γ-transcranial alternating current stimulation improves upper-limb motor function in stroke patients and modulates oscillatory imbalance: a pilot randomized controlled trial

Abstract

Background

Post-stroke cortical oscillatory activity is frequently disrupted, characterized by increased low-frequency oscillation and reduced high-frequency oscillation, which are associated with stroke severity and motor dysfunction. γ-tACS has the potential to entrain brain oscillations and enhance motor performance in healthy individuals, yet evidence in stroke patients is still limited.

Methods

This randomized, double-blind, sham-controlled pilot study enrolled 42 patients with post-stroke upper-limb motor impairment. Participants were randomly assigned to either the γ-tACS group or the sham-tACS group. High-definition tACS was applied over the M1 hotspot (70 Hz, 2 mA peak-to-peak, 20 min/session) for 10 sessions, once daily, in addition to conventional rehabilitation. The primary outcome was the Fugl–Meyer Assessment-Upper Extremity (FMA-UE). Secondary outcomes included the Modified Barthel Index (MBI) and resting-state EEG measures, including spectral power, power-ratio indices, and functional connectivity.

Results

A total of 42 participants were included in this trial (γ-tACS, n = 21; sham-tACS, n = 21), and the baseline characteristics were comparable between groups. FMA-UE improved in both groups, but the improvement was greater in γ-tACS group (ΔFMA-UE = 4.64 ± 4.55, P = 0.039, 95%CI=[0.154, 5.799]), whereas between-group differences in MBI improvement were not significant. Baseline EEG showed ipsilesional–contralesional asymmetry, with higher delta power, lower alpha power and elevated delta alpha ratio (DAR), delta beta ratio (DBR), and (delta+theta)/(alpha+beta) ratio (DTABR) on the ipsilesional side. After treatment, γ-tACS reduced ipsilesional delta power and lowered DBR and DTABR. Greater reductions in delta power and DTABR were associated with greater improvement in FMA-UE. In addition, γ-tACS increased ipsilesional beta-band connectivity (FC3-C3 wPLI), and this increase was also associated with motor improvement.

Conclusions

Compared with sham-tACS, γ-tACS applied over the ipsilesional M1 improved upper-limb motor function in patients with stroke. These clinical improvements were accompanied by restored oscillatory balance and enhanced network connectivity. Quantitative EEG parameters may serve as useful biomarkers for tracking neurophysiological changes and treatment response during stroke recovery.

Trial registration This study was registered in the Chinese Clinical Trial Registry (ChiCTR2300074898, date of registration: 2023/08/18).

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12984-026-02103-6.

Background

Stroke remains a leading cause of long-term disability, reduced quality of life, and considerable socioeconomic burden worldwide [1]. Approximately 80% of stroke survivors suffer from upper limb motor dysfunction [2], which severely limits daily activities and independence [3]. Although rehabilitation strategies have advanced considerably, motor recovery often remain incomplete, underscoring the need for innovative therapeutic approaches to enhance neural repair and functional outcomes.

In stroke survivors, oscillatory brain activity is often disrupted and reflects impaired cortical communication and altered network reorganization [4]. Previous EEG studies have shown that, compared with healthy controls, stroke patients exhibit increased low-frequency (delta and theta) and decreased high-frequency (alpha and beta) oscillations [5]. Enhanced low-frequency activity has been associated with larger infarct volume and more severe neurological deficits [6–8]. This abnormal slow oscillatory pattern is thought to reflect impaired cortico-cortical connectivity, reduced excitatory-inhibitory balance, and diminished capacity for efficient information processing in the lesioned hemisphere. Because these pathological slow rhythms interfere with motor network communication, restoring the balance between low- and high-frequency oscillations is considered a critical step toward re-establishing functional cortical dynamics and facilitating motor recovery.

Transcranial alternating current stimulation (tACS), a noninvasive brain stimulation (NIBS) technique, delivers weak alternating electric currents through the scalp to alter the probability of a neuron generating action potentials [9]. tACS is believed to entrain endogenous neural oscillations in a frequency-dependent manner, thereby influencing cortical synchronization and promoting neuroplasticity [10]. Because of its ease of application, safety, and high tolerability, tACS holds promise for integration into stroke rehabilitation.

Recent studies have explored the therapeutic potential of tACS in stroke rehabilitation. Xie et al. applied θ-tACS over the supplementary motor area (SMA) in patients with chronic post-stroke aphasia and reported improvements in speech production and auditory comprehension [11]. Schuhmann et al. found that differences in interhemispheric alpha activity reflect attentional deficits, and showed that α-tACS can improve hemispatial neglect in stroke survivors [12]. These findings suggest that tACS can modulate cortical oscillations in a frequency-dependent way and lead to functional improvements. However, evidence supporting the use of tACS for post-stroke motor rehabilitation is still limited.

