GSA-PSO: A Hybrid Optimization Algorithm using WPT and RBFNN for Accurate Epileptic Seizer Detection

Epilepsy is a life-risk mental illness caused by irregularities in the electrical signal flow in the brain that cause other parts of the brain to respond differently called seizures. For the last two decades, neurologists and researchers have been brainstorming to detect seizure signals accurately. Presently metaheuristic algorithms and machine learning algorithms are widely used to design powerful EEG devices. In this proposed experiment, a hybrid optimization algorithm, GSA-PSO: “Gravitational Search Algorithm and Particle Swarm Optimization Algorithm” is used with a Radial Basis Function Neural Network (RBFNN) classifier to detect seizure signals. For experimenting short-term Bonn University EEG dataset, Wavelet Packet Transform (WPT) for feature extraction and ANOVA test for feature selection are used. The outcomes of the performance measurements are an average accuracy of 100%, sensitivity of 100%, specificity of 100%, and area under curve (AUC) values as 1. This experiment can be tested with other publicly available datasets in the future