Study design

This study is a participant and outcome assessor-blinded, parallel-group, randomized controlled trial (RCT). Eligible participants will be randomly allocated to either the priming iTBS group (cTBS + iTBS), the non-priming iTBS group (sham cTBS + iTBS), or the sham stimulation group (sham cTBS + sham iTBS) in a 1:1:1 ratio. All groups will receive REX exoskeleton training as conventional therapy to facilitate sensorimotor loop integration through proprioceptive feedback. The intervention period will last for 4 weeks (5 days of intervention per week), with outcome assessments conducted at baseline (T0), midway at 2 weeks (T1), and immediately post-intervention at 4 weeks (T2), and a follow-up assessment will be conducted at 6 weeks (T3). This study is designed as a superiority trial. The trial will be conducted at the Department of Rehabilitation Medicine, Yancheng NO.1 People’s Hospital, Yancheng, Jiangsu, China. Fig 1 and Table 1 outline the trial procedures in detail.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Schedule of enrolment, interventions, and assessments.

https://doi.org/10.1371/journal.pone.0349578.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of the study design.

https://doi.org/10.1371/journal.pone.0349578.g001

This study has been approved by the Ethics Committee of Yancheng NO.1 People’s Hospital (approval number: 2025-K-184) and registered at the Chinese Clinical Trial Registry (ChiCTR2500110279). All participants will provide informed consent directly.The timeline for this study is as follows: recruitment of participants will commence on 01/02/2026 and is expected to be completed by 01/08/2026.; data collection is projected to conclude by 01/11/2026, with the aim of obtaining preliminary results by 01/12/2026.This study protocol will adhere to the principles of Helsinki and the Standard Protocol Items for Randomized Trials (SPIRIT) statement, and the checklist of this protocol is provided in the relevant supplementary materials (S1 File). Enrolling 60 participants within 6 months is feasible, supported by the approximately 500 annual stroke admissions at Yancheng NO.1 People’s Hospital, which ensures a sufficient candidate pool.

Patient recruitment

Participants will be recruited from inpatient and outpatient stroke rehabilitation wards at Yancheng NO.1 People’s Hospital. Recruitment strategies include: (1) Displaying study posters and distributing flyers in relevant clinical areas. (2) Regular presentations about the trial to clinical staff (physicians, therapists) to facilitate patient referrals. (3) Screening of hospital admission records for potentially eligible patients by the research coordinator. An independent research coordinator will screen individuals for eligibility and obtain written informed consent prior to enrolment.

Inclusion criteria

(1) Stroke diagnosed by CT or MRI. (2) Age 40–80 years old. (3) Stable condition, first-ever stroke (patients with a history of recurrent strokes will be explicitly excluded), duration 1–3 months. (4) Able to walk independently or with assistance for more than 10m, Brunnstrom stage III-V lower limb. (5) Cognitive clarity, ability to follow simple verbal commands or instructions, MoCA ≥ 26. (6) Informed consent signed by the patient or his/her relatives. (7) No history of epilepsy. (8) No contraindications to TMS examination.

Exclusion criteria

(1) Individuals with relevant contraindications to TMS (pacemakers, intracranial implants, implanted drug pumps etc.). (2) Significant speech, attention, auditory, visual (lateral neglect, hemianopsia, etc.), intellectual, psychiatric or cognitive dysfunction. (3) Prior history of organic brain disease, neuropsychiatric history, drug abuse and alcoholism. (4) Patients with severe failure of vital organs (e.g., heart, lung, liver, kidney). (5) Patients with pre-existing lower limb motor dysfunction prior to the stroke. (6) Those who are currently participating in other clinical trials. (7) Those who cannot co-operate with the functional assessment and have poor compliance. (8) Inability to induce a TMS response in the tibialis anterior muscle prior to formal testing. (9) Pregnancy. (10) Use of drugs that affect cortical excitability (e.g., benzodiazepines, baclofen, antiepileptics, certain antidepressants) that cannot be safely tapered or washed out prior to study participation.

Randomization and blinding

A total of sixty eligible stroke patients will be enrolled in this study. Following the baseline assessment, a computer-generated randomization sequence will be used to assign all participants in a 1:1:1 ratio to one of three parallel intervention groups: the priming iTBS group (n = 20), the non-priming iTBS group (n = 20), or the sham stimulation group (n = 20). The allocation sequence will be concealed from the researchers enrolling participants. After a participant provides informed consent and all baseline assessments (T0) are completed, the research coordinator will access the central web-based system. The system will then reveal the group allocation only to the unblinded physical therapist responsible for administering the intervention. This ensures that outcome assessors and participants remain blinded.

