Stroke is a major public health concern in China, characterized by high rates of morbidity, disability, mortality, and recurrence, along with substantial economic burden [1]. It remains the leading cause of death and long-term disability among Chinese adults [2]. Acute basal ganglia cerebral infarction (BGCI), a common stroke subtype involving lesions in the basal ganglia, often leads to motor deficits, impaired higher cognitive function, and emotional-behavioral abnormalities, severely affecting patients’ ability to live independently [3]. Motor impairment is the most frequent cause of post-stroke disability, yet motor recovery varies considerably among individuals [4]. These behavioral deficits may stem not only from localized tissue damage but also from disruptions in functional brain networks [5].

Recent advances in neuroimaging have advanced the understanding of brain function after stroke. Resting-state functional magnetic resonance imaging (rs-fMRI) has been widely used to investigate functional abnormalities in stroke patients, helping to reveal underlying mechanisms of recovery [6]. rs-fMRI captures low-frequency spontaneous neural activity via blood oxygenation level-dependent (BOLD) signals. Many studies have focused on functional connectivity within or between networks—such as the motor network and other functional systems—in stroke patients [7, 8]. Fewer studies have examined local spontaneous brain activity, which is essential for a comprehensive understanding of post-stroke neurophysiological changes.

One common metric for assessing spontaneous regional brain activity is the amplitude of low-frequency fluctuation (ALFF), which quantifies the power of BOLD oscillations within the 0.01–0.08 Hz frequency band and is thought to reflect the intensity of local neural activity [9]. ALFF alterations in specific brain regions have been linked to motor impairment severity in acute stroke, suggesting its utility in evaluating brain tissue viability after ischemic injury [10].

However, ALFF is sensitive to the scale of the BOLD signal, limiting its applicability in group-level analyses. To overcome this limitation, the percent amplitude of fluctuation (PerAF) has been proposed as a scale-independent alternative suitable for cross-group comparisons [11]. PerAF has shown promise in neuropsychiatric disorders such as Alzheimer’s disease, where increased PerAF in frontal and temporal regions has provided insights into disease mechanisms [12]. Such results enhance investigators’ understanding of the underlying neural mechanisms of Alzheimer’s disease and may also contribute to its further exploration [12]. Nevertheless, until now, PerAF has not been applied to study dynamic brain changes during stroke recovery.

Moreover, conventional analyses often assume signal stationarity, overlooking the dynamic nature of brain activity. The dynamic amplitude of low-frequency fluctuation (dALFF) approach addresses this by using sliding windows to capture temporal variability in local neural activity, offering a more nuanced view of brain dynamics [13]. Although dALFF has been used in subacute stroke populations [14], its application in acute BGCI patients across multiple recovery phases remains unexplored.

In this prospective longitudinal study, a homogeneous cohort of acute BGCI patients was recruited, and rs-fMRI scans were performed at multiple time points to investigate temporal changes in PerAF and dALFF across acute, early convalescent, and mid-convalescent stages. It was hypothesized that distinct patterns of motor and cognitive recovery would be reflected in spatiotemporal alterations of local brain activity, thus providing new insights into the dynamic process of post-stroke neural remodeling.

rs-fMRI was chosen as the core tool for its unique advantages: It is non-invasive and suited to longitudinal tracking, enabling repeated measurements across recovery stages without biological risks, while uniquely capturing spontaneous low-frequency neural activity to reflect intrinsic brain function and functional reorganization (even in structurally intact regions) that other imaging methodologies like structural MRI or Positron Emission Tomography (PET) fail to fully reveal [26]. This longitudinal study investigated functional brain remodeling in patients with acute BGCI by combining PerAF and dALFF analyses of rs-fMRI data. To the authors’ knowledge, this is the first study to apply both methods to track neural changes over a three-month period following cerebral infarction. Findings indicate that while PerAF and dALFF capture distinct aspects of brain activity, they provide complementary insights. dALFF results were robust using a 50-TR (100 s) window with 1-TR (2 s) steps. Both methods revealed longitudinal alterations in cortical and cerebellar regions across acute (day 1–7), early convalescent (day 30), and mid-convalescent (day 90) stages, with several changes significantly correlating with clinical scores. These regions span key functional networks—default mode network (DMN), frontoparietal network (FPN), motor network, and cerebellar network—shedding light on functional reorganization after BGCI.

