A Closed-Loop ta-VNS System Synchronized with BCI-Based Motor Training for Post-Stroke Upper Limb Rehabilitation

Summary

This study demonstrates a standardized protocol for a closed-loop ta-VNS system synchronized with BCI-based motor training. Furthermore, it compares the efficacy of the closed-loop ta-VNS system synchronized with BCI-based motor training with that of sham ta-VNS, providing significant evidence for its clinical application in stroke rehabilitation.

Abstract

Transcutaneous auricular vagus nerve stimulation (ta-VNS) involves applying electrical stimulation via electrodes to the auricular concha. This activates vagal afferent fibers, initiating an ascending pathway from the periphery to the brainstem, which ultimately stimulates central vagal projections and promotes neural plasticity. Previous studies have demonstrated that combining ta-VNS with motor training offers synergistic benefits for motor recovery after stroke. However, these combined approaches typically employ open-loop stimulation with fixed parameters, lacking real-time closed-loop responsiveness to dynamic neural activity. To address this limitation, we developed a novel closed-loop ta-VNS system synchronized with electroencephalography (EEG)-triggered brain-computer interface (BCI) motor training. This system was designed to enhance corticospinal coupling and promote synaptic plasticity. We established a standardized protocol for applying this closed-loop ta-VNS system synchronized with BCI-based motor training in stroke patients. Using EEG-based functional assessment, we compared the effects of the closed-loop ta-VNS system synchronized with BCI-based motor training to those of sham ta-VNS synchronized with BCI-based motor training. This work provides the methodological and theoretical groundwork for the clinical application of this approach in stroke rehabilitation.

Introduction

Transcutaneous auricular vagus nerve stimulation (ta-VNS) is a non-invasive neuromodulation technique used to ameliorate post-stroke motor impairments¹. This method entails stimulating the auricular concha with electrodes to activate the auricular branch of the vagus nerve. Such stimulation initiates the nucleus tractus solitarius (NTS)-locus coeruleus (LC) pathway², which ascends to central vagal projections in the motor cortex, frontal lobe, thalamus, and cerebellum. This process enhances neural plasticity and network reorganization³. Compared with traditional invasive vagus nerve stimulation, ta-VNS preserves therapeutic efficacy while mitigating surgical risks and complication rates⁴. These advantages establish it as a prominent research focus within rehabilitation science^5,6,7.

Clinical evidence supports the efficacy of combined ta-VNS and motor training in enhancing post-stroke motor function. Chang et al.⁸ randomized 36 stroke patients with upper limb impairment into two groups: active ta-VNS paired with robotic motor training versus sham ta-VNS with robotic training. The active ta-VNS group demonstrated significantly greater improvement in upper limb spasticity and electromyogram activity than the sham group, indicating enhanced motor recovery. Complementing these findings, Zheng et al.⁹ employed motor evoked potentials to demonstrate that stroke patients receiving ta-VNS synchronized with robotic training exhibited increased primary motor cortex (M1) excitability (amplitude) and improved conduction efficiency (reduced latency) compared to those receiving robotic training alone. This suggests that combined ta-VNS promotes functional cortical reorganization to enhance motor recovery.

However, existing combined interventions typically employ fixed-parameter open-loop stimulation, lacking dynamic closed-loop responsiveness to neural states. To address this limitation, we designed a novel electroencephalography (EEG)-triggered closed-loop system that integrates ta-VNS with brain-computer interface (BCI)-based motor training (Figure 1), comprising four components: a synchronization system, an EEG acquisition unit, a pneumatic glove for hand rehabilitation, and a ta-VNS device. The EEG acquisition unit captures motor-intent-related brain engagement by decoding EEG signals, utilizing real-time engagement characteristics as a biomarker to trigger the pneumatic glove and ta-VNS device¹⁰. Both devices connect to the synchronization host via wired serial ports. When brain engagement reaches a predefined threshold, the devices activate synchronously, enabling closed-loop ta-VNS intervention precisely matched to motor training. This synchronization is hypothesized to enhance synaptic efficacy between the motor cortex and spinal motor neurons, as prior research suggests that ta-VNS paired with motor training can modulate cortical excitability and synaptic function¹¹. Conversely, ta-VNS is not triggered if the brain engagement value is below the threshold. It is anticipated that this system may strengthen corticospinal coupling and promote synaptic plasticity, given that closed-loop ta-VNS synchronized with movement has shown potential in enhancing motor recovery outcomes¹², thereby holding significant promise for neurorehabilitation studies.

The system is primarily suitable for stroke patients experiencing upper limb motor dysfunction. Eligibility requires patients to have sufficient cognitive capacity for task compliance as well as stable vital signs. Key contraindications include the presence of implanted electronic devices, skin lesions or abnormalities at the auricular stimulation site, diagnosed cervical vagus nerve pathology, or specific arrhythmias. Operational constraints involve the necessity for trained personnel to administer the intervention and continuous monitoring during initial sessions to ensure stimulation tolerance and procedural safety.

Given the novelty of this integrated system, a standardized application protocol is paramount to ensure consistent administration and replication of the method. Therefore, the primary aim of this methodological article is to present a standardized protocol for the closed-loop ta-VNS system synchronized with BCI-based motor training. Furthermore, using EEG-based functional assessment in a representative case, we aimed to preliminarily explore the neuromodulatory effects of the closed-loop ta-VNS system synchronized with BCI-based motor training compared to sham ta-VNS synchronized with BCI-based motor training. This work establishes the methodological groundwork and demonstrates the initial feasibility of applying this synchronized intervention in a clinical setting.

Protocol

The study was approved by the Medical Ethics Association of the Seventh Affiliated Hospital of Sun Yat-sen University (Shenzhen, China) (approval number: KY-2025-318-01) on April 30, 2025. The trial was registered in the Chinese Clinical Trials Registry (registration number: ChiCTR2500104317) on June 16, 2025. The subject signed the informed consent form.

1. Recruitment

1. Inclusion criteria
   1. Recruit subjects with stroke diagnosis that meets the criteria (encompassing cerebral hemorrhage and infarction) established by the Fourth National Cerebral Vascular Disease Academic Conference.
   2. Recruit subjects with first-episode, confirmed cerebral hemorrhage or cerebral infarction by cranial CT or MRI, 2 weeks ≤ disease duration ≤ 6 months.
   3. Recruit subjects aged 18-80 years.
   4. Recruit subjects with Mini-Mental State Examination (MMSE) ≥21.
   5. Recruit subjects with post-stroke motor dysfunction of the upper limbs and hands.
   6. Recruit subjects with stable vital signs.
   7. Recruit subjects with voluntarily signed informed consent.
2. Exclusion criteria
   1. Exclude subjects with a history of bilateral/unilateral vagus nerve injury or surgery.
   2. Exclude subjects with pacemakers or neurostimulators.
   3. Exclude subjects with cognitive dysfunction.
   4. Exclude subjects with infection, ulcers, or scarring at stimulation sites.
   5. Exclude subjects with heart rate <60 bpm.
   6. Exclude women during pregnancy and lactation.
   7. Exclude subjects with severe visual or auditory impairment that prevents perception of the experimental cues.

2. Basic information collection

1. Collect and record the basic information of the subject, including age, gender, disease duration, stroke type, hemiplegic side, hand Brunnstrom stage, and Fugl-Meyer Assessment of Upper Extremity Motor Function (FMA-UE).
2. Implement a within-subject crossover design, in which the subject completes both intervention conditions: closed-loop ta-VNS synchronized with BCI-based motor training and sham ta-VNS synchronized with BCI-based motor training, separated by a 6-day washout period.

