Fifteen healthy right-handed volunteers (7 males; mean age 27.13 ± 4.52 years) participated in this exploratory crossover study from August 2020 to January 2021. All participants were free of neurological, psychiatric, or upper limb sensorimotor disorders and had no contraindications to TMS. The study protocol was approved by the Institutional Review Board of Huashan Hospital, Fudan University (reference number: #2019-609), and written informed consent was obtained from all participants prior to enrolment.

The study was conducted in a separate and quiet room. Each participant completed two experimental sessions in a randomized order: one involving NMES and the other PMS applied to the right wrist extensors. Sessions were separated by a washout period of at least 48 h to minimize carryover effects. The experimental timeline is summarized in Figure 1.

At the beginning of each session, a 5-min resting-state fNIRS recording was acquired. Within each session, the following sequence was performed: (1) a pre-intervention assessment of corticospinal excitability using TMS; (2) the peripheral stimulation intervention (NMES or PMS), during which continuous fNIRS data were recorded; and (3) an immediate post-intervention TMS assessment. Participants remained seated comfortably with their eyes open and were instructed to stay relaxed throughout. Subjective comfort was rated after each stimulation session.

NMES was delivered using a biphasic constant-current stimulator (ES-521, ITO Co., Ltd, Tokyo, Japan) via a single channel. Stimulating electrodes were placed distal to the common extensor origin and halfway down the extensor surface of the right forearm, covering both extensor carpi ulnar and extensor carpi radialis muscles. The protocol was conducted at a frequency of 50 Hz with 15 repetitions of 10 s on/30 s off stimulation for 10 min, ramp-up and ramp-down taking 1 s. The current intensity was individually adjusted to elicit maximal, visually confirmed wrist dorsiflexion that was comfortable and pain-free (mean± SD: 12.30 ± 3.78 mA).

PMS was applied using a monophasic magnetic stimulator (OSF-pTMS, O.SELF Company, Wuhan, China) with a figure-of-eight coil (outer diameter 70 mm). The coil was held tangentially over the muscle belly of the right forearm extensors to induce maximal wrist dorsiflexion. The stimulation protocol matched the NMES timing structure: 10 Hz pulses delivered in 15 blocks of 10 s on/30 s off, yielding a total of 1,500 pulses over 10 min. Intensity was titrated from a low baseline (e.g., 20% of maximum stimulator output, %MSO) upward to the lowest level that consistently produced a maximal, painless wrist dorsiflexion, with its amplitude visually matched to that achieved during NMES (mean ± SD: 30.47 ± 4.78 %MSO). This titration to the lowest effective intensity mitigates any risk of overstimulation, which is pertinent given the absence of cutaneous discomfort with PMS.

Corticomotor excitability was assessed using single-pulse TMS delivered with an OSF-pTMS magnetic stimulator (O.SELF Company, Wuhan, China) with a figure-of-eight coil. Surface electromyography was recorded from the right first dorsal interosseous (FDI) muscle using Ag/AgCl electrodes placed in a belly–tendon montage. The raw EMG signal was amplified, band-pass filtered (20–1,000 Hz), and sampled at 5,000 Hz for offline analysis. The coil was placed tangentially over the left M1 at a 45° angle to the midline to induce a posterior-anterior current. The hotspot for eliciting Motor evoked potential (MEP) in the FDI was identified by systematically moving the coil in 5 mm increments and was marked on a tight-fitting electrocap to ensure consistent coil placement across pre- and post-intervention measurements.

Resting motor threshold (RMT) was defined as the minimum intensity required to produce MEPs with a peak-to-peak amplitude >50 μV in at least 5 out of 10 consecutive trials in the relaxed FDI muscle. To quantify corticospinal excitability, MEP amplitude was assessed at a fixed suprathreshold stimulus intensity. This intensity was individually determined at baseline as the level required to evoke an approximately 1 mV peak-to-peak MEP amplitude. The same stimulus intensity was then maintained for both pre- and post-intervention assessments to ensure a stable and sensitive measure of changes in corticospinal excitability. At each time point, 10 single-pulse TMS stimuli were delivered with an inter-stimulus interval of at least 5 s.

fNIRS data were preprocessed and analyzed using HomER2 (version 2.8), a MATLAB-based graphical interface for processing continuous-wave NIRS data (Huppert et al., 2009). The raw optical intensity signals were processed through a standardized pipeline. First, the hmrIntensity2OD function was applied to convert the raw intensity signals into optical density (OD). Motion artifacts were then identified using the HmrMotionArtifactByChannel algorithm with thresholds set at STDev = 10 and AMP = 5, and corrected via spline interpolation. A bandpass filter (0.01–0.1 Hz) was applied to remove low-frequency drift and high-frequency physiological noise. Subsequently, the hmrOD2Conc function was employed to convert OD data into concentration changes of oxygenated ([HbO]) and deoxygenated ([HbR]) based on the modified Beer–Lambert law. For task-based analysis, trial-wise concentration changes of [HbO] and [HbR] were baseline-corrected by subtracting the mean signal from the 5 s interval immediately preceding each stimulation block. Hemodynamic responses were then time-locked to stimulus onset and block-averaged across all trials within each condition using the hmrBlockAvg function, yielding mean response time courses from −5 s to 40 s relative to stimulus onset.

