A sequential learning model with GNN for EEG-EMG-based stroke rehabilitation BCI

Frontiers in Neuroscience, 2023 · DOI: 10.3389/fnins.2023.1125230 · Published: April 17, 2023

Simple Explanation

This paper introduces a new method for brain-computer interfaces (BCIs) that uses both EEG and EMG signals to help stroke patients recover motor function. The method employs a sequential learning model that incorporates a Graph Isomorphic Network (GIN) to process data from these signals. The model divides movements into smaller sub-actions and predicts them separately. By analyzing the sequence of these sub-actions, the system can provide more accurate feedback to patients, potentially improving their rehabilitation outcomes. The study showed that this new approach achieved higher accuracy in classifying movements compared to existing methods, suggesting it could be a valuable tool for developing more effective rehabilitation systems.

Study Duration

Not specified

Participants

5 healthy volunteers and 2 stroke patients

Evidence Level

Original Research

Key Findings

1
The proposed GIN-based model achieved a classification accuracy of 88.89% on an EEG-EMG synchronized dataset for push and pull movements.
2
The GIN-based model significantly outperformed the benchmark method’s performance of 73.23%.
3
Both EEG and EMG signals contribute to the performance of the model, with the combination achieving significantly better classification accuracy than using a single data source.

Research Summary

This paper introduces a sequential learning model incorporating a Graph Isomorphic Network (GIN) that processes a sequence of graph-structured data derived from EEG and EMG signals for stroke rehabilitation BCI. The method achieves a classification accuracy of 88.89% on an EEG-EMG synchronized dataset, significantly outperforming the benchmark method’s performance of 73.23%. The proposed model can provide sequential compensatory feedback to patients, allowing them to receive better feedback on the quality of their movements and potentially rebuild their neural circuitry.

Practical Implications

Improved Stroke Rehabilitation

The hybrid EEG-EMG brain-computer interface can provide patients with more accurate neural feedback to aid their recovery.

Personalized Rehabilitation Plans

By using both EEG and EMG signals, a more comprehensive rehabilitation plan can be tailored to the specific needs of each patient.

Enhanced Movement Assessment

The method can be used to assess movements at a finer level of intensity to determine whether they are being performed correctly at each stage.

Study Limitations

1
The interpretation of sub-actions is weak.
2
The model struggled to accurately predict data for more distant sub-actions of the same movement for all subjects.
3
The SPMI calculation has a high computational complexity.

Your Feedback

Was this summary helpful?

Back to Neurology