Modified Gorilla Troops Optimization with Deep Learning Based Epileptic Seizure Prediction Model on EEG Signals

Abstract

Approximately 50 million people worldwide suffer from Epileptic Seizure (ES), a persistent neurological disorder that cannot spread from person to person. Electroencephalography (EEG) is a tool that is often used to identify and diagnose epilepsy by observing how the brain works. However, analyzing EEG recordings to identify epileptic activity can be difficult, time-consuming, and requires specialist expertise. However, a precise and early diagnosis of epilepsy is necessary to start anti-seizure medication treatment and reduce the risk of consequences from recurrent episodes. In this paper, a modified Gorilla Troops Optimization with a Deep Learning based ES Prediction model (MGTODL-ESP) using EEG signals is implemented. The proposed MGTODL-ESP model comprises two main processes: feature selection and prediction. The MGTODL-ESP model uses a modified gorilla troops optimization (MGTO) based feature selection algorithm to select the optimal subset of features. The MGTO-based Gated Recurrent Unit (GRU) model predicts different types of ES. Finally, the Grey Wolf Optimizer (GWO) algorithm was used to tune the parameters of the MGTODL model. The outline of the MGTO-ESP-based feature selection and Grey Wolf Optimizer (GWO)-based parameter tuning indicates the novelty of this research. A comprehensive empirical study was conducted using a benchmark CHB-MIT scalp EEG database from IEEE DataPort to investigate the improved prediction performance of the MGTODL-ESP model. A comparison of the different methods showed that the MGTODL-ESP approach was the most accurate, with an accuracy rate of 98.50%.