Physiotherapists and researchers delivering the conventional interventions will have access to the patient’s corresponding intervention group number through the system but will not be blinded to the specific intervention type. Blinding will be implemented for participants and outcome assessors throughout the trial. Blinding will be maintained until the end of the trial, after data locking, at which point unblinding will be performed.

Emergency unblinding will be permitted only in the event of a serious adverse event (SAE) where knowledge of the participant’s allocation is deemed necessary for their medical management. The request for unblinding will be made by the treating physician to the principal investigator, who will then access the randomization system. All instances of unblinding will be documented and reported to the Ethics Committee.

Conventional rehabilitation protocol

All participants in the three groups will receive a standardized conventional rehabilitation regimen in addition to their respective trial interventions. This comprehensive foundational therapy is tailored to the individual’s needs but consistently includes the following core components:

Physical Therapy: Sessions target the improvement of gross motor functions, including bed mobility, transfers (e.g., from bed to chair), balance control in sitting and standing, and gait training. A key focus is placed on strength training for major muscle groups of the upper and lower limbs, utilizing modalities such as resistance bands, free weights, and weight-bearing exercises to address hemiparetic weakness. Additionally, therapists employ techniques to mitigate spasticity, facilitate normal movement patterns, and improve joint range of motion and flexibility.

Occupational Therapy: Interventions are designed to promote independence in Activities of Daily Living (ADLs), such as feeding, grooming, dressing, and bathing. Therapists also address upper limb function through targeted exercises for fine motor skills, coordination, and dexterity.

Furthermore, all patients will undergo a standardized robot-assisted training regimen utilizing the REX self-balancing exoskeleton rehabilitation robot (Rex Bionics Ltd., New Zealand). All robot-assisted training sessions are conducted after the TBS intervention. The interval between them should be minimized (less than 15 minutes). Each daily 30-minute session comprises three distinct 10-minute programs:

Program 1: Standing Activity Training: This component focuses on improving trunk control and upper-lower limb coordination during stable standing, exemplified by tasks such as targeted reaching.

Program 2: Resistance Training with Elastic Bands: Employing Proprioceptive Neuromuscular Facilitation (PNF) patterns, this program is designed to facilitate affected lower limb extension and enhance muscular strength and coordination.

Program 3: Lower Limb Functional Training: This segment aims to improve functional strength, balance, and mobility through exercises including single-leg weight-bearing, lateral stepping, and squats.

All robot-assisted training is supervised by experienced physical therapists. They dynamically adjust the level of robotic support and task difficulty based on real-time assessment of patient performance and tolerance, while providing concurrent verbal feedback. Detailed operational procedures, parameters, and monitoring protocols are rigorously specified in the study’s Standard Operating Procedure (SOP) (S2 File). Fig 2 shows the REX exoskeleton robot device and relevant treatment content.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. REX exoskeleton rehabilitation robot.

https://doi.org/10.1371/journal.pone.0349578.g002

iTBS intervention protocol

Prior to each robot-assisted training session, the iTBS intervention will be administered using the Wuhan Yiruidian paired transcranial magnetic therapy device (model: N50000) with its matching double-cone coil. The stimulation will target the “hotspot” within the lower limb M1 area. This site will be determined by identifying the location that elicits the maximum motor evoked potential (MEP) amplitude in the tibialis anterior muscle, with the resting motor threshold (rMT) tested to assist in localization.

The intervention protocol will consist of two sequential parts: priming (using cTBS) followed by excitation (iTBS). The cTBS segment will deliver 600 pulses over approximately 40 seconds, using 200 clusters of uninterrupted stimulation at a 50 Hz pulse frequency and a 5 Hz repetition frequency. Subsequently, the iTBS segment will be administered. It will consist of a 2-second train of 3 pulses at 50 Hz, followed by an 8-second rest period. This 10-second cycle will be repeated 20 times, resulting in a total of 600 pulses over approximately 192 seconds. All stimulus intensities will be set at 80% of the individual’s active motor threshold (aMT), which is defined as the minimum intensity required to elicit at least 5 out of 10 peak-to-peak MEPs exceeding 200 μV in the tibialis anterior muscle.

For sham stimulation, the same double-cone coil used in the active condition will be employed. It will be configured to operate in a special sham mode, which mimics the audible “clicking” sound of active transcranial magnetic stimulation without delivering a clinically significant magnetic field to the scalp. The coil will be placed at the identical scalp position and held at the same angle as during active stimulation, using the same holder to ensure consistent positioning and blinding. This setup will replicate the acoustic characteristics of real TBS, generating a similar loud sound without delivering effective stimulation, thereby ensuring participant blindness to the intervention. The specific training operation details will be shown in the Standard Operating Procedure (SOP).