For subsequent general linear model (GLM) and functional network analyses, data were processed within the NirSpark toolbox (NirSmart, Danyang Huichuang Medical Equipment, China) in the MATLAB environment. Preprocessing in NirSpark applied parameters consistent with the HomER2 pipeline (motion correction: STDev-thresh = 10, AMP-thresh = 5; bandpass filter: 0.01–0.1 Hz; partial pathlength factor = 6). The GLM was implemented to estimate condition-specific [HbO] responses at both individual subject and group levels, yielding beta values for each condition. A 4 s full-width-at-half-maximum Gaussian smoothing kernel was applied to the [HbO] time series to suppress short-duration, high-frequency noise.

Functional connectivity was assessed by constructing brain networks based on inter-nodal correlation. For the 64-channel configuration, Pearson correlation coefficients were computed between all channel pairs. Network edges were defined by applying a series of similarity thresholds (r ≥ 0.5, 0.6, 0.7, and 0.8). ROI-based and channel-based connectivity matrices were generated for both NMES and PMS conditions during stimulation. Differences in connectivity strength (ROI-ROI and channel-channel) between the two stimulation conditions were statistically compared using paired t-tests, performed separately for the resting-state and task-period data.

Based on standard Montreal Neurological Institute (MNI) spatial coordinates corresponding to the fNIRS source-detector array, channels were grouped into six anatomical ROIs: Left PFC: channels 6, 7, 8, 9, 11, 13; Right PFC: channels 1, 3, 4, 5, 14, 15; Left SMC: channels 20, 22, 23, 24, 25, 26, 35, 36; Right SMC: channels 17, 18, 27, 28, 29, 30, 31, 33; Left OC: channels 51, 54, 56, 61, 62, 63; Right OC: channels 43, 45, 47, 49, 50, 58.

Given the study's focus on cortical responses to peripheral wrist extensor stimulation, channels located over the primary motor representation of the forearm were selected as seeds for detailed analysis. According to MNI coordinates, channels 23 and 25 corresponded to the left forearm motor cortex, and channels 18 and 30 to the right forearm motor cortex.

Statistical analyses were performed using IBM SPSS Statistics (version 22.0). All continuous data are expressed as mean ± standard deviation, with normality confirmed by the Shapiro-Wilk test. Corticospinal excitability was quantified using raw MEP amplitudes and, where appropriate, expressed as percentage change relative to the pre-intervention value (MEP% of baseline) to facilitate within-subject comparisons across stimulation conditions. Additionally, paired t-tests were conducted to compare MEP% of baseline values between the NMES and PMS conditions. For fNIRS data, mean concentrations of [HbO] and [HbR] were calculated for the stimulation period (5–10 s post-stimulus onset) and the subsequent rest period (35–40 s post-stimulus onset). Regional activity within each ROI was derived by averaging signals across all constituent channels. Paired t-tests were used to compare [HbO] and [HbR] between stimulation and rest periods within each ROI and condition. Direct comparisons between NMES and PMS were performed for stimulation-period [HbO] and [HbR] values within SMC channels. Pearson's correlation analyses were conducted to examine associations between MEP changes and stimulation intensity, between MEP changes and RMT, and between MEP% of baseline values across the two stimulation conditions. To account for multiple comparisons, the Benjamini–Hochberg procedure was applied to control the false discovery rate. Statistical significance was set at a two-tailed p value < 0.05.

RMT measured prior to stimulation did not differ significantly between the NMES (37.73 ± 11.81 %MSO) and PMS (36.60 ± 12.94 %MSO) sessions (p = 0.697). A significant positive correlation was observed between individual RMT measured in the two sessions (r = 0.656, p = 0.008), indicating good within-subject consistency (Figure 3B).

MEP amplitudes recorded before and after each intervention are presented in Figure 3A. No significant change in MEP amplitude was observed NMES (p = 0.674) or PMS (p = 0.794). Consistent with these findings, the relative change in MEP amplitude did not differ significantly between the two stimulation conditions.

Resting-state functional connectivity analysis revealed no significant differences in either ROI-ROI or channel-channel connectivity strength between the NMES and PMS conditions prior to stimulation, indicating comparable baseline network organization across sessions. Task-evoked hemodynamic responses are in Figure 4.