PerAF reflects voxel-wise BOLD signal variability. It allowed time-dependent PerAF changes to be observed across networks: PerAF in left cerebellar lobules 4–5 (lCbe4_5) increased linearly from acute to mid-convalescent phases, whereas PerAF in the left PreCG decreased significantly by the mid-convalescent stage. PerAF in the PCUN.L was notably elevated at day 90.

Schoch et al. [27] found that the anterior cerebellum, specifically lobules IV–V (Cbe4_5), is involved in upper limb locomotion. The cerebellum is anatomically and functionally interconnected with the basal ganglia and cerebral cortex [28]. This cerebellar–cortical loop is integral to motor planning, execution, and higher cognitive and emotional functions [29]. Previous electrophysiological study has shown that Purkinje cells in anterior cerebellar lobules IV–V are activated during arm movement in macaques [30]. Recent research also suggests that gray matter volume changes in this region are associated with motor and cognitive function in patients with multiple system atrophy [31]. The findings highlight the influential role of the anterior cerebellum in motor reorganization. Our observation of a linear increase in lCbe4_5 PerAF suggests enhanced synchronization of neural activity in this region, possibly reflecting compensatory plasticity following BGCI. This implies that left anterior cerebellar lobules IV-V may contribute to motor recovery after basal ganglia infarction. Therapies targeting motor areas, such as transcranial magnetic stimulation, have been reported to improve motor function in acute stroke patients and may be a promising avenue for future intervention in BGCI.

In this study, PerAF in the PreCG.L decreased significantly by the mid-convalescent phase compared to the acute phase, while PerAF in the PCUN.L increased notably by day 90. These findings illustrate how basal ganglia injury affects neural activity in distant regions. PreCG and PCUN belong to the sensorimotor network (SMN) and DMN, respectively. The SMN, which includes pre- and postcentral gyri, shows decreased functional connectivity over time in Parkinson’s disease patients, correlated with motor impairment [32]. PreCG.L is a primary motor control center, and disruption of its connections can lead to motor deficits [33]. The decrease in left PreCG PerAF was negatively correlated with upper limb FMA and BI scores, suggesting that reduced local BOLD signal is associated with better neurological recovery at three months. This indicates that PerAF changes in the PreCG.L are linked to functional remodeling in BGCI patients.

Alternatively, stroke patients typically exhibit reduced DMN activity and lower functional connectivity in the precuneus and other regions compared to healthy controls [34]. Similarly, here it was found that PerAF in the PCUN.L (a DMN node) was significantly lower in the acute phase than in early and mid-convalescent phases. The DMN exhibits high activity during rest and is involved in emotional regulation, self-awareness, and sustained attention [35]. The precuneus is a central DMN hub, and its dysfunction is considered an early marker of cognitive impairment [36]. Thus, the observed changes in PCUN.L PerAF may represent a neuroimaging correlate of post-stroke cognitive changes.

Notably, the observed brain functional remodeling is likely a synergistic effect of intrinsic neuroplasticity (e.g., synaptic reorganization, compensatory network recruitment) and standardized neurorehabilitation received by patients, as structured rehabilitation therapies are known to modulate neuroplasticity and guide adaptive functional reorganization [37, 38].

Further, a positive correlation between PHQ-9 scores (assessing depressive symptoms) and PerAF was observed in the lCbe4_5 (r = 0.398, p = 0.036). This may reflect that enhanced neural synchronization in lCbe4_5—originally interpreted as compensatory plasticity for motor recovery—could also be associated with unresolved depressive symptoms, suggesting a dual involvement of this region in motor and emotional remodeling after basal ganglia infarction.

Two-sample t-tests based on dALFF revealed significant differences among acute patients, mid-convalescent patients, and HCs. Dynamic activity in the right Crus II was significantly reduced in BGCI patients during the acute and mid-convalescent phases compared to controls, consistent with previous findings [39]. The right Crus II is functionally connected to the FPN and DMN and is activated during cognitive tasks [40, 41]. As motor recovery involves relearning motor skills, Crus II may support this process after stroke [42]. In this study, reduced dALFF in right Crus II was observed at both 30-TR and 50-TR windows, suggesting it has a role in motor recovery in BGCI patients. The 100-TR window results did not survive GRF correction, likely due to excessive window width smoothing of temporal dynamics. The lack of correlation between dALFF values and clinical scores may be due to the small number of significant voxels or limited sample size.

This study has several limitations. First, the sample size was relatively small. Future studies should explore frequency-specific changes in larger cohorts of acute BGCI patients. Second, the inclusion of predominantly mild-to-moderate stroke patients limits the generalizability of the findings to the broader stroke population. Although infarct distribution was similar between groups, lesion locations varied between the left and right basal ganglia, which may influence network topology. Future studies should enroll patients with more homogeneous infarct locations. Third, although a longitudinal design was used, the three-month follow-up period was relatively short. An extended follow-up would help further investigate the relationship between functional network dynamics and stroke recovery. Fourth, although the lesions were primarily located in the basal ganglia, distant from DMN nodes, variations in infarct location may still affect functional network measures. Fifth, baseline NIHSS score was not included as a control variable in the correlation analysis between PerAF and clinical functional scores, considering the study’s focus on the strong intrinsic correlation between baseline NIHSS score and clinical outcome scores. Future studies are warranted to comprehensively explore the potential confounding or mediating role of baseline NIHSS score in this relationship. Finally, although demographic and motion confounders were adjusted for, differences in medical history—including unmeasured vascular risks—may affect between-group comparisons and require cautious interpretation.

This study demonstrates that acute BGCI leads to altered spontaneous neural activity in regions including the left Cbe4_5, left PreCG, left PCUN, and right Crus II, as measured by PerAF and dALFF. Notably, the left Cbe4_5 emerges as a key node in cognitive-motor recovery networks, with its dynamic PerAF changes linked to both motor improvement and emotional regulation [43]. These changes are associated with motor and cognitive performance at three months, suggesting they reflect adaptive brain remodeling underlying functional recovery. The cortical–basal ganglia–cerebellar circuit, particularly the cerebellum as a promising target for NIBS techniques, may represent a direction for neuromodulation therapies in BGCI, meriting further investigation [44, 45].

From a translational perspective, our findings support the potential use of PerAF/dALFF patterns as imaging-based stratification biomarkers. For instance, patients showing early elevation of PerAF in the left Cbe4_5 with subsequent normalization may represent a subgroup with better adaptive plasticity and motor recovery. Conversely, persistent hyperactivity in the left PreCG may indicate maladaptive reorganization associated with poorer outcomes. Such stratification could help tailor neuromodulation therapies, such as cerebellar non-invasive brain stimulation, to patients most likely to benefit.

Furthermore, the cortical–basal ganglia–cerebellar circuit—particularly the cerebellum—represents a promising target for neuromodulation in BGCI. Future studies should explore whether baseline PerAF/dALFF profiles can predict response to interventions such as repetitive transcranial magnetic stimulation or transcranial direct current stimulation, thereby advancing toward personalized neurorehabilitation.

Despite the longitudinal insights provided by PerAF and dALFF in capturing post-stroke brain dynamics, the present study was not designed to evaluate their diagnostic classification performance. Future studies with larger, independent cohorts should employ machine learning approaches—such as support vector machine (SVM) or other pattern recognition methods—to examine whether PerAF/dALFF patterns can reliably stratify patients into recovery subtypes or predict therapeutic response. Such efforts would help translate these dynamic neuroimaging markers toward personalized prognosis and intervention planning in stroke rehabilitation.

In summary, our work not only highlights key neural correlates of recovery but also provides a framework for imaging-guided patient stratification and circuit-targeted therapy design, moving closer to precision medicine in stroke rehabilitation.

The datasets supporting the findings of this study, including processed neuroimaging metrics and clinical assessment scores, are available within the article and its supplementary materials. Additional raw data (e.g., de-identified resting-state fMRI data) are available from the corresponding author upon reasonable request.

Conceptualization, SJC and JZ; methodology, YYL; software, SJC and JZ; validation, SJC and JZ; formal analysis, YYL and XMQ; investigation and interpretation of data for the work, QY; visualization and resources, LYP; data curation, LYP; writing—original draft preparation, SJC; writing—review and editing, and design of the work: ZJL; visualization, SJC; supervision, ZJL; project administration, ZJL; funding acquisition, SJC, JZ and ZJL. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Studies involving human patients were reviewed and approved by the First Affiliated Hospital of Guangxi Medical University Review Board (Reference Number: 2021 (KY-E-184)). All procedures were performed in accordance with the ethical standards of the Declaration of Helsinki. Informed consent was obtained from all subjects involved in the study.

We thank the Hangzhou NaoHai Technology Laboratory for the technical support provided by its scientific research platform in this study.

This study was supported by grants from the National Natural Science Foundation of China (82260243, 82560240). Additional support was provided by the Guangxi Natural Science Foundation (2025GXNSFAA069038) and the Liuzhou Science and Technology Bureau (2024YB0103A014).

The authors declare no conflicts of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/JIN47725.