3. The closed-loop ta-VNS system synchronized with BCI-based motor training

1. Subject preparation
   1. Before the formal intervention, all participants received a comprehensive explanation of the experimental procedure and completed a brief practice session to ensure they understood the motor imagery task and could perceive all visual and auditory cues clearly.
2. Pre-test preparation
   1. Prepare a 3% NaCl electrolyte solution by adding 4.5 g of sodium chloride to 150 mL of pure water, then shake vigorously until fully dissolved.
   2. Prepare the electrolyte solution in a container, then immerse 32 cotton pads and soak at room temperature for 3–5 min.
   3. Squeeze excess water from the cotton pads.
   4. Insert the cotton pad holder into the electrode base along the card slot.
   5. Pinch the electrode base while rotating clockwise until a click is heard.
   6. Sequentially install 30 EEG electrodes and 2 EOG electrodes, positioning the reference electrode at CPz and the ground electrode at FPz.
   7. Wear the EEG cap and locate the Cz point using the international 10-10 system.
      1. Acquire the EEG signals using a 32-channel saline-based wireless multi-channel EEG acquisition and analysis system. According to the international 10-10 system, position the reference electrode at A1 (left earlobe), and place the ground electrode at AFz (frontal midline). The setup included 30 recording electrodes and 2 EOG electrodes.
      2. Set the amplifier specifications as follows: input impedance of 1 GΩ, input noise < 0.4 µV, and a common-mode rejection ratio of -120 dB. Maintain the electrode-skin interface impedance below 5 kΩ during recordings, and sample the EEG signals at 250 Hz.
   8. Position the Cz electrode at the vertex—the intersection of the interauricular line (connecting both earlobes) and the nasio-bregmatic line (extending from the nasal tip to the glabella)
   9. Install the amplifier by gently inserting it into the EEG cap plug and securing it to the cap.
   10. Connect the ear clip stimulation electrode to the ta-VNS host.
   11. Remove the alcohol swab and cleanse the subject's auricular concha.
   12. Apply the stimulation electrode to the auricular concha.
   13. Don the pneumatic glove and adjust for a comfortable fit to prevent over-tightening or loosening from compromising training outcomes.
3. Software Implementation
   1. Power on the ta-VNS host device by pressing and holding the green button located on its right side.
   2. Connect the ta-VNS stimulator to the BCI-based motor training system via the USB interface.
      NOTE: The ta-VNS stimulator was connected to a USB-to-TTL module, by which it established serial communication with the PC via a USB cable to synchronize with the BCI-based motor training system. The serial port was configured with a baud rate of 38400 bps, 1 stop bit, and no parity.
   3. Launch the brain function testing software (BCIServer-V1.0.0).
   4. Click the Start Device button in the upper right corner, followed by the 'Impedance Detection' button on the left side, ensuring electrode impedance remains below 5 kΩ. Do not proceed with the run until all electrodes show an impedance of ≤ 5 kΩ and the indicator is green/yellow.
      1. If high impedance (red indicator) is detected, adjust by: (1) gently twisting the electrode base, (2) reapplying saline to the scalp, and (3) sequentially optimizing each electrode until the indicator turns green/yellow.
      2. For persistently high impedance at any electrode, remove the cotton pads tray, re-saturate it with saline, reload it into the electrode base, and twist gently until impedance reduction is achieved.
   5. Launch the closed-loop ta-VNS system synchronized with BCI-based motor training software (HandGame-V1.0.0) and enter the subject's demographic information: name, gender, and date of birth.
   6. Access the training settings and configure: contraction speed: high, training site: left hand, motor imagery difficulty: low, and training duration: 20 min.
      NOTE: The contraction speed parameter defines the angular velocity of the pneumatic glove's finger joints. The 'high' setting corresponds to approximately 68°/s, while the 'low' setting corresponds to 30°/s. The selection should be adjusted based on the patient's muscle tone; for example, 'high' speed is unsuitable for patients with spasticity. For the present case, a 'high' setting was selected as the patient (Brunnstrom Stage I, flaccid stage) exhibited low muscle tone. The motor imagery difficulty is defined by the minimum brain engagement threshold required to activate the BCI-based motor training system within the 5 s task period. The threshold is set to 50 for the 'low' level, 60 for the 'medium' level, and 80 for the 'high' level. For this patient's initial training session, the 'low' difficulty level was selected. Subsequent adjustments to the difficulty level are made based on the patient's performance, primarily considering the number of successful motor imagery triggers achieved during the training session.
   7. Adjust the current intensity using a default start of 1 mA and an allowable range of 0–8.0 mA, Increase the current in increments of 0.2 mA until the patient reports a strong but comfortable tingling sensation without any pain or discomfort. Immediately cease stimulation if the patient experiences pain or any discomfort (e.g., burning, muscle twitching, dizziness, palpitations, or dyspnea).
      NOTE: The ta-VNS preset parameters are fixed at: frequency: 20 Hz, pulse width: 300 µs, and stimulation duration: 5 s.
   8. Initiate training.
   9. Execute cyclic motor imagery training.
   10. Proceed with individual motor imagery training tasks sequentially in a structured cycle, as illustrated in Figure 2:
   11. Rest phase (5 s): Record the baseline EEG data during this phase for subsequent calibration and analysis.
   12. Motor imagery and EEG acquisition (5 s): 
       1. Perform visual/auditory cues induced hand grasping imagery training with synchronous EEG acquisition. The brain engagement is a dimensionless metric derived from the EEG signal to quantify attentional focus during motor imagery. 
       2. Sample the EEG signals at 250 Hz and process them by an 8–30 Hz band-pass filter to retain the frequencies of interest. Apply a sliding window (length: 4 s, step: 1 s).
       3. Compute the power spectral density (PSD) via Fast Fourier Transform (FFT) for each window. Calculate the energy of the alpha (8-13 Hz) and beta (14-30 Hz) bands (E_α and E_β), and derive their ratio R = E_β / E_α. Then, the ratio R was normalized to a value between 0 and 100. Subsequently calculate the brain engagement value using the formula: brain engagement = 100 × (R - min) / (max - min), where the empirical constants max = 3.50 and min = 0.35 were determined from a prior calibration dataset¹³.
   13. Feedback and stimulation phase (5 s, contingent on success): If the brain engagement value was greater than or equal to the predefined threshold of 50 within the task period, initiate the pneumatic glove movement and verum ta-VNS simultaneously for a duration of 5 s, constituting the synchronized BCI-ta-VNS intervention. Otherwise, neither pneumatic feedback nor ta-VNS occurred, and auditory cues were provided to encourage patient to persist.
   14. Terminate the closed-loop ta-VNS system synchronized with BCI-based motor training software post-training.