Stimulation frequency is a key determinant of tACS efficacy. Identifying the optimal frequency is therefore critical to maximizing its therapeutic potential. In our previous study, healthy participants received three types of stimulation (20 Hz β-tACS, 70 Hz γ-tACS, and sham), and the results indicated that γ-tACS (70 Hz) was most effective in enhancing motor cortical excitability [13]. These findings are supported by other studies showing that γ-tACS outperforms other frequencies in augmenting motor cortical excitability and improving motor performance [14–16].

Despite these promising findings, few studies have systematically investigated the therapeutic efficacy and neurophysiological mechanisms of γ-tACS in stroke patients. Whether modulation of cortical oscillatory dynamics contributes to functional improvement remains unclear. Therefore, we hypothesized that γ-tACS could modulate abnormal cortical oscillations and promote upper limb motor recovery after stroke. To verify this hypothesis, we conducted a randomized, double-blind, sham-controlled pilot study combining clinical and EEG assessments to evaluate both the behavioral efficacy and electrophysiological effects of γ-tACS in stroke patients with upper limb hemiparesis.

Methods

Study design and participants

This study was a randomized, double-blind, sham-controlled trial conducted at the Department of Rehabilitation Medicine, The Third Affiliated Hospital of Sun Yat-sen University. We recruited hospitalized patients aged 18 years and older, with a first-ever, unilateral hemisphere stroke confirmed by CT, MRI or both, 2 weeks to 6 months after stroke, with hemiparesis of upper limb. Patients were excluded if they had any of the following condition: moderate-to-severe spasticity of extensor and flexor muscles of elbow, hand and wrists(Modified Ashworth Scale ≥ 2); severe aphasia or cognitive impairment, cranial bone defects, metallic or electronic implants, uncontrolled epilepsy or a history of epileptic episodes within the last 3 months, pregnancy, or any condition that, in the medical judgment, may interfere with the proper development of the treatment. The protocol was approved by the institutional ethics committee (RG2023-279-02), and written informed consent was obtained from all participants. This trial was registered at the Chinese Clinical Trial Registry (ChiCTR2300074898).

Interventions

tACS was administered using an electrical stimulation device (VOLCAN, VC-8000 F, Nanjing, China) with a 4 × 1 high-definition (HD) electrode montage (12 mm, AgCl, ring electrodes). The central electrode was placed over the motor hotspot, surrounded symmetrically by four return electrodes positioned 3 cm away from the center. A biphasic alternating current waveform was used, with a current intensity of 2 mA (peak-to-peak) applied for 20 min. Before stimulation, electrode–skin contact quality was checked using the device-specific Quality Unit (QU) index, and stimulation was initiated only when the QU value was below 0.50, according to the manufacturer’s calibration standard. The stimulation phase was set at 0°. Each session lasted 20 min, administered once daily for a total of 10 sessions.

Transcranial Magnetic Stimulation (TMS) device (YIRUIDE, YRD CCY-I, Wuhan, China) was used to locate M1 hotspot. The motor hotspot was identified as the scalp location that produced the largest Motor Evoked Potentials (MEP) in the first dorsal interosseous (FDI) muscle, recorded by surface electromyography (EMG). Participants were instructed to sit comfortably while accepting tACS treatment. The stimulation site was located using a 10–20 EEG cap, with the central electrode aligned over the hotspot. In patients without elicitable MEP on the affected side, the motor hotspot was identified in the unaffected hemisphere and its symmetrical location was used for stimulation over the affected hemisphere.

In addition to the study intervention, both groups received standard upper-limb rehabilitation after each tACS/sham stimulation session. The rehabilitation program included task-oriented training for reaching and grasping, range-of-motion and stretching exercises, strengthening and coordination training, and functional activities of daily living practice. The therapists responsible for delivering rehabilitation were blinded to group allocation. No other noninvasive brain stimulation (e.g., tDCS or TMS) was permitted during the intervention period. Medication regimens remained unchanged after enrollment, with no dose adjustments throughout the trial.

Randomization and blinding

Participants were randomly assigned in a 1:1 ratio to the γ-tACS group or the sham-tACS group using a computer-generated randomization sequence. Group allocation was concealed in sealed, opaque envelopes prepared by an independent researcher not involved in participant recruitment, stimulation delivery, or outcome assessment.

