Multiresolution approach for artifacts removal and localization of seizure onset zone in epileptic EEG signal

Abstract

Epilepsy is major challenge to medical fraternity. This research article has come up with a novel method for pre-processing of Epileptic Electroencephalogram (EEG) signals for artifact removal and localizing the epileptic seizure onset zone. The presence of physiological artifacts alters the true EEG information which can considerably affect the epileptic seizure classification accuracy. Hence, it is of prime importance to eliminate these contaminations. The analysis of epileptic EEG signal requires localization of the seizure onset zone in the epileptic EEG signal. The objective of this paper is to present an approach to eliminate physiological artifacts and to localize the epileptic region. This paper suggests a hybrid model based on Multiresolutional Analysis and Adaptive Filtering [MRAF]. Initially the EEG signal is decomposed using Discrete Wavelet Transform (DWT) which effectively helps to localize the epileptic region. Multiresolutional soft thresholding is applied to these decomposed wavelet components to remove the abrupt variations. Further, adaptive filtering is applied to remove the low frequency components which represents physiological artifacts. It is observed that the MRAF method effectively eliminate physiological artifacts and localizes the epileptic seizure zone present in the seizure EEG information. The accuracy of the proposed MRAF model is found to be 86.66% with a precision of 88.88%, since it is able to retain most of the seizure signal present in the tested datasets.