4. The sham ta-VNS synchronized with BCI-based motor training

1. Follow steps 3.1 to 3.3 identically, but in step 3.3.7, reduce the ta-VNS current intensity for BCI-based motor training to zero. All parameters except ta-VNS current remain identical to those described in section 3. The training duration was 20 min.

5. Offline EEG data processing and analysis

NOTE: Offline EEG data processing and analysis were conducted using MATLAB R2021a in conjunction with the EEGLAB (v2024.2) and ERPLAB (v12.00) toolboxes. The processing pipeline comprised two main stages: preprocessing and analytical analysis.

1. Data preprocessing
   1. Perform data import and channel configuration:
      1. Convert the acquired continuous EEG data in .csv format to .mat format and import them into EEGLAB. Set the sampling rate to 250 Hz.
      2. Load the channel location file corresponding to the EEG cap, and plot the electrode positions to verify their coordinates. Remove the unused EOG electrodes.
   2. Filtering: Apply a bidirectional FIR band-pass filter (0.1–40 Hz) to the task-state EEG data from both interventions to remove baseline drift and suppress high-frequency noise.
   3. Perform epoching.
      1. Segment the data based on event markers. Marker 1 indicated the start of the motor imagery task. Extract epochs from -2000 ms to +5000 ms relative to Marker 1, but select only those trials that were successfully triggered (i.e., those that led to a 'success' outcome, as the primary analysis focused on the brain state during effective trials) for subsequent analysis.
      2. Apply a baseline correction using the interval from -2000 ms to -1000 ms pre-stimulus. Export an EVENTLIST to document the number and timing of markers.
   4. Remove artifact.
      1. Inspect the epoched data visually to identify and reject anomalous segments and bad channels. Interpolate identified bad channels. Subsequently, subject the data to Independent Component Analysis (ICA).
      2. Use the ICLabel toolbox to automatically classify independent components, and remove those identified as ocular or muscular artifacts with a probability exceeding 90%. This automatic rejection was followed by manual inspection to confirm the removal of artifact-related components.
   5. Re-referencing: Re-reference the data to the average of all channels.
2. Data analysis
   1. Time-frequency analysis: Apply a wavelet transform to the task-state EEG data for time-frequency analysis. Use the period from -2000 ms to -1000 ms for baseline correction. Analyze the event-related desynchronization (ERD) phenomenon at electrode C4 within a time window of 0 ms to 5000 ms.
   2. Power spectral and topographic map analysis: Compute the average power in the alpha and beta bands (8–30 Hz) using FFT over a time window of 500 ms to 3000 ms post-cue. Plot topographic maps of the average band power to visualize the spatial distribution of sensorimotor cortex activation.

Results

A 37-year-old male stroke patient successively underwent two interventions: a closed-loop ta-VNS system synchronized with BCI-based motor training, and a sham ta-VNS system synchronized with BCI-based motor training. Baseline characteristics are presented in Table 1. System Intervention Parameters are presented in Table 2.