Both participants and outcome assessors were blinded to group allocation. The investigator responsible for administering stimulation was not involved in clinical assessments or data analysis.

In the sham-tACS group, current was delivered only during a 15-second ramp-up at the beginning and a 15-second ramp-down at the end of the session, with no stimulation applied during the remaining treatment period.

After the intervention period, participants were asked whether they perceived electrical stimulation to assess the success of blinding. As shown in Supplementary Data 1.2, the distribution of responses did not differ significantly between the γ-tACS and sham-tACS groups, suggesting that participant blinding was adequately maintained.

Outcome evaluation

Functional evaluation

The primary outcome was the mean change in upper limb motor recovery, evaluated using the Fugl Meyer Assessment-upper extremity (FMA-UE). The scale included 33 items and with a maximum score of 66. Each item is rated on a scale of 0, 1, or 2, where 0 indicates complete inability to perform the task, 2 indicates full performance of the task, and 1 indicates partial ability. A higher score reflects better motor performance.

The Modified Barthel Index (MBI) was used to assess the patient’s independence in daily living activities, comprising 10 items with a maximum score of 100.

The evaluation was performed by an experienced occupational therapist who was blinded to the group allocation. Evaluations were performed on the day before treatment and on the second day after treatment completion.

EEG assessment

Patients were instructed to sit comfortably in a chair, close their eyes, and wear the EEG cap (EGI-64 MicroCel Geodesic Sensor Net, Electrical Geodesics Inc., USA) equipped with Ag/AgCl electrodes. The electrode cap was connected to an amplifier (Net Amps 400, Electrical Geodesics Inc., USA). The sampling rate was set to 1000 Hz, with 64 electrodes in total. For online signal quality monitoring during recording, the signals were displayed with a low-pass filter at 30 Hz, a high-pass filter at 0.01 Hz, and a notch filter at 50 Hz; these filters were applied to the online display only and did not alter the stored raw data. Medical conductive paste, medical alcohol, and scrubbing cream were used to reduce impedance, which was kept below 5 kΩ. Resting-state EEG (RS-EEG) signals were recorded for 3 min.

EEG recordings and clinical assessments (FMA-UE and MBI) were performed on the same day, including baseline (before treatment) and post-intervention assessment.

EEG data analysis

EEG data preprocessing and analysis were performed using MATLAB R2024b (MathWorks, Natick, MA, USA).

Preprocessing: A high-pass filter at 0.1 Hz, low-pass filter at 50 Hz, and notch filter at 50 Hz were applied. Bilateral mastoid electrodes were used as reference electrodes. The data were segmented into 2-second epochs. Channels with poor data quality were identified by visual inspection based on excessive noise or persistent artifacts and were subsequently interpolated. Epochs with amplitudes larger than 100 µV were excluded. Artifacts from eye movements, blinks and head motion were removed by using Independent Component Analysis (ICA). For participants with right-hemisphere stroke, EEG channels were left–right relabeled to align the ipsilesional hemisphere across participants (i.e., ipsilesional channels were mapped to the left side) for further analysis.

Fast Fourier Transform (FFT) with a Hamming window was applied to convert time-domain EEG signals into the frequency domain and estimate power spectral density. Relative power was calculated for the following bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–45 Hz). To reduce the influence of single-channel signal-to-noise ratio (SNR), the power of each region of interest (ROI) was computed as the average of the target channel and its four surrounding electrodes (e.g., C3 averaged with C5, FC1, FC3, and CP1).

The following power ratio index (PRI) were calculated [17]: Delta/Alpha Ratio (DAR), Delta/Beta Ratio (DBR), Theta/Alpha Ratio (TAR), Theta/Beta Ratio (TBR), and Delta+Theta/Alpha+Beta Ratio (DTABR).

For functional connectivity (FC) analysis, the EEG data were re-referenced to common average reference. To avoid the influence of volume conduction effects, we calculated the Weighted Phase-Lag Index (wPLI) at the ROI level for further analysis [18]. Selected ROIs were:

M1(left: CP1, C3, C5; right: CP2, C4, C6),

PMC (left: FC1, FC3, FC5; right: FC2, FC4, FC6),

M1 hotspot area (left: F1, F3, F5; right: F2, F4, F6),

SMA (C1, Cz, C2).

Sample size and statistical analysis

The sample size was calculated using G*Power 3.1. We considered a medium effect size (E = 0.25), with a statistical power (1-β) of 0.8 and a significance level (α) of 0.05, the sample size was 34 participants (17 in each group). We compensated for a 20% drop-out rate, therefore, 42 participants were recruited.