During NMES, a significant reduction in [HbO] was observed in the PFC, right SMC, and OC during the stimulation period compared with rest (all p < 0.05; Figure 4A). In contrast, channels 23 and 35 within the left motor cortex exhibited a non-significant increase in [HbO]. Concurrently, [HbR] levels was significantly reduced in selected channels within the right PFC (ch4: p = 0.020; ch15: p = 0.013) and bilateral SMC (left ch35: p = 0.010, right ch31: p = 0.001).

During PMS, [HbO] in the left SMC demonstrated an increasing trend, whereas decreases were observed in the OC and right SMC. However, none of these [HbO] changes survived correction for multiple comparisons (p > 0.05; Figure 4B). In contrast, [HbR] exhibited significant reductions in multiple SMC channels during stimulation relative to rest, indicating a more spatially localized but consistent deoxygenation response under PMS.

Direct comparison between stimulation modalities revealed distinct hemodynamic patterns within the SMC channels. Specifically, [HbO] concentrations in the left motor cortex (channels 23 and 35) were significantly higher during PMS than during NMES (ch23: p = 0.005; ch35: p = 0.022). In parallel, [HbR] levels were significantly lower under PMS at several SMC channels (ch23: p = 0.007; ch26: p = 0.012; ch35: p = 0.039), further highlighting modality-specific differences in cortical oxygenation dynamics (Figures 4C–F). GLM-derived [HbO] activation maps are presented in Figure 5. Both NMES and PMS elicited a lateralized activation pattern within the SMC, characterized by positive activation in the left motor cortex accompanied by relative deactivation in the right hemisphere. Notably, the magnitude of [HbO] activation in the left motor cortex was visibly greater during PMS than during NMES, indicating a stronger task-related hemodynamic response under PMS.

Task-related functional connectivity analysis revealed no significant differences between the NMES and PMS conditions. Specifically, neither the overall strength of ROI-ROI connections nor the number of significant channel-channel connections differed between stimulation modalities across similarity thresholds ranging from 0.5 to 0.8 (Figure 6).

No significant changes in MEP amplitude were observed following either a single session of NMES or PMS. This finding should be interpreted within the context of the specific stimulation parameters employed. Previous studies have established a clear dose–response relationship for NMES-induced cortical excitability changes. Sasaki et al. (2017) reported that 20 min of median nerve NMES at 30 Hz significantly increased MEP amplitude when delivered above MT, whereas stimulation at 90% of MT was ineffective. Similarly, Huang et al. (2019) demonstrated that a significant increase in cortical [HbO], measured by fNIRS, occurred only when NMES current intensity exceeded 20 mA, showing a clear dose-dependent pattern. Furthermore, stimulation duration is critical, as NMES protocols capable of inducing sustained facilitatory effects typically require intervention durations exceeding 20 min (Andrews et al., 2013; Wu et al., 2005). In comparison, the intensity and duration of NMES used in the present study were relatively conservative and may have been insufficient to elicit detectable after-effects.

Similarly, the effects of PMS on cortical excitability are highly parameter-dependent (Gallasch et al., 2015; Nito et al., 2021). Evidence indicates that rPMS delivered at frequencies ≥25 Hz for at least 15 min is effective in enhancing M1 excitability and improving motor performance, whereas 10 Hz rPMS failed to induce changes in cortical excitability (Nito et al., 2021). Moreover, 25 Hz rPMS has been shown to be superior to 10 Hz in inducing long-term potentiation–like plasticity within the SMC (Gallasch et al., 2015) and demonstrates greater clinical efficacy in improving upper limb motor function and reducing hand edema post-stroke (Jiang et al., 2022; Jia et al., 2021; Fujimura et al., 2025). In contrast, inducing measurable after-effects with 10 Hz PMS typically requires significantly prolonged intervention durations, with studies suggesting the need for continuous stimulation of at least 45 min (McKay et al., 2002) or cumulative application times exceeding 2 h (Ridding et al., 2000; Kaelin-Lang et al., 2002). Therefore, the short-duration 10 Hz PMS protocol used in the present study likely did not reach an effective stimulation dose.

Although stimulation was applied to the wrist extensors, MEPs were recorded from the FDI muscle to assess excitability changes across the broader corticospinal system rather than localized effects restricted to the target muscle. Previous studies indicate that peripheral stimulation can induce heterotopic modulation of corticospinal excitability via intracortical network mechanisms within M1 (Yamamoto et al., 2022). While this strategy is suitable for capturing system-level modulation, it may reduce sensitivity to highly focal effects. Thus, the absence of MEP changes in the FDI does not preclude the possibility of more localized excitability changes within the cortical representation of the stimulated muscle.

Despite the absence of lasting excitability changes, the real-time hemodynamic responses captured by fNIRS suggest that the two stimulation modalities engage sensorimotor networks in distinct, modality-specific ways during the intervention.

