Exploring Hermite transformation in brain signal analysis for the detection of epileptic seizure

Abstract

Automatic detection of epileptic seizure from brain signal data (e.g. electroencephalogram (EEG)) is very crucial due to dynamic and complex nature of EEG signal (e.g. non-stationarity, aperiodic and chaotic). Owing to these natures, manual interpretation and detection of epileptic seizure is not reliable and efficient process. Hence, this study is intended to develop a new computer-aided detection system that can automatically and efficiently identify epileptic seizure from huge amount EEG data. In this study, Hermite Transform is introduced for extracting discriminating information from EEG data for the detection of epileptic seizure. The analysis is performed in three stages: EEG signal transformation into a new form by Hermite Transform; computation of three types of features, namely permutation entropy, histogram feature and statistical feature; and classification of obtained features by least square support vector machine. The classification outcomes reveal the presence of epileptic seizure. The proposed method is evaluated on a benchmark Epileptic EEG database (Bonn University data) and the performance of this method is compared with several state-of-art algorithms for the same database. The experimental results demonstrate that the proposed scheme has the ability to efficiently detect epileptic seizure from EEG data outperforming competing techniques in terms of overall classification accuracy.