Outcome measures

Assessments will be conducted at baseline (T0), 2 weeks (T1),4 weeks (T2) after intervention, and 6 weeks of follow-up (T3). Outcome measures will be collected by trained assessors blinded to group allocation. Assessors will have at least 5 years of clinical experience and receive specific training on standardized administration of all assessment tools prior to the trial start.

Primary outcome

The primary outcome is the change in the Fugl-Meyer Assessment of Lower Extremity (FMA-LE) score from baseline (T0) to post-intervention (T2, 4 weeks). The FMA-LE quantifies the impairment and recovery of lower extremity motor function by systematically evaluating the dimensions of reflexes, synergistic movements, and coordination. The FMA-LE is scored on a 3-point ordinal scale: 0 for severe dysfunction, 1 for partial dysfunction, and 2 for no dysfunction, with a maximum total score of 34 points. With good reliability and validity, the FMA-LE has been widely recognized in clinical practice.

Secondary outcomes

Neurophysiological and biomechanical outcomes

Functional near-infrared spectroscopy (fNIRS) tests will be performed using the NirSmart-3000B device (Danyang Huichuang Medical Equipment Co., Ltd., China) to observe changes in the concentrations of oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) in various brain regions. The fNIRS system consists of 16 light sources and 18 detectors (resulting in 40 active channels) with dual wavelength detection at 730nm and 850nm and a sampling frequency set to 11 Hz. Assessments will be conducted in baseline (T0), mid-intervention (T1, 2 weeks), and (T2, 4 weeks). Prior to fNIRS data acquisition, participants will remain in a quiet and relaxed state for approximately 1 minute to ensure a stable hemodynamic baseline. Subsequently, an 8-minute resting-state fNIRS scan will be conducted with participants maintaining a wakeful, motionless condition. Fig 3 shows the layout of the fNIRS system.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Layout of the fNIRS optodes and measurement channels.

https://doi.org/10.1371/journal.pone.0349578.g003

Electroencephalography signals will be acquired using a high-density 64-channel system (GES 400, EGI, USA). Electrodes will be arranged according to the international 10–20 system, with Cz as the reference, to record scalp EEG from 64 channels along with signals from bilateral mastoids. The system specifications (input impedance ≥1000 MΩ, sampling rate ≥8 kHz per channel, and input noise ≤1 μV) will ensure the capture of high-fidelity data. These signals will be analyzed to investigate treatment-induced changes in cortical activation and functional network dynamics as neural correlates of motor recovery. Prior to EEG signal acquisition, participants will remain in a quiet and relaxed state for approximately 1 minute to ensure a wakeful, relaxed, and motionless baseline condition. Subsequently, a 5-minute EEG recording will be conducted under eyes-closed resting conditions. Assessments will be conducted at baseline (T0), mid-intervention (T1, 2 weeks), post-intervention (T2, 4 weeks). Fig 4 shows the layout of the EEG system.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Configuration of the high-density EEG electrode system.

https://doi.org/10.1371/journal.pone.0349578.g004

Gait data collection was performed using the Qualisys three-dimensional motion capture system. This system operates at a frequency of 100 Hz to accurately measure the three-dimensional coordinates of reflective markers placed at specific anatomical locations within the coronal, sagittal, and horizontal planes, thereby recording spatiotemporal and kinematic parameters that describe the body’s spatial movement during walking. Equipped with three-dimensional motion sensing and adaptive technology, the system utilizes computer learning algorithms to automatically track patients’ step frequency and gait characteristics, enabling precise detection of multi-dimensional angular changes. During walking tests, it automatically collects and analyzes basic gait parameters and continuously records usage data after a single power-on, including information such as the number of steps and time per step. Assessments were conducted at baseline (T0), mid-intervention (T1, week 2), and post-intervention (T2, week 4).

Surface electromyography signals will be recorded using a dedicated system (DELSYS, Trigno™, Wireless Biofeedback System, USA) from four lower limb target muscles (rectus femoris, biceps femoris, tibialis anterior, and gastrocnemius) to evaluate the effects of different intervention methods. Assessments will be conducted at baseline (T0), mid-intervention (T1, 2 weeks), post-intervention (T2, 4 weeks).

