Revealing the Neuromechanism of Interhemispheric Balance Restoration in Chronic Stroke Patients through EMG-driven Robot Hand Training: Insights from Dynamic Causal Modeling

Stroke is a common cause of disability, with most stroke survivors suffering from upper limb paralysis. The consequences of upper limb functional impairment can persist for over six months, with only a few survivors (less than 12%) achieving full recovery. Researchers have been dedicated to developing post-stroke motor rehabilitation programs to restore daily living abilities and improve the quality of life for these patients.

In recent years, research on upper limb rehabilitation using robot-assisted devices has garnered significant attention. Robot rehabilitation provides a consistent, intensive, and interactive training experience that can actively engage patients. Meta-analyses show that individuals undergoing robot-assisted training exhibit significant improvements in Fugl-Meyer Assessment (FMA-UE) scores and upper limb functional activities. However, the efficacy of robots targeting wrist and hand functions has been limited in terms of motor control and daily living activities improvements. With the advent of intention-driven robots, these devices have become popular in the past five years, aiding in the improvement of wrist and hand functions. By utilizing electrophysiological signals such as EEG or EMG, individuals can trigger robot-assisted motor tasks through voluntary motor intentions. Clinical studies have found that compared to continuous passive motion control groups, EMG-driven robot hand training can lead to higher FMA-UE scores and wrist and hand function improvements in chronic stroke patients.

However, the precise neurophysiological mechanisms underlying the motor recovery induced by these devices are still limited. This study aims to use Dynamic Causal Modeling (DCM) to analyze the effective connectivity (EC) of the central nervous system, revealing the neural mechanisms of motor recovery in chronic stroke patients through EMG-driven robot hand training.

Research Source and Authors

This research article was published in IEEE Transactions on Neural Systems and Rehabilitation Engineering, Vol. 32, 2024. The authors include Chun-hang Eden Ti, Chengpeng Hu, Kai Yuan, Winnie Chiu-wing Chu, and Raymond Kai-yu Tong, and it was collaboratively written by researchers from the Department of Biomedical Engineering and Department of Imaging and Interventional Radiology at The Chinese University of Hong Kong. Raymond Kai-yu Tong is the corresponding author. The study was partially supported by the Hong Kong Research Grants Council and the General Research Fund.

Research Methods and Process

1. Subjects and Recruitment

This was a single-center, randomized controlled trial that recruited 19 chronic stroke survivors. One participant withdrew for medical reasons, and another was excluded from the analysis due to abnormal fMRI activation patterns. Inclusion criteria included first-time stroke, over six months post-stroke onset, unilateral brain injury, and moderate to severe upper limb dysfunction with Action Research Arm Test (ARAT) scores between 6 to 52. Exclusion criteria involved a family history of epilepsy, alcohol or drug abuse, and other neurological dysfunctions such as spatial neglect, aphasia, and apraxia. All subjects signed informed consent, confirming their understanding of the study procedures and impact.

2. Training Intervention

Each subject completed 20 sessions of combined EMG-driven robot hand training at a local community center. Prior to each session, subjects received 20 minutes of active or sham stimulation. During training, subjects wore the robot hand and sat in front of a computer screen. They were instructed to keep their elbow still and slightly bent at approximately 130 degrees. EMG signals were recorded from the flexor and extensor muscles, and the training tasks involved the robot hand performing gripping and hand opening actions triggered by EMG signals surpassing 10% of the maximum voluntary contraction (MVC).

Training consisted of three modules, each lasting 15 minutes with a 5-minute break in between. During each module, subjects had to trigger robot-assisted gripping and opening actions, repeating these tasks until the module ended.

3. MRI Data Acquisition and Analysis

MRI scans were conducted using a 20-channel head coil to obtain high-resolution anatomical T1 images, resting-state (rs-fMRI), and task-based functional MRI (tb-fMRI) data. Each subject underwent MRI scans before and after the 20 training sessions. During tb-fMRI, subjects alternately gripped a tennis ball with their left or right hand based on on-screen textual prompts.

4. Dynamic Causal Modeling and Functional Connectivity Analysis

Dynamic Causal Modeling (DCM) was used to estimate effective connectivity (EC) between brain regions. We focused on changes in EC between brain regions during task periods before and after training in chronic stroke patients.

Additionally, resting-state functional connectivity (FC) was used to study the interactions within the brain’s motor network, providing task-independent brain reorganization information. By comparing the differences before and after training, we revealed the impact of the training.

Research Results and Discussion

1. Training Effect

Training data indicated that all subjects’ trigger counts increased over the training sessions. Mixed-effects ANOVA analysis showed a significant positive effect of time on ARAT scores, indicating that EMG-robot hand training improved upper limb function.

2. Dynamic Causal Modeling Results

Bayesian model selection identified the model that best explained task-regulated EC and training effects in chronic stroke patients. We found that during training, voluntary motor intentions effectively reduced inhibitory influence from the contralesional M1 to the ipsilesional M1, while increasing functional connectivity of the ipsilesional M1. These results suggest a positive correlation between reduced interhemispheric inhibition and functional improvement, highlighting the role of interhemispheric interaction in motor recovery.

3. Resting-State Functional Connectivity

Functional connectivity analysis showed a significant increase in FC between bilateral M1 post-training, indicating the restoration of interhemispheric balance in resting-state brain activity. Although no significant correlation was found between changes in FC and ARAT improvement, there was a significant association between changes in EC between M1 and motor function improvement.

4. Task-based Functional MRI

Task-based fMRI activation maps showed a general decrease in activation in the contralesional hemisphere and an increase in the ipsilesional hemisphere post-training. This indicates an increased role of the ipsilesional hemisphere and reduced dependence on the contralesional hemisphere during motor tasks.

Research Conclusions and Significance

This study demonstrates that EMG-driven robot hand training can significantly improve upper limb function in stroke patients by reducing interhemispheric inhibition and promoting ipsilesional hemisphere activation. These results highlight the value of EMG-driven robot hand training in restoring motor function and promoting neuronal plasticity in chronic stroke patients. The study provides new insights into the interhemispheric balance restoration induced by EMG-driven robot training and offers valuable references for future stroke rehabilitation research.

This report summarizes the results of an important study in stroke rehabilitation, delving into the neural mechanisms and recovery effects of EMG-driven robot hand training for chronic stroke patients. The study indicates that EMG-driven robot training can significantly enhance upper limb function and promote neural recovery by reducing interhemispheric inhibition, offering valuable perspectives for future stroke rehabilitation research.