Design and Development of Prediction Model to Detect Seizure Activity Utilizing Higher Order Statistical Features of EEG signals.

Abstract

Clinical data is complex, context-dependent, and multi-dimensional, and such data generates an amalgamation of computing research challenges. To extract and interpret the useful information from raw data is a challenging job. This study aims at developing an automated predictive model to diagnose the state of an epileptic patient using EEG signals. The segmented EEG signals are utilized to extract various statistical features which are used for prediction. Strategically, we have designed a fully automated neural network model, capable of classifying the seizure activity into ictal, interictal and normal state with an accuracy as high as 99.3%, maximum sensitivity of 100% and specificity as high as 98.3% for all the classes. For the different set of parameters and optimum number of neurons in hidden layer, ANN model revealed a superior model for validating the classification.