Identification of epileptic seizures from scalp EEG signals based on TQWT

Abstract

In this work, we propose a method for epileptic seizure detection from scalp electroencephalogram (EEG) signals. The proposed method is based on the application of tunable-Q wavelet transform (TQWT). The long duration scalp EEG signals have been segmented into one-second duration segments using a moving window-based scheme. After that, TQWT has been applied in order to decompose scalp EEG signals segments into multiple sub-band signals of different oscillatory levels. We have generated two-dimensional (2D) reconstructed phase space (RPS) plot of each of the sub-band signals. Further, the central tendency measure (CTM) has been applied in order to measure the area of the 2D-RPS plots. These computed area measures have been used as features for distinguishing seizure and seizure-free EEG signal segments. Finally, we have used a feature-processing technique which clearly discriminates epileptic seizures in the scalp EEG signals. The proposed method may also find application in the online detection of epileptic seizures from intracranial EEG signals. © Springer Nature Singapore Pte Ltd 2019.