates a shift toward a more randomized network topology, which is generally less efficient for the specialized yet integrated information processing characteristic of a healthy brain. In the clinical context of stroke, this could reflect a diffuse, less organized neural effort to maintain function after a focal insult. In spite of this, the fact that both the BGIS group and the HC group had Sigma >1 for all sparsity degrees indicates that the fundamental small-world architecture of the functional brain network was maintained in both groups. The small-world property is a characteristic of efficient neural information processing, in which there is a balance between local clustering and global integration. In a small-world network, nodes can quickly communicate with neighboring nodes (high local clustering) and also have relatively short paths to communicate with nodes in other regions (short path length) (28-30). This indicates that despite the presence of BGIS, the overall organization of the brain network still retains a basic structure that is conducive to efficient information flow at the global level.

Impaired local network organization in BGIS

Reduction in standardized clustering coefficient and modularity

The lower values of the standardized clustering coefficient (γ) and modularity in the BGIS group compared to the HC group suggest that the local organization of the brain network in patients with BGIS is disrupted. The clustering coefficient reflects the density of local connections among nodes. A lower γ implies that the local connections are less clustered, meaning that the local information-processing units in the brain may not be as tightly integrated as in the healthy state. Modularity represents the degree to which the network can be divided into distinct, functionally related modules. A decrease in modularity indicates that the distinctiveness and efficient cooperation of these functional modules are affected. For example, in the context of BGIS, the regions involved in motor control and related cognitive functions may have their local connection patterns and modular organizations disrupted, potentially leading to problems such as motor deficits and cognitive impairments.

Stability of information-transfer efficiency and connection preferences

No difference in efficiency-related and assortativity measures

The absence of significant differences in Eloc, Eglob, Cp, Lp, Lambda, and assortativity between the two groups indicates that the basic architecture for efficient information processing is preserved at the global level in the acute phase. The unchanged shortest path length and standardized characteristic path length indicate that the overall “distance” for information to travel between different nodes in the network remains relatively stable. Additionally, the unaltered assortativity shows that the preference of nodes to connect with other nodes of similar characteristics is not affected. This may suggest that the basic connection rules and patterns in the brain network are still in place despite the damage caused by the infarct.

These results, along with the lower standardized clustering coefficient and modularity, indicate that BGIS leads to a disruption of the local organization of the brain network. However, the overall small-world topology and the efficiency of information transfer, as well as the basic connection preferences, are maintained, likely through the activation of compensatory mechanisms.

Altered brain connectivity and functional redistribution

DC changes

The higher DC of the right precentral gyrus in the BGIS group suggests that this region exerts a greater function related to the connection with other brain regions after BGIS (31,32). The precentral gyrus is involved in motor control, and thus this change might indicate a compensatory mechanism in which the brain attempts to maintain normal motor functions by strengthening the connections of this area.

The lower DC of the left vPFC and the right dorsolateral sup frontal indicates that the connections of these regions with other brain areas are weakened. The left vPFC is related to functions such as emotional regulation and decision-making, and the right dorsolateral sup frontal is associated with working memory. The reduced connections may lead to disruptions in these cognitive functions as the channels for information exchange between these regions and other brain areas become narrower.

Betweenness centrality alterations

The elevated betweenness centrality of the right precentral gyrus in the BGIS group shows that it figures prominently as an information-passing “intermediary” (31,32). This could be due to the fact that after BGIS, the information-passing paths are modified, and more information needs to be relayed through the right precentral gyrus to bypass the obstacles caused by the infarcted area.

The lower betweenness centrality of the right parietal gyrus, the left vPFC, and the right dACC means that their roles as information intermediaries are weakened. The right parietal gyrus is related to spatial perception, the left vPFC is involved in cognitive functions, and the right dACC plays a part in emotional and cognitive control. Such changes may affect the information transfer related to these functions, leading to problems in spatial perception, cognition, and emotional control in the brain.

Eigenvector centrality variations

The higher eigenvector centrality of the right precentral gyrus in the BGIS group suggests its heightened importance in the overall brain network. This might be associated with the brain’s attempt to compensate for the impact of BGIS on motor functions, making the right precentral gyrus more prominent in the brain network.

