An Innovative Method for Refractory Epilepsy Diagnosis Based on Microstate Analysis and Graph Convolutional Network.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Informatics

Background:

• Refractory epilepsy (RE) poses diagnostic challenges.
• Traditional electroencephalogram (EEG) analysis has limitations in clinical settings.
• EEG microstates offer a novel approach to understanding brain dynamics.

Purpose of the Study:

• To systematically investigate EEG microstate alterations in refractory epilepsy patients across seizure stages.
• To develop and validate a novel EEG microstate analysis framework for seizure recognition and classification.
• To compare the proposed framework with traditional methods for enhanced diagnostic accuracy.

Main Methods:

• Utilized two independent datasets to extract microstate features across four seizure stages.
• Constructed a directed microstate graph structure.
• Employed a directed graph convolutional network (DGCN) for classification, termed MsG-GCN.
• Compared MsG-GCN performance against traditional methods like Support Vector Machine (SVM).

Main Results:

• The proposed MsG-GCN framework achieved a classification accuracy of 80.2%, outperforming the best traditional method (SVM) at 74.3%.
• Microstates A and C demonstrated significant differences across seizure stages.
• Average microstate occurrence rate showed higher discriminative power than duration or coverage.

Conclusions:

• The study presents a novel, highly interpretable framework (MsG-GCN) for automated epileptic seizure classification.
• Graph neural networks effectively model dynamic epileptic microstate transitions.
• The framework offers a promising tool for intelligent, auxiliary diagnosis of neurological disorders like epilepsy.