Identifying epileptic EEGs and congestive heart failure ECGs under unified framework of wavelet scattering transform, bidirectional weighted (2D)(2)PCA and KELM

摘要

In order to achieve the accurate identifications of various electroencephalograms (EEGs) and electrocardiograms (ECGs), a unified framework of wavelet scattering transform (WST), bidirectional weighted two-directional two-dimensional principal component anal-ysis (BW(2D)2PCA) and grey wolf optimization based kernel extreme learning machine (KELM) was put forward in this study. To extract more discriminating features in the WST domain, the BW(2D)2PCA was proposed based on original two-directional two-dimensional principal component analysis, by considering both the contribution of eigen-value and the variation of two adjacent eigenvalues. Totally fifteen classification tasks of classifying normal vs interictal vs ictal EEGs, non-seizure vs seizure EEGs and normal vs congestive heart failure (CHF) ECGs were investigated. Applying patient non-specific strat-egy, the proposed scheme reported ACCs of no less than 99.300 +/- 0.121 for all the thirteen classification cases of Bonn dataset in classifying normal vs interictal vs ictal EEGs, MCC of 90.947 +/- 0.128 in distinguishing non-seizure vs seizure EEGs of CHB-MIT dataset, and MCC of 99.994 +/- 0.001 in identifying normal vs CHF ECGs of BBIH dataset. Experimental results indicate BW(2D)2PCA based framework outperforms (2D)2PCA based scheme, the high-performance results manifest the effectiveness of the proposed framework and our proposal is superior to most existing approaches.(c) 2023 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences. Published by Elsevier B.V. All rights reserved.