Continuous variables are presented as mean ± standard deviation (SD) for descriptive statistics. For group-level graphical presentation, data are shown as mean ± standard error of the mean (SEM) or as box plots, as appropriate. The specific data presentation format is described in each figure legend. Qualitative data are presented as frequencies (percentages). All statistical inferences were based on two-tailed tests, with p-value < 0.05 considered statistically significant. 95% confidence intervals were used for parameter estimates. The normality and homogeneity of variance of the data were tested using Shapiro-Wilk test and Levene’s test. Inter-group effects were compared using independent samples t-test, and intra-group effects were compared using paired t-test. The Wilcoxon rank-sum test is used when data do not meet the assumptions of normality or homogeneity of variance. For EEG analyses involving multiple comparisons across frequency bands and connectivity measures, false discovery rate (FDR) correction was applied to control for multiple testing. We correlated any significant changes between brain activity and behavioral clinical scales by using Pearson correlation analysis.

Results

Participant enrollment and flow

A total of 238 stroke patients were screened for eligibility. 42 patients met the inclusion criteria and were enrolled in the study.

During the intervention period, 4 participants withdrew from the study. In the γ-tACS group, 2 participants discontinued due to early discharge and inability to complete outcome assessments. In the sham-tACS group, 2 participants withdrew because of medical reasons unrelated to the intervention and personal reasons. No withdrawals were due to adverse events related to the intervention (Fig. 1).

Fig. 1.

Consort patient flow chart

Baseline characteristics of stroke participants

There were no significant differences in baseline demographic or clinical characteristics between the γ-tACS and sham-tACS groups. Specifically, the two groups were comparable in terms of gender distribution (P = 0.485), age (P = 0.243), stroke type (P = 0.179), stroke location, and affected hemisphere (P = 0.744). No group differences were observed in the prevalence of hypertension (P = 0.724) or diabetes (P = 0.461). Time from stroke onset, FMA-UE, MBI scores, MMSE and MEP recruitment rates were also similar between groups (P > 0.05), indicating good baseline comparability (Table 1).

Table 1.

Baseline demographic and clinical characteristics of the patients

Variable	Group		T or χ² Value	P Value
	γ-tACS (n = 19)	sham-tACS (n = 19)
Gender: n(%)
Male	14 (73.7)	12 (63.2)	χ²=0.487	0.485
Female	5 (26.3)	7 (36.8)
Age, years	52.26 ± 14.41	57.79 ± 14.27	T=-1.188	0.243
Stroke type: n(%)
Ischemic	10 (52.6)	14 (73.7)
Hemorrhagic	9 (47.4)	5 (26.3)	χ²=1.810	0.179
Stroke location: n(%)
Subcortical	15 (78.9)	15 (78.9)
Mixed	4 (21.1)	4 (21.1)
Affected hemisphere, n (%)
Right	9 (47.4)	8 (42.1)
Left	10 (52.6)	11 (57.9)	χ²=0.106	0.744
Hypertension: n(%)
With	12 (63.2)	17 (89.5)
Without	7 (36.8)	2 (10.5)	χ²=0.124	0.062
Diabetes: n(%)
With	6 (31.6)	4 (21.1)
Without	13 (68.4)	15 (78.9)	χ²=0.543	0.461
Time from the onset, Weeks	11.05 ± 11.15	13.16 ± 11.04	T=-0.585	0.562
MMSE	20.00 ± 4.84	21.43 ± 5.77	T=-0.844	0.404
FMA-UE	23.36 ± 16.10	25.26 ± 12.64	T=-0.417	0.679
MBI	61.21 ± 25.76	63.44 ± 24.11	T=-0.252	0.803
MEP recruitment rates
Recruited	9 (47.4)	7 (36.8)
Unrecruited	10 (52.6)	12 (63.2)	χ²=0.432	0.511

MMSE: Mini-mental State Examination; FMA-UE: Fugl-Meyer Assessment Upper Extremity; MBI: Modified Barthel Index; MEP: Motor Evoked Potential

Functional improvement after tACS treatment

As shown in Table 1, the baseline levels of FMA-UE and MBI were comparable before treatment. Table 2 shows that both groups showed within-group improvements in FMA-UE scores (γ-tACS: ΔFMA-UE = 4.64 ± 4.55; sham-tACS: ΔFMA-UE = 1.67 ± 3.27), with the γ-tACS group showing a significantly greater improvement than the sham-tACS group (T = 2.154, P = 0.039, 95%CI: [0.154, 5.799]). However, no significant between-group difference was observed in MBI score changes.

