A New Epileptic Seizure Detection Method Based on Degree Centrality and Linear Features

扫码查看

原文链接

NETL
Springer Nature

外文摘要：With the increasing incidence of epilepsy， we need to detect the epilepsy with high efficiency to avoid the disease attack。 In this paper， we proposed two novel feature extraction methods for automatic epileptic seizure detection with high performance based on the statistic properties of complex network。 One is the degree centrality combined with the linear features as the features to classify the epileptic EEG signal。 Firstly， we transformed the time series into complex network by using horizontal visibility graph (HVG)。 Then we extracted the degree centrality of the complex network combined with the fluctuation index and variation coefficient as the three-dimensional features and the classification accuracy is up to 95。98%。 To enhance the difference of the degree centrality feature， we put the other new feature。 That is the improved degree centrality and chose the improved degree centrality as the single feature to classify the signal。 Experimental results showed that the classification accuracy of this single feature is 96。50%。

外文关键词：

Feature extraction methodDegree centralityHorizontal visibility graphFluctuation indexVariation coefficientEpileptic seizure detection

作者：

Haihong Liu、Qingfang Meng、Yingda Wei、Qiang Zhang、Mingmin Liu、Jin Zhou

展开 >

作者单位：

School of Information Science and Engineering, University of Jinan, Jinan 250022, China,Shandong Provincial Key Laboratory of Network Based Intelligent Computing, Jinan 250022, China

Institute of Jinan Semoconductor Elements Experimentation, Jinan 250014, China

会议名称：

International symposium on neural networks

会议地点：

Sapporo(JP);Muroran(JP)

会议母体文献：

Advances in neural networks - ISNN 2017

页码：

439-446

出版时间：

2017

DOI：

10.1007/978-3-319-59081-3_51