Corticomuscular coupling alterations in subacute stroke patients: insights from fNIRS and sEMG Authors: Xiupan Wei, Xiangshan Wei, Zhi Li, Suhail Jamid, Hongxing Wang Published: 2025 ABSTRACT AbstractBackgroundCorticomuscular coupling (CMC) reflects the neural communication between the central and peripheral nervous systems, particularly during motor control. However, the corticomuscular coupling characteristics during upper limb motor tasks in stroke patients remain unclear. This study aims to explore the differences in corticomuscular coupling characteristics between stroke patients and healthy subjects.MethodsA total of 25 stroke patients in the subacute phase with dominant hemisphere and 20 age-matched healthy controls were enrolled in the study. Participants performed an isometric wrist extension task at 50% of their maximum voluntary contraction (MVC) for 15 s, during which surface electromyography (sEMG) signals measured from the flexor carpi radialis and extensor carpi ulnaris, as well as functional near-infrared spectroscopy (fNIRS) data from the prefrontal cortex, supplementary motor area, and primary motor cortex, were simultaneously recorded. Corticomuscular coupling metrics (phase synchronization index (PSI), coherence), fNIRS measures (cortical activation, functional connectivity), and sEMG parameters (root mean square (RMS), median frequency (MF), and fuzzy approximate entropy (fApEn)) were analyzed.ResultsStroke patients demonstrated significantly reduced PSI and coherence values in specific corticomuscular couplings compared to age-matched healthy controls. Moreover, enhanced cortical activation, stronger functional connectivity, and diminished muscle activation (with lower sEMG complexity) were observed in specific cortical regions and assessed muscles.ConclusionThis study reveals differences in the mapping relationship between cortical activation and sEMG signals across time-domain and frequency-domain indices in stroke patients compared to healthy subjects, providing new theoretical insights into the interaction between brain activity and motor execution. These results underscore the disrupted corticomuscular coupling in stroke patients and suggest its value in characterizing the central-peripheral interaction during motor execution, which may inform future rehabilitation assessment strategies.Keywords: Stroke, Functional near-infrared spectroscopy, Surface electromyography, Corticomuscular coupling, Motor control Introduction Stroke is the leading cause of adult disability worldwide, with approximately 80% of survivors experiencing upper limb dysfunction. This impairment severely limits their ability to perform daily activities and significantly reduces quality of life, making motor recovery a central challenge in rehabilitation [1]. Such dysfunction arises not only from focal neurological damage, but also from the disruption of information transmission pathways between the cortex and the peripheral muscular system [2, 3]. Stroke patients often exhibit task-related cortical reorganization and altered coordination patterns between brain activity and muscular output [4]. Elucidating the functional characteristics of this central-peripheral pathway during motor execution is therefore essential for understanding the mechanisms underlying post-stroke neural impairment and for informing individualized rehabilitation strategies [5]. The human motor control system functions as a closed-loop network, wherein the central and peripheral nervous systems dynamically interact to achieve precise motor regulation [6]. Corticomuscular coupling (CMC) refers to the bidirectional communication between cortical motor areas and peripheral muscles, integrating top-down motor commands with bottom-up sensory feedback [7]. This coupling reflects the integrity of descending and ascending pathways that support voluntary movement [8]. This synchronization reflects the functional integrity of the corticospinal tract [9]. Following stroke, weakened corticomuscular coupling reflects a diminished capacity of the brain to effectively control muscle activity [10, 11]. This disruption compromises upper limb muscle coordination and contributes to motor dysfunction. Previous studies have demonstrated a significant reduction in CMC within the ipsilesional motor cortex of stroke survivors [12], which is closely associated with the degree of motor impairment. Therefore, CMC may serve as a valuable marker for investigating the reorganization of central-peripheral functional networks after stroke and throughout