Abstract

Info

Title:: Classification of epileptic electroencephalogram signal using graph convolutional neural network with multiple features and multiple relations

Author(s):: YU Yongqiang1; LIU Jian2; SUN Rencheng 1; SUI Yi 1; 1. College of Computer Science and Technology, Qingdao University, Qingdao 266071, China 2. Qingdao Stomatological Hospital, Qingdao 266000, China

Keywords:: Keywords: electroencephalogram epilepsy graph convolutional neural network multiple-relation graph

Abstract:: Abstract: Objective To present a graph convolutional neural network (GCNN) based on multi-feature and multi-relations for the electroencephalogram (EEG) signal classification in epilepsy for promoting the application of GCNN in epileptic EEG signal classification and improving the classification accuracy. Methods One frequency domain feature, 3 time frequency domain features and 2 nonlinear dynamical features of epileptic EEG signals were taken as features of model nodes. The spatial similarity and spectral similarity between EEG channels were extracted and fused as the edge relationship matrix between the overall graph nodes. Results Experiments were conducted on the TUEP dataset to evaluate the classification performance. The proposed model achieved an accuracy, precision, recall, F1-score, and AUC of 0.87±0.02, 0.91±0.04, 0.82±0.04, 0.86±0.02, and 0.90±0.03, respectively. Conclusion The proposed model is advantageous over the single-feature and single-relation GCNN in the classification of epileptic EEG signal.