Detection of Epileptic EEG Signals Based on an Improved AFSA-GA Algorithm

Epilepsy is a neurological disorder that occurs worldwide and is characterized by sudden abnormal electrical discharges in the brain. The main benefit of detecting epileptic seizures through electroencephalogram (EEG) is to assist in the evaluation and diagnosis of epilepsy patients. The aim of this study is to establish a hybrid model based on an improved artificial fish swarm algorithm (AFSA) and genetic algorithm (GA) to determine the optimal feature combination for detecting epileptic seizures. Firstly, the geometric features of ten epilepsy EEG signals were extracted using the second-order difference plot (SODP) method across the entire frequency range. Then, an optimization algorithm for feature selection, called AFSA-GA, was proposed by combining the improved AFSA with GA, which was based on a food strategy-based update step. Finally, the feature combination was inputted into a naive Bayes classifier for identifying epileptic and non-epileptic seizures. The proposed method achieved a classification accuracy of 92.94% using a ten-fold cross-validation strategy in frequency band crossing, and is an effective approach for detecting epilepsy that can assist clinicians in quickly diagnosing epilepsy based on SODP analysis and feature selection optimization algorithms.