For the stroke patient, the number of successful motor imagery trials was recorded for both intervention protocols. During the 20 min closed-loop ta-VNS synchronized with BCI-based motor training, 54 successful triggers were recorded, each initiating 5 s of simultaneous pneumatic glove movement and verum ta-VNS, yielding a total effective stimulation time of 270 s (calculated as the number of successful triggers × 5 s). In the 20 min, sham ta-VNS synchronized with BCI-based motor training, 43 successful triggers activated the pneumatic glove for 5 s per trigger with concurrent sham ta-VNS (0 mA), resulting in a total effective glove activation time of 215 s.

Prior to data acquisition, electrode impedance was verified to ensure signal quality. A screenshot of the system's software interface, confirming that all electrode impedances were at or below the 5 kΩ threshold before the task began, is provided in Figure 3.

The EEG signals were acquired using an amplifier with an input impedance of 1 GΩ, an input noise of <0.4 µV, and a common-mode rejection ratio of -120 dB. The signals were sampled at 250 Hz. For the online calculation of the brain engagement index, the raw EEG signal was processed in real-time by the BCI software with an 8-30 Hz band-pass filter to isolate the frequencies of interest for motor imagery.

For offline analysis, the acquired continuous EEG data were exported and processed in MATLAB R2021a. The data were first band-pass filtered between 0.1 and 40 Hz to remove baseline drift and suppress high-frequency noise.

Wavelet-based band-pass filtering was applied to task-state EEG data from both interventions. Analysis of ERD at electrode C4 (sensorimotor cortex of the lesioned hemisphere) revealed stronger ERD modulation during closed-loop ta-VNS synchronized with BCI-based motor training than during sham ta-VNS synchronized with BCI-based motor training (Figure 4).

Topographic mapping of average alpha/beta band (8–30 Hz) power revealed greater contralateral sensorimotor power reduction during closed-loop ta-VNS synchronized with BCI-based motor training than during sham ta-VNS synchronized with BCI-based motor training (Figure 5). This aligns with enhanced ERD patterns observed under the closed-loop ta-VNS system synchronized with BCI-based motor training.

Figure 1: Schematic of the closed-loop ta-VNS system synchronized with the BCI-based motor training system. Abbreviations: ta-VNS, transcutaneous auricular vagus nerve stimulation; NTS, nucleus tractus solitarius; LC, locus coeruleus; SM1, primary sensorimotor area. Please click here to view a larger version of this figure.

Figure 2: Session timeline for a single trial in the intervention. The diagram illustrates the sequential stages of a single trial: (1) Rest Phase (5 s): Baseline EEG data were recorded during this phase for subsequent calibration and analysis. (2) Motor Imagery & EEG Processing (5 s): The patient performs hand grasping motor imagery. EEG is acquired and processed in real-time to calculate the brain engagement index. (3) Threshold Check & Feedback (Contingent, 5 s): At the end of the 5 s imagery period, if the brain engagement value meets or exceeds the threshold (≥50), it triggers the simultaneous activation of the pneumatic glove and ta-VNS stimulator for 5 s. If the threshold is not met, an auditory encouragement cue is played, and no feedback or stimulation is delivered. Please click here to view a larger version of this figure.

Figure 3: Pre-task electrode impedance verification. Please click here to view a larger version of this figure.

Figure 4: Time-frequency maps at electrode C4 under two interventions. (A) Closed-loop ta-VNS system synchronized with BCI-based motor training. (B) Sham ta-VNS synchronized with BCI-based motor training. Please click here to view a larger version of this figure.

Figure 5: Average power topography maps of alpha and beta frequency bands under different interventions. (A) Closed-loop ta-VNS system synchronized with BCI-based motor training. (B) BCI-based motor training. Please click here to view a larger version of this figure.

Variables	Baseline information
Age (years)	37
Gender	Male
Duration (months)	4
Type of stroke	Cerebral infarction
Side of hemiparesis	Left
Hand Brunnstrom staging	I
FMA-UE	12

Table 1: Baseline information of the subject. Abbreviation: FMA-UE, Fugl-Meyer Assessment of Upper Extremity Motor Function.

Parameter Category	Specific Parameter	Set Value
Software Settings	Training Duration	20 min
	Contraction Speed	High (~68 °/s)
	Training Site	Left Hand
	Motor Imagery Difficulty	Low
Threshold Settings	Brain Engagement Threshold	50
Glove Parameters	Size	M
	Single Grasp Duration	5 s
ta-VNS Parameters	Frequency	20 Hz
	Pulse Width	300 μs
	Single Stimulation Duration	5 s
	Current Intensity	3 mA (Closed-loop ta-VNS synchronized with BCI-based motor training)
		0 mA (Sham ta-VNS synchronized with BCI-based motor training)

Table 2: System intervention parameters. Abbreviations: ta-VNS, transcutaneous auricular vagus nerve stimulation; BCI, brain-computer interface.