the recovery process [13], offering critical insights into the underlying mechanisms of motor deficits. Comparative studies involving healthy control groups have demonstrated in patients with acute or chronic stroke, that the normalization of CMC toward physiological levels is significantly associated with motor improvement [14, 15]. These findings highlight the potential of CMC as a critical biomarker for assessing the reorganization of central-peripheral functional networks following stroke. It may also reflect plastic changes and the restoration of functional integrity within the cortico-muscular pathway. Nevertheless, research on CMC during the subacute phase-a crucial period characterized by heightened neuroplasticity-remains limited, particularly in the context of active motor task engagement. This lack of evidence constrains a comprehensive understanding of the mechanisms underlying motor recovery during this pivotal stage. Although the analysis of CMC based on EEG is relatively well established, its application in post-stroke rehabilitation remains constrained by several technical limitations. EEG inherently suffers from limited spatial resolution and is highly susceptible to contamination from motion artifacts, particularly during active motor training or simultaneous multi-channel recordings [16]. In contrast, functional near-infrared spectroscopy (fNIRS), a non-invasive neuroimaging technique based on neurovascular coupling, offers higher spatial localization and greater resistance to motion artifacts [17, 18], making it particularly suitable for dynamic, continuous task monitoring during rehabilitation training. fNIRS enables the detection of detect task-related low-frequency activation at the cortical level and thus provides a valuable complement to EEG in studies of post-stroke cortico-muscular coupling, particularly in addressing limitations related to spatial resolution and movement compatibility [19, 20]. This study investigates corticomuscular coupling in subacute stroke patients during voluntary motor tasks, using simultaneous fNIRS and surface electromyography (sEMG) recordings. We hypothesize that, compared with healthy controls, stroke patients show reduced CMC strength in the ipsilesional motor cortex, reflecting impaired cortical output and altered sensorimotor integration. These changes may indicate early motor system reorganization and provide preliminary evidence of disrupted central-peripheral communication, offering insights into post-stroke motor impairment and rehabilitation targets. Methods Participants This cross-sectional observational study included 25 patients and 20 age-matched healthy subjects, all recruited from the Department of Rehabilitation Medicine, Zhongda Hospital (Table 1). All individuals were confirmed to be right-handed based on the Edinburgh Handedness Inventory [21] and exhibited sufficient communicative and cognitive abilities to comply with task requirements. The inclusion criteria of patients were: (1) age between 40 and 75 years; (2) first-ever unilateral stroke confirmed by neuroimaging; (3) stroke duration of 3 to 6 months with Brunnstrom stage III to V; (4) ability to extend the wrist from 80° flexion to 10° extension; (5) ability to understand simple commands and cooperate with treatment, and tolerance for 1 to 1.5 h of testing and assessment; (6) completion of the informed consent process in written form. The exclusion criteria of patients were: presence of intracranial mental implants, seizures, serious cardiovascular diseases, or infectious diseases, inability to tolerate the study, and participation in other clinical studies. Table 1Demographic and clinical baseline characteristics of the participantsStroke group (n = 25)Healthy group (n = 20)P valueAge51.41 ± 13.9145.50 ± 12.350.28Gender(male/female), n13/1212/80.76Stroke onset time, months3.85 ± 0.72N/AN/AStroke type (Ischaemic/Haemorrhagic)16/9N/AN/AStroke lesion (cortical/subcortical)8/17N/AN/AFMA-UE38.06 ± 12.01N/AN/AMAS-wrist extensionMedian (IQR)1.00(0.00, 1.00)N/AN/AMAS-wrist flexion Median (IQR)1.00(0.00, 1.50)N/AN/A Demographic and clinical baseline characteristics of the participants MAS-wrist extension Median (IQR) This study was conducted in accordance with the principles of the Declaration of Helsinki, and written informed consent was obtained from all participants prior to participation. Experimental design and procedure Experimental design and procedureBaseline demographic data were collected prior to the experiment. Before data collection, the operator