The lower eigenvector centrality of the left vPFC and the right parietal gyrus suggests their reduced importance in the brain network. This may affect the cognitive and spatial perception functions they are responsible for, as their influence in the overall layout of the brain network becomes smaller.

Stability of local connectivity and subgraph structure

The fact that the node clustering coefficient and subgraph centrality was not significantly different between the two groups indicates that the local connection tightness and the central structure of the local subgraphs in the brain network are not significantly affected by BGIS. This may imply that the brain maintains relatively stable local information integration and the core structure of local modules to a certain extent, with the local information-processing functions of the brain persisting.

A key challenge in interpreting poststroke network changes is distinguishing compensatory adaptation from pathological dysregulation. The significantly elevated centrality metrics in the right precentral gyrus likely reflect a compensatory mechanism, with the increased functional integration of this motor-related hub contributing to the mitigation of the motor deficits caused by the subcortical lesion, potentially through increased reliance on nonlesioned cortical pathways and transcallosal disinhibition. This aligns with its role as a primary site for motor execution and its known involvement in poststroke recovery (31,32). Conversely, the decreased centrality in cognitive hubs (e.g., left vPFC and right dorsolateral sup frontal) can be more readily interpreted as impaired network integration contributing to cognitive deficits. However, the long-term functional outcome of such “compensatory” hyperconnectivity is not predetermined and could evolve into efficient recovery or, if excessive and poorly regulated, potentially contribute to maladaptive plasticity. Longitudinal studies tracking these metrics alongside recovery are needed to definitively determine their functional valence.

Correlation between clinical scales and network topological features

The lack of direct correlation between network metrics and clinical scales in our acute-stage cohort does not negate the clinical relevance of our findings. First, the acute phase is characterized by dynamic processes of damage, diaschisis, and early compensation, creating a disconnection between initial network disruption and measurable behavioral output, which may be stabilized by compensatory strategies. Second, despite standardized clinical scales such as NIHSS, FMA, and MoCA being gold standards, they may lack the sensitivity to capture subtle network-mediated deficits or be influenced by compensatory mechanisms that mask the underlying network disruption. Third, the relationship between network function and behavior is likely nonlinear and multifaceted; a single network node might support multiple functions and vice versa. This dissociation between acute network changes and behavior has been reported in another stroke neuroimaging study (33) and is often attributed to the factors outlined above. The clinical meaningfulness of our findings lies in the identification of the potential neural substrates and mechanisms of both impairment (e.g., cognitive network disintegration) and compensation (e.g., motor network hyperconnectivity). These network biomarkers may be more sensitive to change than gross clinical scores in the early phase and could prove valuable for predicting long-term outcomes or stratifying patients for targeted therapies (e.g., neuromodulation of the precentral gyrus) in future longitudinal studies, even if they do not directly correlate with cross-sectional acute assessments.

Temporal evolution of network pathology

The acute-phase findings likely reflect initial network disruption before secondary degeneration, as BGIS typically induces delayed transneuronal degeneration in connected cortical areas via diaschisis. We expect that chronic-phase GTA would reveal exacerbated small-world disruption (further reduced Sigma/Gamma) due to fragmented local clustering from secondary degeneration, alongside strengthened compensatory hubs (higher precentral gyrus centrality), if neurorehabilitation harnessed early plasticity and would reveal emergence of maladaptive changes such as DMN integrity reduction if degeneration proceeded unchecked. This temporal trajectory underscores a critical therapeutic window for ultraearly interventions (<2 weeks), where network-guided strategies such as neuromodulation targeting the precentral gyrus could amplify compensatory mechanisms while activity-dependent neuroprotection could mitigate secondary degeneration, thereby maximizing the functional recovery potential before irreversible network deterioration occurs.

