Spatial Enhanced Pattern Through Graph Convolutional Neural Network for Epileptic EEG Identification.

Area of Science:

• * Neuroscience
• * Machine Learning
• * Signal Processing

Background:

• * Feature extraction is crucial for epilepsy detection and recognition in clinical settings.
• * Electroencephalogram (EEG) signals are multichannel, offering opportunities to analyze inter-channel associations.
• * Existing methods may not fully leverage the spatial patterns inherent in multichannel EEG data.

Purpose of the Study:

• * To propose a novel graph convolutional neural network (GCNN)-based framework for classifying epileptic seizures from non-seizures in EEG.
• * To capture and utilize spatial enhanced patterns from multichannel EEG signals.
• * To develop a GCNN classifier capable of discriminating between normal, ictal, and interictal EEG states.

Main Methods:

• * Development of a GCNN framework to analyze spatial patterns in multichannel EEG signals.
• * Application of the GCNN for feature extraction and classification of EEG data.
• * Comparative experiments against state-of-the-art techniques to validate the proposed approach.

Main Results:

• * The proposed GCNN framework demonstrated high performance in discriminating ictal and interictal EEG signals.
• * Achieved a sensitivity of 98.33%, specificity of 99.19%, and accuracy of 98.38% for ictal/interictal discrimination.
• * The GCNN framework successfully visualized salient regions within EEG sequences.

Conclusions:

• * The GCNN-based framework is effective for classifying epileptic seizures from non-seizures using EEG data.
• * The approach excels at capturing spatial patterns, offering a novel method for EEG analysis.
• * This technique shows significant potential for improving clinical epilepsy detection and diagnosis.