explained the process to each participant until they fully understood it. The experimental setup in this study is shown in Fig. 1A. Each participant sat comfortably in a quiet, isolated room for the one-hour session. The subject’s forearm is fixed to the slot of the self-made armrest. Experimental design. A Experimental setup for the simultaneous acquisition of fNIRS and sEMG signals. B Probe layout of fNIRS channels, with sources marked by red dots and detectors by blue dots. A total of 16 sources and 16 detectors form 39 channels, covering six brain regions: LPFC, RPFC, LSMA, RSMA, LM1, and RM1. C Procedure protocol for fNIRs and sEMG data acquisition. Each trial included a 15-second task followed by a 25-second rest, repeated six times for each session First, the subjects were asked to perform three maximum autonomic contractions (MVC), each lasting 5 s, repeated three times with a 5-minute rest interval to avoid muscle fatigue. Second, the subjects completed wrist extension of 50% MVC equal length on the dominant side under visual feedback adjustment. Wrist extension plays a critical role in hand opening and object manipulation, both essential for activities of daily living (ADLs) and commonly impaired after stroke [22]. As a key movement reflecting motor control and functional recovery, we used this task to explore the mechanisms underlying motor control and rehabilitation. We selected 50% MVC because this moderate intensity is achievable for stroke patients and effectively induces β-band corticomuscular coherence [23]. It also avoids the cognitive load associated with low-intensity contractions and the muscle fatigue caused by high-intensity contractions, both of which could obscure true corticomuscular coupling results [24]. Each exercise consists of a 15-second task phase and a 25-second rest phase, repeated six times. The experimental process is shown in Fig. 1C. fNIRS and EMG were synchronized. fNIRS data acquisition and processing The 39-channel fNIRS system (NirSmart, Huichuang Medical Equipment Co. Ltd., Danyang, China) was used to cover the right and left prefrontal cortex, the right and left supplementary motor areas, and the right and left primary motor cortex. The near-infrared light is transmitted at 730 nm and 850 nm wavelengths with a sampling rate of 11 Hz. Figure 1B shows the fNIRS probe layout using the standard reference of the international 10–20 system. A source-detector separation of 30 mm was employed to enable signal acquisition from the cortical gray matter. Based on the spatial registration estimated on the MNI, a total of six regions of interest (ROI) were selected in both cerebral hemispheres. Probabilistic registration methods are used to map the channels for each ROI. In this study, we quantified changes in activity in the left (LPFC) and right (RPFC) prefrontal cortex (LPFC: CH 8, 9, 10, 11, 12, 24, 25, 26 and 27; RPFC: CH 3, 4, 5, 6, 7 19, 20, 21 and 23), the left (LSMA) and right (RSMA) supplementary motor area (LSMA: CH 28, 29, 35, 38 and 39; RSMA: CH 1, 15, 18, 31 and 32), and left (LM1) and right (RM1) primary motor cortex (LM1: CH 34 and 37; RM1: CH 30 and 33). fNIRS data were preprocessed using the Homer2 package. Raw signals were first converted into optical density values and corrects for motion artifacts induced by head movements. The motion-corrected data were subsequently processed with a bandpass filter ranging from 0.01 to 0.1 Hz to remove baseline drift and reduce physiological interferences such as cardiac and respiratory signals. Based on the modified Beer-Lambert law, the filtered optical density signals were transformed into relative concentration changes of oxygenated hemoglobin (HbO₂), deoxygenated hemoglobin (HbR), and total hemoglobin (HbT). To characterize the hemodynamic response function (HRF), we subtracted the average baseline value (calculated during the last 5 s of the resting phase) from the mean signal amplitude within the response plateau (5–15 s following motion onset). Considering that ΔHbO₂ generally provides a superior signal-to-noise ratio and greater sensitivity to regional cerebral blood flow changes, it was chosen as the primary metric for subsequent HRF analyses [25, 26]. sEMG data acquisition and processing To collect the sEMG signals, the skin over the target muscles of the right upper limb was first cleaned. The extensor carpi ulnaris (ECU) and flexor carpi radialis (FCR) muscles were selected for data collection using a wireless EMG telemetry system (Dino Medical Equipment Co., Ltd., Shanghai, China). Ag-AgCl