Clinical implications and future directions

The distinct network reorganization patterns observed in this study, notably the hyperconnectivity of the right precentral gyrus and the hypoconnectivity within cognitive networks (FPN and DMN), offer potential biomarkers for poststroke deficits and recovery trajectories. The right precentral gyrus was identified as a critical hub for motor recovery, suggesting it could be a prime target for neuromodulatory interventions such as repetitive transcranial magnetic stimulation or transcranial direct current stimulation aimed at enhancing this compensatory plasticity (34). Conversely, the observed hypoconnectivity in the vPFC and dorsolateral sup frontal provides a network-level correlate for the cognitive and emotional impairments frequently reported in patients with BGIS. This hypoconnectivity provides a network-level correlate for the executive dysfunction, working memory deficits, and emotional dysregulation frequently reported in patients with subcortical strokes affecting the basal ganglia-thalamocortical circuits (4,5). Future studies could investigate whether cognitive remediation therapies can normalize these connectivity patterns. Although our cross-sectional design did not yield robust clinical correlations, likely due to the complex, multifaceted nature of poststroke recovery, longitudinal studies tracking these network changes and detailed clinical profiling are essential to validating their prognostic utility and establishing causal links to functional outcomes.

Limitations

Several limitations to this study should be acknowledged. First, the cross-sectional nature of the analysis prevented us from evaluating the dynamic evolution of network changes and their relationship to long-term functional recovery. Longitudinal tracking is needed to determine whether the observed compensatory hyperconnectivity stabilizes into efficient recovery or evolves into maladaptive patterns. Second, the construction of functional networks without corresponding structural connectivity data (e.g., from diffusion tensor imaging) limits our ability to infer how functional changes are constrained or enabled by the underlying structural wiring. Future research integrating multimodal imaging would provide a more comprehensive view. Third, the absence of robust clinical correlations, despite compelling network findings, highlights the challenge of linking complex network alterations to standard behavioral measures in the acute phase and should be a key consideration for interpreting similar studies. Fourth, while we controlled for gender, age, and education in our statistical models, other potential confounding factors such as specific vascular risk factors (e.g., hypertension, diabetes), lesion volume (beyond location), and medication effects were not explicitly modeled and could have influenced FC measures. Finally, although we performed lesion laterality analysis, the moderate sample size within each subgroup might have the limited power to detect more subtle lateralized effects on network reorganization. Furthermore, while our sample size was substantial for a neuroimaging study of a specific stroke subtype, a formal a priori power analysis for graph theory metrics was not conducted, which is a common challenge in the field due to the lack of established effect sizes.

Conclusions

Our findings revealed that BGIS disrupts the balance between functional integration and segregation in brain networks. The reduced Sigma and Gamma values suggest a tendency toward randomization, likely driven by basal ganglia damage and subsequent compensatory rewiring. The right precentral gyrus emerged as a critical hub, possibly facilitating motor recovery through neuroplastic reorganization. Conversely, weakened centrality in the vPFC and dorsolateral sup frontal may underlie cognitive deficits frequently observed in patients with BGIS. These findings illuminate the neuroplastic mechanisms underlying poststroke recovery and provide a framework for developing network-based interventions. Longitudinal studies integrating multimodal imaging and detailed behavioral assessments are necessary to clarify the dynamic recovery trajectories and validate the prognostic value of these network biomarkers.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-317/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-317/dss