Table 2.

Statistic results of FMA-UE and MBI

Scale		γ-tACS	sham-tACS	T Value	P Value
FMA-UE	Pre	23.36 ± 16.10	25.26 ± 12.64	-0.417	0.679
	Post	28.00 ± 18.12	27.17 ± 14.34
	T Value	3.817	2.161
	P Value	0.002**	0.045*
	Post-pre	4.64 ± 4.55	1.67 ± 3.27	2.154	0.039*
MBI	Pre	61.21 ± 25.76	63.44 ± 24.11	-0.252	0.803
	Post	71.86 ± 23.89	69.28 ± 21.52
	T Value	4.915	3.084
	P Value	< 0.001***	0.007**
	Post-pre	10.64 ± 8.10	5.83 ± 8.02	1.675	0.104

FMA-UE: Fugl-Meyer Assessment Upper Extremity; MBI: Modified Barthel Index

* P < 0.05; ** P < 0.01; *** P < 0.001

Lateralization of motor cortex oscillatory power in stroke patients

To characterize oscillatory imbalance after stroke, baseline EEG data from all participants (n = 38) were analyzed. Relative power was extracted from the ipsilesional (C3) and contralesional (C4) regions with four surrounding electrodes. As shown in Fig. 2A-C, stroke patients exhibited hemispheric asymmetry in oscillatory activity. The ipsilesional hemisphere showed significantly higher delta power (T = 3.304, P = 0.003) and lower alpha power (T = -3.400, P = 0.002) compared to the contralesional side. Additionally, quantitative EEG indices reflecting the ratio between low- and high-frequency bands were significantly elevated on the ipsilesional side, including DAR (T = 2.470, P = 0.021), DBR (T = 3.497, P = 0.002), and DTABR (T = 3.174, P = 0.004). These findings indicate a shift toward enhanced low-frequency and reduced high-frequency activity in the lesioned hemisphere after stroke.

Fig. 2.

Lateralization of EEG relative power in stroke patients. A Group-averaged power spectra at the C3 and C4 for the ipsilesional (Ips) and contralesional (Con) hemispheres, shown on a logarithmic scale (log₁₀ µV²/Hz). Shaded areas indicate ± SEM. Colored backgrounds denote the canonical frequency bands; B Scalp topographies of spectral power for each frequency band. Color scales indicate the spatial distribution of band-specific power across electrodes; C Raincloud plots comparing band-averaged relative power (delta and alpha) and power ratio indices (DAR, DBR, DTABR) between Ips and Con hemispheres. Dots represent individual participants, and boxes indicate median and interquartile range. (* P < 0.05; ** P < 0.01)

Frequency-specific spectral changes and behavioral correlation

Figure 3A shows the dB-transformed pre-post changes in the power spectrum at the ipsilesional motor ROI (C3). The γ-tACS group demonstrated an overall reduction in low-frequency power with a concomitant increase in higher-frequency power, whereas no comparable spectral shift was observed in the sham-tACS group. Figure 3B presents scalp topographies of band-specific power distributions for the pre- and post-intervention conditions.

Fig. 3.

Frequency-specific spectral changes and behavioral relevance A dB-transformed changes in power spectra (ΔPSD = 10·log₁₀[post/pre]) at the C3 ROI in the γ-tACS and sham-tACS groups. The γ-tACS group showed a reduction in low-frequency power and a concomitant increase in high-frequency power after the intervention, whereas no comparable pattern was observed in the sham-tACS group. Shaded backgrounds denote the canonical frequency bands. B Scalp topographies of band-specific power in each frequency band, shown separately for the pre- and post-intervention conditions in the γ-tACS and sham-tACS groups. The same color scale was applied to the γ-tACS and sham-tACS conditions to allow direct visual comparison of scalp power distributions. C Raincloud plots showing pre- to post-intervention changes in delta power and power ratio indices (DBR and DTABR). Each dot represents an individual participant, and boxes indicate median and interquartile range. D Correlations between changes in EEG spectral indices (post-pre) and improvements in motor function. Shaded areas represent 95% confidence intervals. (* P < 0.05, ** P < 0.01)

At the C3 ROI, pre-post comparisons indicated no significant changes in band power in the sham-tACS group (P > 0.05). However, DAR decreased significantly after sham intervention (T = -2.45, P = 0.034), while other ratio indices showed no significant changes (Supplementary data 1.6). In contrast, the γ-tACS group showed a significant reduction in delta power after treatment (T = -2.82, P = 0.012), with no statistically significant changes in theta, alpha, beta, or gamma power. In addition, DBR and DTABR decreased significantly following γ-tACS (T = -2.66, P = 0.017; T = -2.30, P = 0.035) (Fig. 3C; detailed statistics in Supplementary data 1.6).

To assess behavioral relevance, Pearson correlation analyses were conducted between changes in EEG indices that showed significant pre-post effects in the γ-tACS group and improvements in motor function (FMA-UE, post-pre). As shown in Fig. 3D, FMA-UE improvement was negatively correlated with changes in delta PSD (r = -0.739, P = 0.002) and DTABR (r = -0.610, P = 0.016). No significant association was observed between FMA-UE improvement and DBR change (r = -0.269, P = 0.333). Additional correlation analyses in the sham-tACS group did not reveal significant associations between FMA-UE improvement and changes in EEG indices (Supplementary Data 1.7).

Functional connectivity enhancement after tACS treatment

ROI-based functional connectivity analysis identified a significant effect in the FC3–C3 connection (Fig. 4A). Using Mann–Whitney U tests, the γ-tACS group showed a significant post-intervention increase in wPLI between FC3 and C3 when compared with the pre-intervention distribution (P = 0.0261; Fig. 4B). In contrast, no significant pre–post difference was observed in the sham-tACS group.

Fig. 4.

Changes in functional connectivity and association with motor improvement. A ROI pairs included in the functional connectivity (FC) analysis. The red arrow (FC3–C3) indicates the significant result. B Group comparisons of the weighted phase lag index (wPLI) for FC3–C3 before and after the intervention in the γ-tACS and sham-tACS groups. Boxes represent the median and interquartile range, and whiskers indicate data dispersion. C Correlation between changes in functional connectivity (ΔwPLI, post-pre) and changes in motor function (ΔFMA-UE, post-pre). The solid line indicates the linear regression fit, and the shaded area denotes the 95% confidence interval. (* P < 0.05)

To assess behavioral relevance, we further examined the association between connectivity changes and motor improvement. The change in FC3–C3 wPLI (post–pre) was positively correlated with the change in FMA-UE scores (post–pre) (r = 0.58, P = 0.0138; Fig. 4C), indicating that larger increases in functional connectivity were associated with greater gains in upper limb motor function.

Safety

No significant adverse events occurred during this trial. Only mild adverse effects were reported, including itching, tingling and burning sensations during stimulation process. These sensations resolved immediately after stimulation without any residual effects.

Discussion

In this randomized, double-blind, sham-controlled pilot study, γ-tACS applied over the ipsilesional M1 produced greater improvement in upper-limb motor impairment than sham stimulation, as reflected by a larger gain in FMA-UE scores. In parallel, γ-tACS induced measurable neurophysiological changes, including reduced delta power, lower slow-to-fast power ratio (DBR and DTABR), and enhanced beta-band functional connectivity (wPLI) within the ipsilesional motor network. Moreover, these electrophysiological changes were associated with behavioral improvement. Taken together, these findings suggest that γ-tACS may promote motor recovery by rebalancing abnormal post-stroke oscillatory activity and strengthening motor-network communication.

Hemispheric asymmetry of brain oscillation after stroke

We analyzed baseline EEG data from all participants to characterize stroke-related changes in cortical oscillations. Compared with the contralesional hemisphere, the ipsilesional hemisphere showed higher delta power and lower alpha power. DAR, DBR, and DTABR were also higher on the affected side. These results are consistent with previous studies showing that stroke is associated with a shift toward low-frequency dominance and relative suppression of faster oscillations. Our baseline findings therefore confirm the presence of a pathological oscillatory pattern in this cohort and provide a physiological basis for interventions aimed at restoring more efficient cortical dynamics.

Previous EEG studies have reported similar hemispheric asymmetry after stroke. Zhang et al. [19] found increased delta activity and reduced beta power in patients with chronic stroke, and delta-band network measures were correlated with upper-limb motor impairment. Cassidy et al. [20] further proposed that delta oscillations may serve as a biomarker of post-stroke brain injury and recovery, because increased delta coherence was associated with more severe neurological deficits. In addition, quantitative EEG (qEEG) indices such as DAR and DTABR have been widely validated as reliable markers of functional impairment and prognosis after stroke [21–22]. Together, these observations support the view that pathological low-frequency dominance is a central electrophysiological feature of post-stroke motor dysfunction.