Discussion

Prior research confirms that synchronizing ta-VNS with traditional motor training optimizes therapeutic efficacy for post-stroke motor recovery, likely through precise neuromodulation that enhances neural reorganization^8,14. Building on this evidence, we established a standardized experimental protocol and employed EEG biomarkers to monitor sensorimotor cortical activation patterns, particularly reflected in alpha and beta band topographic distributions during the intervention.

The implementation and experimental integrity of the closed-loop system rely on several critical procedures: First, precise EEG electrode placement according to the international 10-10 system (steps 3.2.7–3.2.8) with impedance verification (step 3.3.4) is essential. Optimal positioning of 30 EEG and 2 EOG electrodes determines signal quality¹⁵, which governs motor imagery decoding accuracy and facilitates synchronized ta-VNS intervention. Second, pre-intervention setup requires individualized titration of ta-VNS current intensity (step 3.3.7) to ensure both safety and efficacy. Third, hardware synchronization via a USB interface (step 3.3.2) coordinates the ta-VNS stimulator and BCI training device. Motor imagery recognition triggers simultaneous interventions through dedicated interfaces: TTL-controlled ta-VNS feedback and serial-controlled BCI activation, establishing precise temporal coupling. Regarding operational efficiency, the workflow requires approximately 20 min of trained operator time for initial setup (EEG cap placement, impedance verification, and current titration), while the subsequent 20 min intervention runs autonomously, shifting the clinician's role to supervisory oversight and thereby enhancing workflow efficiency. This structured protocol balances technical precision with the benefit of automated, synchronized intervention delivery.

During system validation, we identified three technical bottlenecks with targeted solutions: First, to resolve EEG signal acquisition latency, we implemented a hybrid network configuration where the EEG amplifier communicated with a local router via 5 GHz Wi-Fi, and the host PC was connected to the same router via a wired Ethernet connection. Second, for suboptimal EEG classification accuracy, we implemented optimal time-frequency subband selection, Riemannian covariance feature extraction, and feature vector fusion for SVM classification¹⁵.

In this study, the subject underwent both the closed-loop ta-VNS system synchronized with BCI-based motor training and sham ta-VNS synchronized with BCI-based motor training. Time-frequency analysis of task-state EEG observed a more pronounced ERD phenomenon in sensorimotor areas during the closed-loop ta-VNS system synchronized with BCI-based motor training compared to sham ta-VNS synchronized with BCI-based motor training. As ERD is considered to reflect the degree of sensorimotor neuron activation and functional reorganization, with greater ERD suggesting higher cortical activation^16,17,18,19, these findings indicate that the closed-loop ta-VNS system synchronized with BCI-based motor training appeared to enhance sensorimotor area activation compared to sham ta-VNS synchronized with BCI-based motor training. This observation points to its potential applicability in patients with post-stroke motor dysfunction. Furthermore, analysis of task-state average power in the alpha and beta frequency bands detected significantly lower power in the contralateral motor area during ta-VNS intervention compared to sham ta-VNS synchronized with BCI-based motor training, aligning with the observed ERD changes. This effect could potentially stem from ta-VNS-induced neural remodeling. Supporting this, Kang et al.²⁰ observed significant differences in alpha and delta band power across multiple EEG channels before and after ta-VNS intervention, suggesting modulated neural plasticity through potential modulation of resting-state network nodes. It is important to emphasize that the observed effects represent acute neuromodulatory responses during the intervention session. Future studies with long-term follow-up assessments are required to determine whether these acute changes can translate into lasting neural plasticity and functional reorganization. Collectively, the findings from this case demonstration suggest that the closed-loop ta-VNS system synchronized with BCI-based motor training may promote sensorimotor activation and reorganization in stroke patients, providing preliminary support for further investigation of its clinical application in stroke rehabilitation.