Funding: This study was supported by the National Natural Science Foundation of China (Nos. 82260243 and 82560240), Guangxi Natural Science Foundation (No. 2025GXNSFAA069641), Youth Science Foundation of Guangxi Medical University (No. GXMUYSF202408), and Key Laboratory of Research on Prevention and Control of High Incidence Diseases in Western Guangxi.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-317/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study obtained approval from the Ethics Committee of The First Affiliated Hospital of Guangxi Medical University {ethics approval number: 2021[KY-E-184]}. Before the initiation of the study, all participants provided written informed consent. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Florio TM, Scarnati E, Rosa I, Di Censo D, Ranieri B, Cimini A, Galante A, Alecci M. The Basal Ganglia: More than just a switching device. CNS Neurosci Ther 2018;24:677-84. [Crossref] [PubMed]
2. Kaas J, Stepniewska I. The basal ganglia are a target for sensorimotor domains in posterior parietal, premotor, and motor cortex in primates. Curr Opin Neurobiol 2023;83:102783. [Crossref] [PubMed]
3. Radanovic M, Mansur LL. Aphasia in vascular lesions of the basal ganglia: A comprehensive review. Brain Lang 2017;173:20-32. [Crossref] [PubMed]
4. Zhao L, Biesbroek JM, Shi L, Liu W, Kuijf HJ, Chu WW, Abrigo JM, Lee RK, Leung TW, Lau AY, Biessels GJ, Mok V, Wong A. Strategic infarct location for post-stroke cognitive impairment: A multivariate lesion-symptom mapping study. J Cereb Blood Flow Metab 2018;38:1299-311. [Crossref] [PubMed]
5. Tang WK, Chen YK, Mok V, Chu WC, Ungvari GS, Ahuja AT, Wong KS. Acute basal ganglia infarcts in poststroke fatigue: an MRI study. J Neurol 2010;257:178-82. [Crossref] [PubMed]
6. Yan CG, Wang XD, Lu B, Deng ZY, Gao QL. DPABINet: A toolbox for brain network and graph theoretical analyses. Sci Bull (Beijing) 2024;69:1628-31. [Crossref] [PubMed]
7. Li Q, Hu S, Mo Y, Chen H, Meng C, Zhan L, Li M, Quan X, Gao Y, Cheng L, Hao Z, Jia X, Liang Z. Regional homogeneity alterations in multifrequency bands in patients with basal ganglia stroke: A resting-state functional magnetic resonance imaging study. Front Aging Neurosci 2022;14:938646. [Crossref] [PubMed]
8. Quan X, Hu S, Meng C, Cheng L, Lu Y, Xia Y, Li W, Liang H, Li M, Liang Z. Frequency-Specific Changes of Amplitude of Low-Frequency Fluctuations in Patients with Acute Basal Ganglia Ischemic Stroke. Neural Plast 2022;2022:4106131. [Crossref] [PubMed]
9. Chen H, Zhan L, Li Q, Meng C, Quan X, Chen X, Hao Z, Li J, Gao Y, Li H, Jia X, Li M, Liang Z. Frequency specific alterations of the degree centrality in patients with acute basal ganglia ischemic stroke: a resting-state fMRI study. Brain Imaging Behav 2024;18:19-33. [Crossref] [PubMed]
10. Nijboer TCW, Buma FE, Winters C, Vansteensel MJ, Kwakkel G, Ramsey NF, Raemaekers M. No changes in functional connectivity during motor recovery beyond 5 weeks after stroke; A longitudinal resting-state fMRI study. PLoS One 2017;12:e0178017. [Crossref] [PubMed]
11. Larivière S, Ward NS, Boudrias MH. Disrupted functional network integrity and flexibility after stroke: Relation to motor impairments. Neuroimage Clin 2018;19:883-91. [Crossref] [PubMed]
12. Medaglia JD. Graph Theoretic Analysis of Resting State Functional MR Imaging. Neuroimaging Clin N Am 2017;27:593-607.
13. Gong Q, He Y. Depression, neuroimaging and connectomics: a selective overview. Biol Psychiatry 2015;77:223-35.
14. Bullmore E, Sporns O. The economy of brain network organization. Nat Rev Neurosci 2012;13:336-49. [Crossref] [PubMed]
15. Nemati PR, Backhaus W, Feldheim J, Bönstrup M, Cheng B, Thomalla G, Gerloff C, Schulz R. Brain network topology early after stroke relates to recovery. Brain Commun 2022;4:fcac049. [Crossref] [PubMed]