MEP assessment will be performed using a transcranial magnetic therapy instrument (Wuhan Iridium, N50000) and a biconical coil (VCZ001). DELSYS surface EMG electrodes will be placed on the affected tibialis anterior muscle (sampling frequency 2000 Hz), and the “hot spot” of the lower limb M1 zone will be localized based on the 10–20 EEG system (with AP direction current stimulation). The rMT and aMT will be measured, then stimulation will be delivered with 130% rMT intensity (interval ≥ 10s). A minimum of 15 valid MEP trials will be acquired at 130% rMT to record amplitude and latency. If recruitment curves (RC) are assessed, stimulus intensities will span 80–150% rMT (5 trials per 10% step) to derive RC parameters including slope. Muscle rest state (baseline RMS < 50 μV) and waveform stability (biphasic morphology, latency variability <10%) will be continuously monitored to ensure data quality. Assessments will be conducted at baseline (T0), mid-intervention (T1, 2 weeks), post-intervention (T2, 4 weeks).

Statistical analysis

Analysis of clinical data

The statistical analyses will be performed using SPSS 22.0 software. Demographic and baseline characteristics will be compared by one-way analysis of variance (ANOVA) or Fisher’s test. Prior to formal analysis, the normality of data distribution will be assessed using the Shapiro-Wilk test. A mixed-design ANOVA was selected as the primary analytical framework because it robustly accommodates both the between-subject factor (the three parallel intervention groups) and the within-subject factor (repeated measures across four time points: T0, T1, T2, T3). This approach allows for the simultaneous evaluation of main effects and the crucial Group × Time interactions, which are essential for determining whether the treatment trajectories differ among the groups.

The significance level will be initially set at p < 0.05. If any significant Group × Time interaction effects are found, pairwise comparisons will be performed using paired t-tests, and the changes from baseline will be compared with a Bonferroni-corrected threshold of 0.017 (for 3 comparisons) to control Type I error in post-hoc testing of the primary outcome.

Furthermore, to control the overall false positive rate resulting from the evaluation of multiple secondary clinical outcomes (e.g., PASS, MBI, gait parameters) and high-dimensional neurophysiological analyses (e.g., fNIRS channels, EEG bands, MEP indices), the False Discovery Rate (FDR) correction (via the Benjamini-Hochberg procedure) will be applied to the resulting p-values across these secondary analyses.

Analysis of neurophysiological and biomechanical data

The raw fNIRS data will be processed and analyzed using the Homer software package. The preprocessing pipeline begins with signal quality inspection and channel exclusion, in which channels with low signal-to-noise ratio or signal loss are flagged and excluded from further analysis. The raw light intensity data are then converted into optical density (OD) units, followed by motion artifact correction using algorithms such as spline interpolation or principal component analysis (PCA). A band-pass filter (0.01–0.2 Hz) is applied to remove low-frequency drift and high-frequency physiological noise (e.g., cardiac and respiratory cycles). Subsequently, the modified Beer–Lambert law is employed to derive time-series changes in oxygenated hemoglobin (HbO₂) and deoxygenated hemoglobin (HbR) concentrations.

Following preprocessing, regional brain activity will be assessed using the fractional amplitude of low-frequency fluctuations (fALFF) to quantify spontaneous neural activity at each channel. To evaluate functional integration within motor-related cortical networks, functional connectivity (FC) will be analyzed based on resting-state HbO₂ signals. Pearson correlation coefficients (r) will be computed between all pairs of channels and then converted to Z-scores via Fisher’s Z transformation to meet parametric assumptions. The resulting Z-scores will be averaged within and between four predefined regions of interest (ROIs)—PMC, SMA, DLPFC, and SMC—to generate ROI-level FC indicators. Between-group differences in FC will be tested using independent samples t-tests, with statistical significance set at p < 0.05.

To minimize the impact of noise and artifacts on data quality, all raw signals will be preprocessed using the EEGLAB toolbox in the MATLAB environment. The preprocessing pipeline includes band-pass filtering and independent component analysis (ICA) to effectively identify and remove physiological artifacts such as ocular and cardiac interference. Following this, a continuous 2-minute high-quality data segment free of significant artifacts will be extracted from each recording for subsequent spectral analysis.

This study focuses on alpha-band (8–13 Hz) activity in brain regions associated with motor function. Based on fast Fourier transform (FFT), the absolute power values within the alpha frequency band will be calculated for the following electrode sites: C3, C4, FC1, FC2, FCz, FC3, and FC4. These values will serve as core EEG metrics. Finally, correlations between the absolute alpha power in these regions and scores from clinical scales will be analyzed to explore the relationship between resting-state brain activity patterns and behavioral performance.

