GRC-Net: Gram Residual Co-attention Net for epilepsy prediction

TL;DR

GRC-Net introduces a novel 3D Gram matrix transformation and multi-level coattention architecture for improved epilepsy prediction from EEG signals, achieving state-of-the-art accuracy on challenging datasets.

Contribution

The paper proposes a new Gram matrix-based 3D representation and multi-granular feature extraction framework for EEG-based epilepsy prediction, enhancing modeling of signal relationships.

Findings

Achieved 93.66% accuracy on the BONN dataset's five-class task.

Outperformed existing methods in epilepsy classification.

Effectively captured global and local EEG signal features.

Abstract

Prediction of epilepsy based on electroencephalogram (EEG) signals is a rapidly evolving field. Previous studies have traditionally applied 1D processing to the entire EEG signal. However, we have adopted the Gram Matrix method to transform the signals into a 3D representation, enabling modeling of signal relationships across dimensions while preserving the temporal dependencies of the one-dimensional signals. Additionally, we observed an imbalance between local and global signals within the EEG data. Therefore, we introduced multi-level feature extraction, utilizing coattention for capturing global signal characteristics and an inception structure for processing local signals, achieving multi-granular feature extraction. Our experiments on the BONN dataset demonstrate that for the most challenging five-class classification task, GRC-Net achieved an accuracy of 93.66%, outperforming existing methods.