16. Zhu H, Zuo L, Zhu W, Jing J, Zhang Z, Ding L, Wang F, Cheng J, Wu Z, Wang Y, Liu T, Li Z. The distinct disrupted plasticity in structural and functional network in mild stroke with basal ganglia region infarcts. Brain Imaging Behav 2022;16:2199-219.
17. Peng Y, Zheng Y, Yuan Z, Guo J, Fan C, Li C, Deng J, Song S, Qiao J, Wang J. The characteristics of brain network in patient with post-stroke depression under cognitive task condition. Front Neurosci 2023;17:1242543.
18. Mu R, Yang P, Qin X, Zheng W, Li X, Huang B, Zhu X. Aberrant baseline brain activity and disrupted functional connectivity in patients with vascular cognitive impairment due to cerebral small vessel disease. Front Neurol 2024;15:1421283. [Crossref] [PubMed]
19. Dosenbach NU, Nardos B, Cohen AL, Fair DA, Power JD, Church JA, Nelson SM, Wig GS, Vogel AC, Lessov-Schlaggar CN, Barnes KA, Dubis JW, Feczko E, Coalson RS, Pruett JR Jr, Barch DM, Petersen SE, Schlaggar BL. Prediction of individual brain maturity using fMRI. Science 2010;329:1358-61. [Crossref] [PubMed]
20. Chinese Society of Neurology, Chinese Stroke Society. Chinese guidelines for diagnosis and treatment of acute ischemic stroke 2018. Chinese Journal of Neurology 2018;51:666-82.
21. Yan CG, Wang XD, Zuo XN, Zang YF. DPABI: Data Processing & Analysis for (Resting-State) Brain Imaging. Neuroinformatics 2016;14:339-51. [Crossref] [PubMed]
22. Yan CG, Craddock RC, He Y, Milham MP. Addressing head motion dependencies for small-world topologies in functional connectomics. Front Hum Neurosci 2013;7:910. [Crossref] [PubMed]
23. Zhang J, Wang J, Wu Q, Kuang W, Huang X, He Y, Gong Q. Disrupted brain connectivity networks in drug-naive, first-episode major depressive disorder. Biol Psychiatry 2011;70:334-42. [Crossref] [PubMed]
24. Fornito A, Zalesky A, Breakspear M. Graph analysis of the human connectome: promise, progress, and pitfalls. Neuroimage 2013;80:426-44. [Crossref] [PubMed]
25. Achard S, Bullmore E. Efficiency and cost of economical brain functional networks. PLoS Comput Biol 2007;3:e17. [Crossref] [PubMed]
26. Dan XJ, Wang YW, Sun JY, Gao LL, Chen X, Yang XY, Xu EH, Ma JH, Yan CG, Wu T, Chan P. Reorganization of intrinsic functional connectivity in early-stage Parkinson’s disease patients with probable REM sleep behavior disorder. NPJ Parkinsons Dis 2024;10:5. [Crossref] [PubMed]
27. Yang H, Chen X, Chen ZB, Li L, Li XY, Castellanos FX, et al. Disrupted intrinsic functional brain topology in patients with major depressive disorder. Mol Psychiatry 2021;26:7363-71. [Crossref] [PubMed]
28. Watts DJ, Strogatz SH. Collective dynamics of ‘small-world’ networks. Nature 1998;393:440-2. [Crossref] [PubMed]
29. Liao X, Vasilakos AV, He Y. Small-world human brain networks: Perspectives and challenges. Neurosci Biobehav Rev 2017;77:286-300. [Crossref] [PubMed]
30. Lv H, Wang Z, Tong E, Williams LM, Zaharchuk G, Zeineh M, Goldstein-Piekarski AN, Ball TM, Liao C, Wintermark M. Resting-State Functional MRI: Everything That Nonexperts Have Always Wanted to Know. AJNR Am J Neuroradiol 2018;39:1390-9. [Crossref] [PubMed]
31. Zuo XN, Ehmke R, Mennes M, Imperati D, Castellanos FX, Sporns O, Milham MP. Network centrality in the human functional connectome. Cereb Cortex 2012;22:1862-75. [Crossref] [PubMed]
32. Farahani FV, Karwowski W, Lighthall NR. Application of Graph Theory for Identifying Connectivity Patterns in Human Brain Networks: A Systematic Review. Front Neurosci 2019;13:585. [Crossref] [PubMed]
33. Liu X, Qiu S, Wang X, Chen H, Tang Y, Qin Y. Aberrant dynamic Functional-Structural connectivity coupling of Large-scale brain networks in poststroke motor dysfunction. Neuroimage Clin 2023;37:103332. [Crossref] [PubMed]
34. Winhuisen L, Thiel A, Schumacher B, Kessler J, Rudolf J, Haupt WF, Heiss WD. Role of the contralateral inferior frontal gyrus in recovery of language function in poststroke aphasia: a combined repetitive transcranial magnetic stimulation and positron emission tomography study. Stroke 2005;36:1759-63. [Crossref] [PubMed]