bipolar electrodes were placed at the muscle bellies of each target muscle in accordance with anatomical landmarks and guidelines recommended by SENIAM [27], maintaining an inter-electrode distance of 20 mm. The sEMG signals were recorded using a high-quality amplifier, and the signals were sampled at a frequency of 2,000 Hz via a data converter. To ensure synchronization, the sEMG and fNIRS devices were synchronized and marked using the E-Prime 2.0 software. All sEMG data were processed using MATLAB (2022b, MathWorks, USA). Raw electromyographic signals from the ECU and FCR muscles were first subjected to a fourth-order Butterworth bandpass filter with a frequency range of 20 to 450 Hz. A 10-second stationary segment from each test cycle was manually selected for further analysis. After full-wave rectification, the signals were segmented into 100ms windows, and the root mean square (RMS) was calculated within each window. To standardize the measurements, RMS values were normalized to the maximum voluntary contraction (MVC). Primary outcome The primary outcome of this study was corticomuscular coupling during wrist extension, which was assessed by calculating two indices: the Phase Synchronization Index (PSI) and coherence between fNIRS and sEMG signals. fNIRS data were used to derive relative changes in oxygenated hemoglobin concentration (ΔHbO₂) from six cortical regions of interest. sEMG data were recorded from the ECU and FCR muscles, and RMS values were used in the coupling analysis. Both time-domain and frequency-domain analyses were conducted to calculate the PSI and coherence values, thereby quantifying the strength and characteristics of the corticomuscular coupling. Phase Synchronization Index (PSI) The Hilbert transform was applied separately to the fNIRS (ΔHbO₂) and sEMG signals to extract their instantaneous phase angles [28]. This transforms the real-valued time-series signals into complex signals, allowing us to compute the instantaneous phase at each time point. The phase difference between the two signals was calculated according to the formula: where and are the instantaneous phases of the fNIRS and sEMG signals at time t, respectively, and n and m are scaling factors (both set to 1 in this analysis). The phase synchronization index was then computed using the following equation: where and represent the time-averaged cosine and sine of the phase difference, respectively. This averaging is performed over a sliding window of reasonable duration to assess synchronization over specific intervals. The resulting PSI quantifies the degree of phase coupling between the cortical activity and muscular activity, with values approaching 1 indicating a high level of synchronization and values closer to 0 indicating low synchronization. Coherence value The short-time Fourier transform (STFT) was applied to both fNIRS and sEMG signals to extract their spectral characteristics. The cross-spectral density between the two signals and the auto-spectral densities of each signal were computed [29, 30]. Coherence (Cxy) was calculated as follows: where Pxy(f) is the cross-spectral density between the fNIRS and sEMG signals, and Pxx(f) and Pyy(f) are the auto-spectral densities of fNIRS and sEMG signals, respectively. Coherence values reflect the frequency-domain coupling between cortical and muscular activities, with values approaching 1 indicating strong synchrony. Both PSI and coherence analyses were used to characterize central-peripheral coupling during the task, with PSI capturing phase-based synchronization and coherence assessing the strength of interaction across different frequencies. Secondary outcome fNIRS measures (cortical activation, functional connectivity) Cortical activation was assessed as the mean ΔHbO₂ amplitude in six cortical ROIs during wrist extension. Intra-cortical functional connectivity was evaluated using Pearson correlation coefficients between ΔHbO₂ time series across ROI pairs. sEMG parameters (RMS, MF, and fApEn) Root mean square (RMS) amplitude and median frequency (MF) were calculated to provide a comprehensive evaluation of neuromuscular activity. RMS value reflects the energy content of the EMG signal and serves as an indicator of the level of muscle activation during contraction [31]. MF reflects the spectral characteristics of the EMG signal and is related to muscle fiber composition and neural conduction velocity. Fuzzy approximate entropy (fApEn) was applied to the EMG signals to quantify signal irregularity [32]. In this study, fApEn-based nonlinear