Detection of the interictal epileptic discharges based on wavelet bispectrum interaction and recurrent neural network

作     者：Nabil SABOR Yongfu LI Zhe ZHANG Yu PU Guoxing WANG Yong LIAN Nabil SABOR;Yongfu LI;Zhe ZHANG;Yu PU;Guoxing WANG;Yong LIAN

作者机构：Department of Micro-Nano ElectronicsShanghai Jiao Tong University Electrical Engineering DepartmentAssiut University MoE Key Lab of Artificial IntelligenceShanghai Jiao Tong University Alibaba DAMO Academy 

出 版 物：《Science China(Information Sciences)》 (中国科学:信息科学(英文版))

年 卷 期：2021年第64卷第6期

页      面：203-221页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 1002[医学-临床医学] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 100204[医学-神经病学] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 10[医学] 

基　　金：supported in part by National Key Research and Development Program of China (Grant No. 2019YFB2204500) National Natural Science Foundation of China (Grant No. 61874171) Alibaba Group through Alibaba Innovative Research (AIR) Program 

主　　题：interictal epileptic discharges epilepsy discrete wavelet transform wavelet bispectrum long short-term memory recurrent neural network 

摘      要：Detection of interictal epileptic discharges(IED) events in the EEG recordings is a critical indicator for detecting and diagnosing epileptic seizures. We propose a key technology to extract the most important features related to epileptic seizures and identifies the IED events based on the interaction between frequencies of EEG with the help of a two-level recurrent neural network. The proposed classification network is trained and validated using the largest publicly available EEG dataset from Temple University *** results clarified that the interaction between β and β bands, β and γ bands, γ and γ bands,δ and δ bands, θ and α bands, and θ and β bands have a significant effect on detecting the IED ***, the obtained results showed that the proposed technique detects 95.36% of the IED epileptic events with a false-alarm rate of 4.52% and a precision of 87.33% by using only 25 significant features. Furthermore,the proposed system requires only 164 ms for detecting a 1-s IED event which makes it suitable for real-time applications.