Coherence-Based Graph Convolution Network to Assess Brain Reorganization in Spinal Cord Injury Patients

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Biomedical Engineering

Background:

• Motor imagery (MI) involves extensive brain networks for action simulation.
• Understanding brain network reorganization post-spinal cord injury (SCI) is vital for assessing neural activity.
• Investigating these changes can reveal mechanisms underlying functional recovery.

Purpose of the Study:

• To examine brain network reorganization during MI in SCI patients.
• To analyze alterations in different frequency bands from resting state to MI.
• To introduce and validate a novel classification algorithm for EEG data.

Main Methods:

• Electroencephalogram (EEG) coherence analysis was performed on SCI patient data.
• Brain network activity was compared between resting state and MI tasks.
• A consistency calculation-based residual graph convolution (C-ResGCN) algorithm was developed for classification.

Main Results:

• EEG coherence analysis revealed decreased [Formula: see text]- and [Formula: see text]-band connectivity during MI compared to rest.
• [Formula: see text]-band connectivity increased in motor regions, while default mode network activity decreased during MI.
• The C-ResGCN algorithm achieved a high classification accuracy of 96.25%.

Conclusions:

• Brain reorganization in SCI patients shifts resources from resting state to MI tasks.
• Effective network reorganization is linked to improved MI performance.
• Findings suggest potential biomarkers for evaluating SCI rehabilitation outcomes.