analysis was employed to assess the complexity of segmented sEMG signals under different conditions. Higher fApEn indicates greater signal complexity (i.e., less regularity), often associated with diverse motor unit recruitment. Lower fApEn suggests reduced complexity (i.e., higher regularity), which may reflect impaired neuromuscular control. Based on previous studies [33, 34], we selected the following parameters for fApEn computation: N = 5000, r = 0.15 × SD, n = 2, and m = 2. The calculation procedure for fApEn was based on previous studies [4] and is summarized as follows: where the function φm is defined as: Statistical analysis Data analyses were conducted using MATLAB (2022b, MathWorks, USA) and GraphPad Prism 8.0 (GraphPad Software, USA). The Shapiro-Wilk test was employed to assess the normality of the data. The homogeneity of demographic variables between the two groups was evaluated using the chi-square test and independent samples t-test. Independent samples t-tests were used to compare cortical activation data, functional connectivity data, and corticomuscular coupling between two groups. A p-value of < 0.05 was considered statistically significant. Where appropriate, the false discovery rate (FDR) correction was applied to control for multiple comparisons. Results Demographic and clinical characteristics A total of 25 stroke patients and 20 age-matched healthy controls were enrolled in this study. There were no significant differences between the two groups in terms of age (stroke group: 51.41 ± 13.91 years; control group: 45.50 ± 12.35 years; p = 0.98) or sex distribution (P = 0.76). In the stroke group, the mean disease duration was 3.85 ± 0.72 months. Detailed demographic and clinical characteristics of both groups are presented in Table 1. Corticomuscular coupling (Coherence and PSI) During the isometric wrist extension, PSI and coherence values for healthy subjects and stroke patients are presented in Fig. 2. The coherence values for LM1-ECU (P = 0.017, Cohen’s d = 1.02), LM1-FCR (P = 0.014, Cohen’s d = 1.05), LPFC-FCR (P = 0.023, Cohen’s d = 0.57), LSMA-ECU (P = 0.0007, Cohen’s d = 0.95), and LSMA-FCR (P = 0.016, Cohen’s d = 1.03) in stroke patients were significantly lower than those in healthy subjects. Similarly, the PSI for LM1-ECU (P = 0.016, Cohen’s d = 1.03), LM1-FCR (P = 0.023, Cohen’s d = 0.97), and LSMA-ECU (P = 0.035, Cohen’s d = 0.89) in stroke patients was significantly lower compared to healthy subjects. Comparison of coherence values (A) and PSI (B) between ROI and ECU, FCR muscles during the task between stroke patients and healthy group. The red indicates significance values (P < 0.05) fNIRS measures (cortical activation and functional connectivity) The average ΔHbO2 values of stroke patients and healthy subjects during isometric wrist extension are shown in Fig. 3. As shown in Fig. 3, the brain activation levels in the stroke group were generally higher than those in the healthy controls across all six ROIs. Figure 4 provides a statistical analysis of these differences. The activation of RSMA (P = 0.003, Cohen’s d = 0.93), LSMA (P = 0.041, r = 0.28), RM1(P = 0.007, Cohen’s d = 0.87), and RPFC (P = 0.019, r = 0.33) in stroke patients was higher than that in healthy subjects. Changes in oxyhemoglobin concentration (ΔHbO2) in healthy individuals (A) and stroke patients (B) Cortical activation differences between stroke patients and healthy individuals during the task. (*P < 0.05, **P < 0.01, L: left upper extremity; R: right upper extremity; PFC: prefrontal cortex; SMA: supplementary motor area; M1: primary motor cortex.) Figure 5A and B show the connection maps of brain regions for healthy subjects and stroke patients, respectively, during isometric wrist extension. These figures visually depict the functional connectivity patterns between regions of interest (ROIs) in the brain, highlighting distinct connectivity profiles for each group. Figure 5C presents the significant differences in functional connections among ROI regions during the extension task period between stroke patients and healthy subjects. Specifically, the connectivity of RSMA ~ RPFC (FDR-corrected P = 0.0064, r = 0.46), RSMA ~ LPFC (FDR-corrected P = 0.016, r = 0.54), LPFC ~ LM1 (FDR-corrected P = 0.042, Cohen’s d = 0.41), RM1 ~ RPFC (FDR-corrected P = 0.023, r = 0.37), and LM1 ~ RPFC (FDR-corrected P = 0.0039, r = 0.40) were stronger in stroke patients than in healthy subjects. Overall, stroke patients had higher FC than healthy subjects. Cortical ΔHbO₂ correlation analysis among