Related Experiment Video
Updated: Jul 31, 2025
09:42
Author Spotlight: Using Motor Imagery Brain-Computer Interface to Improve Motor and Cognitive Function in Stroke Patients
Published on: September 1, 2023
1.3K
A sequential learning model with GNN for EEG-EMG-based stroke rehabilitation BCI
Haoyang Li1, Hongfei Ji1, Jian Yu1
1Translational Research Center, Shanghai Yangzhi Rehabilitation Hospital (Shanghai Sunshine Rehabilitation Center), School of Electronic and Information Engineering, Tongji University, Shanghai, China.
Frontiers in Neuroscience
|May 4, 2023
Summary
This study introduces a novel sequential learning model for brain-computer interfaces (BCIs) to enhance motor rehabilitation in stroke patients. The new model significantly improves the accuracy of predicting complex movements using EEG and EMG signals.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Current brain-computer interfaces (BCIs) offer limited precision for complex motor tasks in stroke rehabilitation due to insufficient EEG signal analysis.
- Accurate decoding of motor intentions is crucial for effective neurofeedback and improved patient recovery.
Purpose of the Study:
- To develop an advanced BCI model for precise motor intention decoding in stroke patients.
- To enhance neurofeedback accuracy for motor rehabilitation by analyzing sequential movement features.
Main Methods:
- A sequential learning model utilizing a Graph Isomorphic Network (GIN) was developed.
- The model processes sequential graph-structured data from EEG and EMG signals, dividing movements into sub-actions for separate prediction.
- Time-based ensemble learning was employed to improve prediction accuracy and movement execution quality scores.
Main Results:
- The proposed model achieved 88.89% classification accuracy on a synchronized EEG-EMG dataset for push and pull movements.
- This significantly outperformed the benchmark method, which achieved 73.23% accuracy.
- The model demonstrated superior prediction results and execution quality scores.
Conclusions:
- The developed sequential learning model offers a more accurate approach to decoding motor intentions for BCIs.
- This technology can advance the development of hybrid EEG-EMG BCIs for more effective neurofeedback in stroke rehabilitation.
