A comparative account of modelling seizure detection system using wavelet techniques

作     者：Swami, Piyush Gandhi, Tapan K. Panigrahi, Bijaya K. Bhatia, Manvir Santhosh, Jayasree Anand, Sneh 

作者机构：Centre for Biomedical Engineering Indian Institute of Technology Delhi and Department of Biomedical Engineering All India Institute of Medical Sciences New Delhi India Department of Electrical Engineering Indian Institute of Technology Delhi New Delhi India Department of Neurology Saket City Hospital New Delhi India Department of Biomedical Engineering University of Malaya Kuala Lumpur Malaysia 

出 版 物：《International Journal of Systems Science: Operations and Logistics》 (Int. J. Syst. Sci. Oper. Logist.)

年 卷 期：2017年第4卷第1期

页      面：41-52页

基　　金：Department of Science and Technology, Government of Kerala Indian Institute of Technology Mandi, IIT, (RP03060) Indian Institute of Technology Mandi, IIT 

主　　题：discrete wavelet transform (DWT) dual-tree complex wavelet transform (DT-CWT) Electroencephalography (EEG) seizure wavelet packet transform (WPT) 

摘      要：Over the last decade, application of wavelet transform (WT) has been realised for extracting features during epileptic seizure detection. Although noteworthy developments have been made in WT algorithms, most of the seizure detection works have been confined to using discrete wavelet transform (DWT). In this present effort, comparisons are made between DWT, wavelet packet transform and dual-tree complex wavelet transform (DT-CWT) for detection of epileptiform patterns in electroencephalography. For the study, combinations of energy, root-mean-square values and standard deviations are used as extracted feature inputs to the general regression neural network classifier. The paper describes unique methodology of using minimal training during K-folds cross-validation to highlight the robustness of the expert model. Classification rates, statistical parameters and computation timings are finally calculated and one-way analysis of variance is applied to validate the results. The results demonstrate statistically significant and ceiling level of classification performances (98%) using DT-CWT extracted coefficients. The present study provides a comparative account for selection of the best model for faithful diagnosis of epilepsy with speed. © 2016, © 2016 Informa UK Limited, trading as Taylor & Francis Group.