regions of interest (ROIs) in healthy controls (A) and stroke patients (B). The color of each line indicates the strength of functional connectivity between ROI pairs. C Highlights significant differences in inter-regional connectivity during the wrist extension task between the two groups sEMG parameters (RMS, MF, and fApEn) Figure 6 illustrates the RMS, MF, and fApEn values of EMG signals recorded from the FCR and ECU muscles during the wrist extension phase. The RMS amplitude of the ECU in the stroke group was significantly lower than that in the healthy control group (P = 0.0003, r = 0.81). The MF of the FCR was also significantly reduced in stroke patients compared to controls (P = 0.0011, Cohen’s d = 1.39). In addition, the MF of the ECU was significantly lower in the patient group than in the healthy group (P = 0.0002, Cohen’s d = 1.67). Comparison of EMG signal characteristics between stroke patients and healthy subjects: (A) RMS values of FCR and ECU; (B) MF values of FCR and ECU; (C) fApEn values of FCR and ECU. Significant differences between conditions are indicated by *P < 0.05, **P < 0.01 Based on previous studies, a window size (N) of 5000 and a tolerance (r) of 0.15 were set, and the mean value of six repetitions was calculated. The results revealed that, under the 50% MVC condition, the fApEn values of both the FCR (P = 0.0019, Cohen’s d = 1.66) and ECU (P = 0.023, Cohen’s d = 1.06) were significantly higher in healthy subjects than in stroke patients. Discussion This study integrated fNIRS and sEMG to investigate the characteristics of corticomuscular coupling between the motor cortex and forearm muscles during an isometric wrist extension task in stroke patients. The results revealed that, compared with healthy controls, stroke patients exhibited significantly reduced coherence and PSI between cortical regions of the ipsilesional hemisphere and the muscles on the affected side, suggesting impaired integrity of the corticospinal descending pathways. In addition, during task execution, stroke patients showed enhanced cortical activation in both the prefrontal and motor areas, accompanied by markedly increased functional connectivity. However, the overall activation level and complexity of electromyographic signals on the paretic side-as masured by RMS, MF, and fApEn, were significantly lower than those observed in healthy individuals. This dissociation, characterized by heightened cortical activation and connectivity alongside diminished corticomuscular coupling, reflects a disruption in central-peripheral coupling following stroke. Our findings are generally consistent with previous studies on corticomuscular coupling in stroke patients. Several EEG-sEMG studies have demonstrated disrupted corticomuscular coupling following stroke. For example, Fang et al. reported significantly reduced β-band coherence between the deltoid and brachialis muscles in patients with chronic stroke during shoulder flexion and elbow extension tasks, suggesting impaired central-peripheral communication [35]. Similarly, Xu et al. observed reduced EEG-EMG coherence in the tibialis anterior muscle of stroke survivor, further supporting the notion of disrupted neuromuscular interaction [10]. Additionally, a previous study [36] using fNIRS-sEMG revealed reduced muscle signal complexity and altered brain-muscle connectivity during elbow flexion and extension in individuals with chronic stroke, highlighting the potential of such coupling abnormalities as biomarkers of upper limb motor dysfunction. These findings collectively indicate reduced transmission efficiency and impaired connectivity between lesioned cortical regions and target muscles, possibly due to partial disruption of corticospinal neurons and altered excitability of local interneurons [37]. However, these studies have primarily focused on chronic phases of stroke while employing a limited repertoire of CMC measures. While existing studies have provided a valuable foundation for understanding CMC in chronic stroke, research focusing on the subacute stage remains limited. By focusing on the subacute stage and integrating both coherence and phase synchronization analysis, our study seeks to offer a more comprehensive perspective on corticomuscular reorganization during this critical period of motor recovery. Our findings revealed that, compared with healthy controls, patients in the subacute stage of stroke exhibited significantly increased activation in RSMA and RPFC. Additionally, during the motor task, stroke patients showed stronger