Coherence based graph convolution network for motor imagery-induced EEG after spinal cord injury

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Spinal cord injury (SCI) significantly impacts motor function and autonomic nervous system regulation.
• Brain-computer interface (BCI) systems utilizing motor imagery (MI) offer promising therapeutic avenues for SCI.
• EEG signal analysis is crucial for developing effective BCI-based rehabilitation strategies.

Purpose of the Study:

• To develop and validate a novel C-GCN method for extracting spatio-temporal-frequency features and functional connectivity from EEG signals.
• To improve the classification accuracy of motor imagery tasks in individuals with SCI.
• To establish a robust framework for BCI applications in SCI rehabilitation.

Main Methods:

• A coherence-based graph convolutional network (C-GCN) was developed to analyze EEG signals.
• The C-GCN method constructs multi-channel EEG features using coherence networks to represent functional connectivity.
• EEG data from SCI patients and healthy controls were analyzed to classify MI tasks.

Main Results:

• The C-GCN method demonstrated superior classification performance compared to traditional approaches.
• The highest classification accuracy achieved was 96.85%, indicating high reliability and stability.
• The analysis successfully identified MI-related functional connections within EEG signals.

Conclusions:

• The proposed C-GCN framework provides an effective method for analyzing EEG signals in BCI applications.
• This approach offers a strong theoretical foundation for advancing rehabilitation treatments for SCI patients.
• The findings highlight the potential of advanced machine learning techniques in improving BCI efficacy for neurological disorders.