Under matched peripheral motor output, PMS primarily elicited focal activation in the contralateral SMC, characterized by increased [HbO] and decreased [HbR], representing a typical pattern of enhanced local neurovascular coupling. This focused activation pattern aligns with the physical properties of PMS and its selective recruitment mechanism for deep proprioceptive afferents. Unlike NMES, the time-varying magnetic field generated by PMS penetrates high-resistance skin tissue, preferentially activating deep Ia-class proprioceptive afferent fibers within the muscle while relatively sparing cutaneous receptors (Beaulieu et al., 2017; Lampropoulou et al., 2012). This relatively “pure” proprioceptive input (Beaulieu et al., 2017) is transmitted primarily via the dorsal column and spinothalamic tracts to the contralateral SMC (Struppler et al., 2004; Okudera et al., 2015; Struppler et al., 2007) and can rapidly modulate excitability in this region within approximately 1 s (Sato et al., 2016). The focal activation pattern observed in the present study may therefore reflect the rapid and efficient recruitment of sensorimotor circuits by PMS.

In contrast, NMES elicited only limited activation in the contralateral SMC, accompanied by more widespread decreases in [HbO] across the PFC, ipsilateral SMC, and occipital regions. This hemodynamic pattern should not be simplistically interpreted as cortical deactivation but rather likely reflects the non-selective nature of transcutaneous electrical stimulation. NMES concurrently activates motor axons, low-threshold tactile fibers (Aβ), and smaller-diameter afferent fibers (Aδ and C; Szecsi et al., 2010), resulting in a complex afferent volley integrating proprioceptive, tactile, and potentially nociceptive signals. Within this context, the widespread recruitment of cutaneous afferents may introduce a degree of “non-physiological noise” that competes with task-relevant proprioceptive input derived from muscle contraction, thereby attenuating effective drive to the sensorimotor cortex (Beaulieu et al., 2017). Furthermore, mild nociceptive input may further suppress M1 excitability by engaging pain-modulatory circuits (Carson and Buick, 2019; Tashiro et al., 2019). These mechanisms may collectively account for the relatively constrained activation observed in the contralateral SMC during NMES in this study.

The signal changes observed in the PFC likely reflect the additional cognitive and integrative load required to process this multimodal sensory information (Muthalib et al., 2015). NMES concurrently engages multiple regions including S1, M1, premotor areas, and the PFC (Carson and Buick, 2019). Therefore, the net effect of NMES on cortical activity likely represents a dynamic balance between excitatory proprioceptive drive and inhibitory influences stemming from cutaneous and nociceptive inputs (Struppler et al., 2007; Beaulieu et al., 2017). an interpretation consistent with prior network-level findings. The widespread [HbO] decreases observed herein may thus reflect the hemodynamic signature of large-scale cortical network reconfiguration or inhibitory regulation in response to processing NMES's complex, mixed afferent input.

In summary, the different hemodynamic patterns induced by NMES and PMS in this study are more likely to originate from fundamental differences in the composition of the afferent volley and the characteristics of network recruitment, rather than from differences in stimulation intensity per se.

No significant differences in task-based functional connectivity were found between the NMES and PMS conditions. This null finding may be attributed to the relatively short stimulation duration, limited sample size, and the sensitivity constraints of fNIRS-based connectivity analysis in healthy populations. More subtle or distributed network changes might require longer stimulation periods or repeated interventions to become detectable.

Notably, the dissociation between TMS and fNIRS findings underscores the complementary value of a multimodal assessment approach. While TMS-evoked MEPs are considered the gold standard for assessing net corticospinal output excitability, they may lack sensitivity to transient, intra-intervention dynamic cortical processing. In contrast, fNIRS captures real-time hemodynamic changes occurring during stimulation (Huang et al., 2019). The present results suggest that the modality-specific cortical engagement patterns characteristic of each peripheral stimulation technique may precede or occur independently of detectable changes in corticospinal excitability.

This study has several limitations. First, the modest sample size, inherent to its pilot and exploratory nature, limits the statistical power to detect small-to-medium effect sizes and increases the risk of Type II errors. Therefore, caution is warranted when interpreting the negative findings. Second, the single-session design precludes conclusions regarding cumulative or long-term neuromodulatory effects, which are more relevant to clinical rehabilitation applications. Third, fNIRS channel localization was based on standardized anatomical coordinates rather than individual MRI data, which may compromise spatial precision. Finally, the limited penetration depth of fNIRS restricts the assessment of deeper cortical and subcortical structures that may be involved in peripheral stimulation-induced neuromodulation.

In summary, a single 10-min session of low-intensity NMES or 10 Hz PMS did not produce measurable after-effects on corticospinal excitability under the specific parameters used. However, the two interventions elicited distinct task-related cortical hemodynamic response patterns: PMS induced focal activation of the contralateral SMC, whereas NMES involved a more widespread cortical network encompassing sensorimotor and prefrontal regions.