Analyzing EEG signal data for detection of epileptic seizure: Introducing weight on visibility graph with complex network feature

Abstract

In the medical community, automatic epileptic seizure detection through electroencephalogram (EEG) signals is still a very challenging issue for medical professionals and also for the researchers. When measuring an EEG, huge amount of data are obtained with different categories. Therefore, EEG recording can be characterized as big data due to its high volume. Traditional methods are facing challenges to handle such Big Data as it exhibits non-stationarity, chaotic, voluminous, and volatile in nature. Motivated by this, we introduce a new idea for epilepsy detection using complex network statistical property by measuring different strengths of the edges in the natural visibility graph theory. We conducted 10-fold cross validation for evaluating the performance of our proposed methodology with support vector machine (SVM) and Discriminant Analysis (DA) families of classifiers. This study aims to investigate the effect of segmentation and non-segmentation of EEG signals in the detection of epilepsy disorder.