Enhanced Epileptic Seizure Detection: Convolution Neural Net and Features Selection in EEG Signals

Epilepsy stands as an inevitable, significant, and serious neurological disorder that affects the human brain. This exploration introduces new strategy for identifying epileptic episodes through EEG signal processing, comprising distinct stages: Feature retrieval, attribute selection, and categorization. Initially, the input EEG signal undergoes preprocessing, followed by the extraction of specific features such as Time Domain Feature, including No-sequential Energy (NE), Positional Entropy (PE), and Weighted Positional Entropy (WPE). Subsequently, a feature selection phase is implemented, employing an enhanced chi-square model to identify relevant features. The opted feature then geared towards classification stage, utilizing an Optimized Convolutional Neural Net. For enhance detection precision, the CNN weights undergo optimal tuning through a shark smell optimization with a focus on weighting factors (SSOWF) model. Adopted method exhibits highest accuracy, approximately 0.9, surpassing existing models.