The efficacy of different intervention methods on lower limb muscle functional recovery will be evaluated by analyzing two key sEMG parameters: the Root Mean Square (RMS) and the Median Frequency (MF). In the acquired signals, the Root Mean Square (RMS), as a reliable parameter in the time-domain analysis, reflects the overall level of electrical activity of motor units during muscle contraction. Its value is associated with the recruitment of motor units and the synchronization of discharges, determines the variation in EMG amplitude, and is often used as a quantitative indicator of the force/torque generated by the muscle. On the other hand, the Median Frequency (MF) is a key parameter in the frequency-domain analysis, referring to the frequency value that bisects the area of the power spectrum of the EMG signal. It reflects the median level of muscle fiber discharge rates and is commonly used to assess local muscle fatigue and the mobilization of different muscle fiber types.

Electrodes will be positioned per SENIAM guidelines [30]: the active electrode centered 3 cm superior to the midpoint between the medial malleolus and tibial tuberosity over the tibialis anterior muscle belly, and a ground electrode on the medial malleolus. The biconical coil will be held tangential to the scalp, handle posteriorly directed at a 45° angle to the midline [31].

The “hot spot” (optimal scalp position for eliciting TA MEPs) will be determined using the 10–20 EEG system as a guide and confirmed by eliciting consistent MEPs with anterior-posterior directed current.

The resting motor threshold (rMT) is defined as the minimum stimulus intensity required to elicit motor-evoked potentials (MEPs) with a peak-to-peak amplitude ≥50 μV in at least 5 out of 10 consecutive stimuli. This is determined using a staircase procedure, starting at 30% of the maximum stimulator output and increasing in 1% increments [32]. The active motor threshold (aMT) is determined using the same criteria during voluntary muscle contraction. All procedures are performed under continuous electromyography (EMG) monitoring to ensure the tibialis anterior muscle remains at rest (root mean square [RMS] value <50 μV during baseline periods), with participants maintaining muscle relaxation aided by visual feedback.

Basic MEP parameters are acquired using a stimulus intensity of 130% rMT, with an inter-stimulus interval ≥10 seconds [33]. A minimum of 15 valid MEPs are obtained per set. If continuous stimulation is applied, inter-train intervals are dynamically adjusted based on heart rate variability, with a mandatory rest period of ≥1 minute enforced after every 10 stimuli. Optionally, a recruitment curve (RC) protocol may be performed: stimulus intensity is incrementally increased from 80% to 150% rMT in 10% steps, with 5 stimuli delivered at each intensity level (inter-stimulus interval ≥10 seconds) [34].

Safety

Adverse events to be observed in this trial will include: hearing problems, local pain, muscle twitching, joint soreness, muscle fatigue, dizziness, and seizures. Seizures are the most severe acute adverse effect of iTBS, with improper parameter settings (e.g., excessive intensity relative to individual motor thresholds) increasing risk, as noted in safety guidelines. Joint soreness and muscle fatigue are more common in early robot-assisted training due to lower limb movement adaptation. Other side effects are generally mild and temporary, subsiding after short rest.

Our team will be fully prepared with an emergency area in the rehabilitation room to handle incidents. Immediate measures will be taken promptly if adverse events occur, with detailed documentation throughout the 4-week intervention.

Patient and public involvement

Stroke patients in this study will participate as participants. No additional patients or members of the public will be involved in the study design, recruitment, or data analysis phases.

Data management

Collected data will be entered into case report forms (CRF) and synchronized to the electronic database. Participants will be identified by unique numbers to ensure privacy and security. Paper documents will be stored in locked cabinets, and electronic data will be stored in password-protected computers. Original datasets will only be accessible to the principal investigator and relevant co-investigators of this study, and will be used solely for research purposes; all operations will strictly comply with medical data confidentiality regulations. All randomized participants will be included in the final primary outcome analysis in accordance with the intention-to-treat (ITT) principle, regardless of protocol deviations, withdrawal, or loss to follow-up, to minimize bias and preserve the randomization benefits. To ensure transparency and reproducibility, the fully de-identified participant dataset, alongside the detailed statistical analysis plan, will be deposited in a publicly accessible repository (e.g., Zenodo or Figshare) upon study completion and manuscript publication.“

To maximize patient retention and minimize dropout rates during the 6-week study cycle, our research coordinators will maintain regular communication with participants via weekly phone calls or WeChat messages. Follow-up assessments will be scheduled flexibly to accommodate patients’ availability, and transportation assistance will be provided if necessary. For data quality control, all clinical assessments will be double-entered into the electronic database by two independent researchers to ensure accuracy. The principal investigator will conduct regular monitoring and logic checks on the case report forms (CRFs) to promptly identify and resolve any missing, inconsistent, or anomalous data.