γ-tACS restores oscillatory balance and improves motor recovery

A major finding of this study is that γ-tACS altered the spectral profile over the ipsilesional motor cortex. After γ-tACS, delta power, DBR, and DTABR all decreased significantly. In contrast, the sham group did not show comparable changes in band power, although DAR decreased modestly. This modest reduction may reflect nonspecific influences, such as conventional rehabilitation, spontaneous recovery, or general changes in arousal. Importantly, greater reductions in delta power and DTABR were associated with greater improvement in FMA-UE. This pattern suggests that reducing pathological slowing is linked to motor recovery rather than representing a secondary epiphenomenon.

Comparable rhythm-dependent effects of γ-tACS have been reported in other neurological and psychiatric conditions, including depression [23] and Parkinson’s disease [24], in which modulation of abnormal oscillatory activity has been linked to symptom improvement. Although stimulation targets and clinical outcomes differ across conditions, these studies support a broader principle that frequency-specific stimulation can reshape maladaptive network rhythms. In the present study, γ-tACS did not primarily manifest as an increase in gamma power itself. Instead, its effects were more clearly reflected in attenuation of pathological slow activity and improvement in slow-to-fast power ratios. This pattern suggests that the physiological impact of γ-tACS may be better captured by network reorganization and spectral rebalancing than by direct enhancement of gamma-band power alone.

These results are consistent with the broader qEEG literature showing that lower slow-to-fast ratios reflect a better cortical functional state and can track rehabilitation progress [25]. From a mechanistic perspective, γ-tACS may act as a frequency-specific signal that helps counteract the post-stroke dominance of slow oscillations. Animal and translational studies further support the relevance of gamma rhythms. Stroke can disrupt gamma generation by impairing inhibitory interneuron networks, whereas successful recovery is accompanied by the restoration of gamma synchronization. Moreover, combining gamma-patterned neuromodulation with rehabilitation training has been shown to accelerate functional recovery in experimental models. This finding supports the view that gamma activity is not only a marker, but also a potential therapeutic target [26].

Although the main effects were observed in delta-related indices rather than in gamma power itself, this does not rule out a gamma-mediated mechanism. Previous studies in other populations have shown that coordinated stimulation at different frequencies can enhance cross-frequency coupling, such as theta–gamma coupling, and improve cognitive performance [27]. It is therefore possible that γ-tACS influenced post-stroke network dynamics through cross-frequency interactions, thereby reducing pathological slow rhythms and facilitating the re-emergence of faster and more efficient communication within motor-related networks. This possibility should be examined directly in future studies using phase–amplitude coupling metrics and other measures of cross-frequency interaction.

Beta-band connectivity enhancement and behavioral relevance

In addition to local spectral changes, we also observed network-level modulation after γ-tACS. Functional connectivity analysis showed a significant post-intervention increase in FC3–C3 wPLI in the γ-tACS group, whereas no significant change was observed in the sham group. Moreover, the increase in FC3–C3 connectivity was positively correlated with improvement in FMA-UE. This suggests that patients with greater strengthening of motor network communication also showed greater motor improvement.

FC3 is commonly considered to reflect premotor-related activity, whereas C3 is often used to index activity over the M1 region. Therefore, increased FC3-C3 connectivity may indicate enhanced synchronization within the ipsilesional frontal-central motor network. However, because FC3 and C3 are spatially adjacent electrodes, this finding should be interpreted cautiously and should not be considered definitive evidence of independent long-range premotor-motor coupling. Rather, it may reflect local ipsilesional motor-network synchronization associated with recovery-related changes.

Direct evidence that γ-tACS modulates beta-band EEG connectivity after stroke remains limited. Therefore, the present connectivity finding should be regarded as preliminary neurophysiological evidence that γ-tACS may influence motor-network coordination, rather than conclusive evidence of a specific connectivity mechanism. Future studies with larger samples, source-level EEG or MEG analysis, and combined EEG-EMG recordings are needed to determine whether γ-tACS can reliably modulate motor-network connectivity or cortico-muscular coupling after stroke.

