Title: A novel hybrid system for detecting epileptic seizure in neonate and adult patients
Authors: Ahmed Adda; Hadjira Benoudnine; Mohamed Daoud; Philippe Ravier
Addresses: Laboratory of Electromagnetism and Guided Optics, University of Abdelhamid Ibn Badis, BP.227, Route Belhacel, 27000, Mostaganem, Algeria ' Laboratory of Electromagnetism and Guided Optics, University of Abdelhamid Ibn Badis, BP.227, Route Belhacel, 27000, Mostaganem, Algeria ' Signals and Systems Laboratory, University of Abdelhamid Ibn Badis Mostaganem, BP.227, Route Belhacel, 27000, Mostaganem, Algeria ' Laboratoire PRISME, INSA-CVL, Univ. Orléans, EA 4229, F45072, Orléans, France
Abstract: Epilepsy is a brain disease characterised by recurrent seizures. Electroencephalography (EEG) remains a prominent tool for detecting seizures. However, visual inspection of EEG traces represents a time-consuming and laborious process. Though some automatic seizure detection methods perform quite well in case of adult patients, they fail in the neonatal case. Therefore, this research proposes an automated system for detecting seizures in patients regardless their ages. The proposed system takes advantage of hybridation between generalised Hurst exponent and approximate entropy features extracted from the envelopes of EEG signals. These features are taken as input parameters of the SVM classifier. To assess the generality of the proposed technique, a binary test (normal vs. seizure) was achieved on two datasets, including EEG records of adults and neonates. Our system detects seizures with an accuracy of 99% in the neonatal case and an accuracy of 100% in the adult case.
Keywords: electroencephalogram; epilepsy; seizure detection; Hilbert transform; signal envelope; multifractal patterns; Hurst parameter; regularity; entropy; support vector machine; SVM.
DOI: 10.1504/IJBET.2023.131711
International Journal of Biomedical Engineering and Technology, 2023 Vol.42 No.2, pp.205 - 223
Received: 27 Nov 2020
Accepted: 23 Aug 2021
Published online: 29 Jun 2023 *