基于时间序列复杂网络的癫痫脑电分类研究  

Research on Epilepsy EEG Classification Based on Time Series Complex Network

作　　者：杨晓利[1] 杨彬 李振伟[1] 吴晓琴 YANG Xiaoli;YANG Bin;LI Zhenwei;WU Xiaoqin(School of Medical Technology and Engineering,Henan University of Science and Technology,Luoyang 471000)

机构地区：[1]河南科技大学医学技术与工程学院,洛阳471000

出　　处：《计算机与数字工程》2023年第12期2814-2820,共7页Computer & Digital Engineering

基　　金：河南省重点研发与推广专项(编号:202102310534)资助。

摘　　要：大脑是一个高度复杂的系统,而且脑电信号噪声背景强,信号微弱,传统的脑电信号特征提取方法不能全面反应脑电信号的特征信息,因此,提出一种与复杂网络理论的相结合,以时间序列为基础构造复杂网络的癫痫脑电分类方法。首先将癫痫脑电信号的时间序列分段处理,每一段作为网络的一个节点,通过Pearson相关计算节点之间的关系来构造网络的连接矩阵,然后通过连接矩阵计算网络特征参数,并对特征参数进行统计分析构造特征向量,最后,使用SVM、逻辑回归和K-NN等分类器进行分类研究。结果显示,该方法对数据集A-E、AB-CDE和ABCD-E的分类准确率分别达到96.67%、94.00%和94.33%。实验结果表明,作为传统时间、频率分析的替代方法,该方法是可用于对脑电信号进行模式识别分类的,能够有效对癫痫脑电信号分类识别。The brain is a highly complex system,and the EEG signal has a strong noise background and weak signal.The tra-ditional EEG signal feature extraction method cannot fully reflect the feature information of the EEG signal.Therefore,an epilepsy EEG classification method combining with complex network theory and constructing complex networks based on time series is pro-posed.First,the time series of epilepsy EEG signals is processed in segments,and each segment is used as a node in the network.The relationship between nodes through Pearson correlation is calculated to construct the connection matrix of the network,and then the network feature parameters is calculated through the connection matrix,and statistical analysis on the feature parameters is per-formed to construct the feature vector.Finally,classifiers such as SVM,logistic regression and K-NN are used for classification re-search.The results show that the classification accuracy of this method for data sets A-E,AB-CDE and ABCD-E reached 96.67%,94.00%and 94.33%,respectively.Experimental results show that,as an alternative to traditional time and frequency analysis,this method can be used for pattern recognition and classification of EEG signals,and can effectively classify and recognize epileptic EEG signals.

关 键 词：复杂网络 时间序列 癫痫 脑电 分类 

分 类 号：TP30[自动化与计算机技术—计算机系统结构]