functional connectivity between RSMA, RM1, and RPFC regions. These results align with previous studies reporting increased contralesional cortical activity during stroke recovery [38, 39], supporting the concept of compensatory reorganization [40]. sEMG results showed that RMS, MF, and fApEn values in the paretic upper limb of subacute stroke patients were significantly lower than those in healthy subjects. Specifically, the fApEn value of the ECU muscle during wrist extension was notably reduced, indicating impaired motor unit recruitment and decreased activation complexity. This reduction in EMG complexity reflects a loss of flexibility in muscle activation patterns, suggesting compromised central motor command transmission [33]. These findings align with evidence that decreased EMG complexity in stroke patients is linked to deficits in central motor control [41]. Overall, the lower RMS, MF, and fApEn values point to impaired motor unit recruitment and reduced motor output complexity, key indicators of disrupted corticomuscular control after stroke. Taken together, our study provides physiological evidence for the “high activation-low coupling” phenomenon, as evidenced by the concurrent observations of enhanced cortical activation (significantly increased HbO concentration in the contralesional SMA and PFC measured by fNIRS) and diminished corticomuscular coupling (significantly reduced coherence and PSI) in stroke patients. Specifically, our findings indicate that although patients recruited broader cortical resources (including the contralesional prefrontal and motor areas) presumably to compensate for the impairment, this enhanced cortical output, which may involve alternative pathways, failed to be effectively translated into functional coupling with the paretic-side muscles [42]. This pattern is consistent with previous neuroimaging studies showing that stroke survivors often recruit broader or more intense cortical activation than healthy individuals to accomplish similar tasks [43]-a phenomenon interpreted as compensatory reorganization through alternative cortical areas. However, enhanced cortical output does not necessarily translate into effective cortical-muscular connectivity [15]. One plausible explanation is that structural damage to the cortex and corticospinal tract compromises the transmission of descending motor commands. As a result, greater cortical effort may be required to drive peripheral muscle activity. However, reduced conduction efficiency impairs corresponding increases in functional coupling [44]. This “high activation vs. low coupling” phenomenon suggests that the structurally compromised motor system has difficulty coordinating efficient motor execution [9]. Furthermore, compensatory recruitment of alternative motor pathways-such as the contralesional cortex or subcortical tracts (e.g., reticulospinal pathways)-may contribute to motor output [45]. However, due to their distinct dynamic properties, these circuits may fail to synchronize effectively with cortical oscillations, thereby further weakening corticomuscular integration. The simultaneous fNIRS-sEMG technique employed in this study offers a powerful complement to the traditional EEG-sEMG approach for probing CMC. Firstly, the fNIRS-sEMG coupling analysis employed in this study possesses a temporal resolution constrained by the slow nature of the hemodynamic response (typically on the order of seconds), which fundamentally differs from EEG that directly captures millisecond-scale electrophysiological activity [46]. This implies that what we capture is not instantaneous neural driving at the millisecond level, but rather the integration and modulation of the overall excitability level of the cortical motor network and the recruitment of the spinal motor neuron pool over a longer time scale, potentially encompassing processes of motor preparation, execution, and maintenance. Therefore, our fNIRS-based CMC metrics and conventional EEG-based high-frequency rhythm synchronization metrics elucidate the integrity of corticomuscular functional connectivity from distinct temporal perspectives, offering complementary insights [47]. Secondly, in terms of frequency and physiological basis, EEG-EMG coherence primarily captures the millisecond-scale relationships between cortical neural ensemble oscillations (such as β rhythms) and motor unit discharge sequences, directly correlating with the instantaneous neural drive of rhythmic movement. In contrast, the physiological rationale for coupling low-frequency fNIRS hemodynamics with high-frequency sEMG signals lies in their joint representation of the complete physiological pathway from motor command generation to muscle execution. Specifically, fNIRS indirectly detects enhanced neural ensemble activity in the motor cortex via “neurovascular coupling” [48]. Subsequently, descending motor commands are transmitted through the corticospinal tract, leading to the excitation of spinal motor neurons and the recruitment of motor units, which is ultimately manifested as the electrical muscle activity captured by sEMG. Crucially, the sEMG signal itself is not devoid of low-frequency components. Research has identified oscillations below 0.5 Hz in sEMG that are coherent with oscillations in force output [49], potentially related to the synchronized recruitment of motor units [50]. Therefore, the slow hemodynamic signals recorded by fNIRS and the low-frequency oscillatory components within sEMG may share a common, slowly varying modulatory source at the level of neural drive. This viewpoint is supported by existing evidence, for instance, low-frequency fNIRS-sEMG coupling has been validated as an effective neurophysiological indicator in patients with spinal cord injury [26]. This study has some limitations. The cross-sectional design and modest sample size limit the ability to track CMC changes over time or infer causal relationships with motor recovery. Moreover, all patients were in the subacute phase with Brunnstrom stage III-IV, which may limit the generalizability of the findings to other stages of stroke recovery. Future studies should include patients across different recovery stages and employ longitudinal, multimodal approaches to better characterize CMC dynamics and their relevance to motor outcomes. Our findings carry significant potential implications for stroke rehabilitation. CMC metrics hold promise for evolution into objective biomarkers for personalized rehabilitation assessment. They offer an objective and quantitative perspective on the functional integrity of motor pathways, complementing existing behavioral scales. In addition, CMC patterns may help differentiate neurofunctional recovery subtypes; for example, patients who consistently exhibit reduced CMC may respond better to interventions aimed at improving central-peripheral conduction efficiency. Finally, CMC could also be incorporated into closed-loop rehabilitation systems, where real-time CMC strength serves as a feedback signal to adjust task difficulty or to trigger neuromodulatory stimulation. Conclusion This cross-sectional study investigated corticomuscular coupling during isometric wrist extension in subacute stroke patients using simultaneous fNIRS-sEMG recording. Compared to healthy controls, stroke patients showed reduced corticomuscular coupling and lower neuromuscular signal complexity, alongside increased activation in contralesional cortical regions. These findings indicate altered central-peripheral interactions following stroke. Further studies are needed to clarify the functional relevance of these coupling metrics in the context of motor recovery. Acknowledgements The authors would like to thank all the participants involved in this study, as well as the Department of Rehabilitation Medicine at Zhongda Hospital for their support. Authors’ contributions XPW: Methodology, software, data collection and organization, visualization, writing of the original draft; XSW, ZL: Data collection, Methodology, data analysis; SJ: Conceptualization, patient training, manuscript review and editing; HXW: Conceptualization, writing review and editing, project management, funding acquisition, supervision. Funding This project is funded by the Zhongda Hospital Affiliated to Southeast University, Jiangsu Province High-Level Hospital Construction Funds (GSP-LCYJFH09). This paper was supported also by the following grants: (1)Key Project of Jiangsu Province's Key Research and Development Program (No.BE2023023-4); (2)Key Project of Jiangsu Province's Key Research and Development Program (BE2023034). Data availability The data supporting the findings of this study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The study was approved by the Ethics Committee of Zhongda Hospital, Southeast University (Approval No. 2024ZDSYLL284-P01). This study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to participation in the study. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data supporting the findings of this study are available from the corresponding author upon reasonable request.