Journal of Computer Applications ›› 0, Vol. ›› Issue (): 370-374.DOI: 10.11772/j.issn.1001-9081.2024030273

• Frontier and comprehensive applications • Previous Articles     Next Articles

Automatic detection method of epileptic EEG based on residual network

Yuchun XU1, Jianjun XU2()   

  1. 1.International Business College,Dongbei University of Finance and Economics,Dalian Liaoning 116025,China
    2.Institute of Supply Chain Analytics,Dongbei University of Finance and Economics,Dalian Liaoning 116025,China
  • Received:2024-03-15 Revised:2024-05-09 Accepted:2024-05-14 Online:2025-01-24 Published:2024-12-31
  • Contact: Jianjun XU

基于残差网络的癫痫脑电自动检测方法

许裕纯1, 许建军2()   

  1. 1.东北财经大学 国际商学院,辽宁 大连 116025
    2.东北财经大学 现代供应链管理研究院,辽宁 大连 116025
  • 通讯作者: 许建军
  • 作者简介:许裕纯(2000—),女,福建漳州人,硕士研究生,主要研究方向:机器学习、数据挖掘、运营管理
    许建军(1969—),男,山东菏泽人,教授,博士,主要研究方向:商业分析、数据驱动的优化、运营管理。

Abstract:

Aiming at the single classification mode of the existing epilepsy detection algorithms, an automatic detection method of epilepsy ElectroEncephaloGram (EEG) based on Residual Network (ResNet) was proposed. Firstly, a one-dimensional ResNet with three residual blocks was built to extract the intrinsic features of EEG signals. Secondly, the fully connected network was used for classification. Finally, the proposed method was tested on the epilepsy EEG database of University of Bonn with seven two-class, five three-class and five-class of all data EEG recognition problems studied, and the detection accuracies of the proposed method were 96.19%-99.32%, 95.28%-97.45%, and 82.34%, respectively. Experimental results show that the proposed method has better universality and classification accuracy, is more suitable for the practical application requirements.

Key words: epilepsy detection, Convolutional Neural Network (CNN), ElectroEncephaloGram (EEG) signal, Residual Network (ResNet), automatic detection

摘要:

针对现有癫痫检测算法分类模式单一的问题,提出一种基于残差网络(ResNet)的癫痫脑电(EEG)自动检测方法。首先,搭建具有3个残差块的一维ResNet提取EEG信号的内在特征;其次,利用全连接网络进行分类;最后,将所提方法在波恩大学癫痫EEG数据库上进行实验,研究了7种二分类、5种三分类和所有数据的五分类的EEG识别问题,所提方法的检测准确率分别为98.30%~99.46%、95.28%~97.45%和82.34%。实验结果表明,所提方法具有较好的普适性和分类准确率,更符合实际应用需求。

关键词: 癫痫检测, 卷积神经网络, 脑电信号, 残差网络, 自动检测

CLC Number: