Hybrid metaheuristic algorithm enhanced support vector machine for epileptic seizure detection. (September 2022)

• Record Type: Journal Article
• Authors: Divya, P.; Devi, B. Aruna
• Is Part Of: Biomedical signal processing and control. Volume 78(2022)
• Publication Date: 2022-09
• Subjects: EEG -- Grasshopper optimization -- Epilepsy classification -- Improved sine cosine algorithm; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103841
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An optimized machine learning model, Support Vector Machine (SVM) for the diagnosis of epilepsy seizure is proposed. A novel Hybrid Grey Wolf Optimizer-Improved Sine Cosine Algorithm (HGWOISCA) was proposed to select the salient features. The HGWOISCAA-SVM –epilepsy diagnosis performance was compared to the performance of earlier methods. The proposed model achieved classification accuracy of 100% for Bonn university data and 99.9% for clinical data. Abstract: Analysis and detection of epileptic seizures are usually done by physicians through the visual scanning of EEG signals, which tends to be subjective, time-consuming, inaccurate, and prone to errors. Various research directions have focused on the detection and classification of epilepsy using machine learning methods. However, the prevailing methods have shortcomings like low classification rates and slow convergence. To address those limitations, this paper designs a fully automated system based on Hybrid Grey Wolf Optimizer-Improved Sine Cosine Algorithm (HGWOISCA) enhanced Support Vector Machine (SVM) named HGWOISCA-SVM for EEG signal classification. Primarily, unprocessed EEG signals are denoised by using an improved wavelet thresholding function to diminish noise and redundant data. Secondarily, three types of features such as wavelet domain features, time-domain features, and chaotic features are computed from the preprocessed EEG signals. In the following phase, Enhanced Grasshopper Optimization Algorithm (EGOA) is adopted to select optimal features with high discriminative power and to reduce dimensionality. Finally, the selected features are input into the HGWOISCA-SVM to differentiate healthy EEG signals from epileptic seizure signals. The effectiveness of the developed system is investigated by using Bonn university data and clinical data. Experiential results prove that the developed system yields a high classification rate of 100% for Bonn university data and 99.9% for clinical data, proving to be a powerful system for epilepsy classification. A comparison of the proposed system with the existing epileptic seizure detection methods is also done for validation.