Limitations

This study has several limitations. First, the sample size was small and the trial was conducted at a single center, and the study was designed as a pilot trial. These factors may limit both the generalizability of the findings and the statistical power to detect more subtle neurophysiological effects. Second, outcomes were assessed only immediately after the intervention, and the durability of the clinical and EEG effects remains unclear. Third, EEG has limited spatial resolution and is susceptible to artifacts, particularly in higher-frequency ranges. Although wPLI was used to reduce the influence of volume conduction on functional connectivity estimation, the short distance between adjacent electrodes, such as FC3 and C3, may still introduce residual spatial dependence between closely located ROIs. In addition, high-gamma activity around the 70 Hz stimulation frequency is characterized by relatively low signal-to-noise ratios and increased susceptibility to muscle and environmental noise in scalp EEG recordings. Therefore, the present study focused on the conventional low-gamma range (30–45 Hz) rather than directly analyzing oscillatory activity at 70 Hz. These limitations may have reduced our sensitivity to detect direct gamma-band entrainment or more complex coupling effects. Future studies combining modalities with higher sensitivity to high-frequency neural activity, such as magnetoencephalography (MEG), may help further elucidate the neurophysiological mechanisms underlying gamma-frequency stimulation. Fourth, the stimulation parameters were not individualized, such as by matching the stimulation frequency to intrinsic rhythms, and inter-individual variability may have influenced treatment responsiveness. Finally, participants received concurrent conventional rehabilitation. Therefore, the observed EEG and behavioral changes may partly reflect rehabilitation-related neuroplasticity or spontaneous recovery, in addition to the effects of γ-tACS. Although the sham-controlled design helps reduce nonspecific effects, it cannot fully separate the independent and interactive contributions of γ-tACS, rehabilitation, and spontaneous recovery. Future studies using factorial designs, rehabilitation-only control groups, stimulation-only conditions, and longitudinal EEG assessments may help better distinguish these effects and clarify whether γ-tACS provides additive or synergistic benefits when combined with conventional rehabilitation.

Conclusion

In this randomized, double-blind, sham-controlled pilot trial, γ-tACS applied over the ipsilesional M1 produced greater improvement in upper-limb motor function than sham stimulation. γ-tACS also reduced pathological slowing, reflected by lower delta power and slow-to-fast power ratios, and enhanced beta-band functional connectivity within the ipsilesional motor network; these electrophysiological changes were associated with gains in FMA-UE. Together, these findings suggest that γ-tACS may be a feasible adjunct to post-stroke upper-limb rehabilitation, while qEEG markers may help track treatment-related neurophysiological changes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

tACS

Transcranial alternating current stimulation

M1

Primary motor cortex

FMA-UE

Fugl–Meyer Assessment-Upper Extremity

MBI

Modified Barthel Index

EEG

Electroencephalography

CI

Confidence interval

DAR

Delta/Alpha Ratio

DBR

Delta/Beta Ratio

DTABR

Delta+Theta/Alpha+Beta Ratio

wPLI

Weighted Phase-Lag Index

qEEG

Quantitative electroencephalography

ROI

Region of interest

RS-EEG

Resting-state EEG

ICA

Independent Component Analysis

FFT

Fast Fourier Transform

PSD

Power spectral density

PRI

Power ratio index

TAR

Theta/Alpha Ratio

TBR

Theta/Beta Ratio

FC

Functional connectivity

PMC

Premotor cortex

SMA

Supplementary motor area

NIBS

Noninvasive brain stimulation

HD

High-definition

TMS

Transcranial Magnetic Stimulation

MEP

Motor Evoked Potential

FDI

First dorsal interosseous

EMG

Electromyography

MMSE

Mini-Mental State Examination

FDR

False Discovery Rate

SD

Standard deviation

SNR

Signal-to-noise ratio

SEM

Standard error of the mean

Ips

Ipsilesional

Con

Contralesional

Author contributions

CJW: Conceptualization, Methodology, Investigation, Formal analysis, Data Curation, Writing - Original Draft, Visualization; WFZ: Investigation (intervention application, EEG collection); LLC: Writing - Review & Editing; SXZ: Investigation (clinical assessment); LH: Investigation (clinical assessment); JJL: Investigation (enrollment); YNA: Resources; HYH: Investigation (allocation); JHX: Writing - Review & Editing; HQZ: Conceptualization, Methodology, Validation, Resources, Writing - Review & Editing, Supervision, Project administration, Funding acquisition.

Funding

This work was supported by the National Key R&D Program of China (Grant No. 2023YFC3603800, 2023YFC3603804); the National Natural Science Foundation of China (Grant: 82272605); the National Natural Science Foundation of China (Grant: NO.82472599); the Science and Technology Projects of Guangzhou (Grant:2024A03J0177).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University (No. RG2023-279-02), and all the subjects volunteered to participate and signed informed consent forms.

Consent for publication

Consent to publish was obtained from all the participants.

Competing interests

The authors declare no competing interests.