However, since the closed-loop ta-VNS system, synchronized with BCI-based motor training, is triggered by the motor imagery paradigm, sensory-impaired stroke patients exhibit significantly impaired motor imagery capacity in affected limbs^21,22. This compromises system responsiveness and limits clinical applicability. Future optimizations should therefore incorporate alternative BCI paradigms to expand patient eligibility. For instance, visual P300 or steady-state visually evoked potentials (SSVEPs) represent two highly viable candidates²³. The P300 paradigm relies on the brain's endogenous response to an attended, rare stimulus within a sequence of common stimuli, which could be mapped to specific motor commands. Conversely, the SSVEP paradigm utilizes the brain's entrainment to externally presented rhythmic visual stimuli, where gazing at an icon flickering at a specific frequency generates a corresponding oscillatory EEG signal that can be classified to trigger an action. The implementation of these paradigms would primarily require modifications to the visual feedback interface and the upstream EEG classification algorithm, while the core closed-loop synchronization architecture between the BCI and the ta-VNS stimulator would remain applicable. Integrating such alternative paradigms could potentially extend the benefits of this closed-loop intervention to a broader stroke population, including those with severe motor imagery deficits.

This study has some limitations. First, this study presented a standardized protocol for a closed-loop ta-VNS system synchronized with BCI-based motor training and described its acute neuromodulatory effects through EEG-based assessments. However, the current protocol was implemented and evaluated in only one stroke patient. Due to the single-case design, the potential influences of key confounding variables—such as stroke type (ischemic or hemorrhagic), lesion location and volume, and detailed treatment history—on the intervention outcomes could not be assessed. Therefore, further verification of its efficacy should be conducted with a larger sample size in the future. Second, this study primarily investigated the immediate effects within a single session; the long-term therapeutic benefits and the impact of a full intervention course require further investigation. Moreover, this system relies on a motor imagery paradigm, which may not be suitable for stroke patients with cognitive dysfunction. BCI under other paradigms should be explored in future clinical practice to extend the applicability of this approach.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Alcohol pads	Steady Medical Group Co., Ltd.		Skin disinfection
Electrolyte bottle	Xi'an Zhentai Intelligent Technology Co., Ltd.		Capacity: 150 mL, used for storing electrolyte solution
Fugl-Meyer Assessment of Upper Extremity Motor Function	The standardized assessment developed by Swedish scholar Fugl-Meyer	FMA-UE	It is mainly used for clinical assessment of the reflex activities, motor control and muscle strength of the upper limbs of stroke patients.
Hand function rehabilitation training and evaluation system	Xi'an Zhentai Intelligent Technology Co., Ltd.	ZhenTec-H1	The training system controlled the rehabilitation training gloves in a pneumatic way, and helped the patients gradually realize the rehabilitation training through multi-mode training Rehabilitation training for hand muscles. The training is highly interesting, highly immersive, and has an evaluation function, which can evaluate the hand function of patients Recovery status.
Heart rate monitors	Shenzhen Huiying Electronic Technology Co., Ltd.	HY-HRV-001	Monitor heart rate
Lingke cotton	Xi'an Zhentai Intelligent Technology Co., Ltd.		Used to absorb salt water and transmit signals
Mini-Mental State Examination	The standardized assessment developed by Marshal F. Folstein et al.	MMSE	Rapid screening of adult cognitive impairment
Pure water	China Resources Tianbo Beverage (China) Co., Ltd.	Yibao Purified Water	150 mL is used for preparing the electrolyte solution
Saline electrode wireless multi-channel EEG acquisition and analysis system	Xi'an Zhentai Intelligent Technology Co., Ltd.	ZhenTec-NT1-G	The EEG acquisition equipment was a 32-lead saline electrode wireless multi-channel EEG acquisition and analysis system. According to the international 10-10 system, there were 30 acquisition electrodes and 2 ophthalmic electrodes. The input impedance of the device was 1 GΩ, the input noise was less than 0.4 μV, and the common mode rejection ratio was -120 dB. The electrode wear impedance was below 5 kΩ, and the EEG sampling rate was 500Hz.
Sodium chloride	Xi'an Zhentai Intelligent Technology Co., Ltd.		4.5 g, used for preparing the electrolyte solution
Transcutaneous auricular vagus nerve stimulator	Shenzhen Huiying Electronic Technology Co., Ltd.	HY-TVNS-001	The product is mainly composed of non-invasive vagus nerve stimulator host, non-invasive vagus nerve stimulator surface electrode, charging cable. The electrodes of the stimulator can be controlled by software, and the stimulation parameters include frequency, pulse width, time, mode, etc.
Wetting bottle	Xi'an Zhentai Intelligent Technology Co., Ltd.		Capacity: 250 mL, used for soaking lint cotton