Classification of Epileptic Seizures Using LSTM Based Zebra Optimization Algorithm with Hyperparameter Tunin

Electroencephalogram (EEG) are the neuro-electrophysiology signals, which are commonly used as a diagnostic tool to measure the seizure activity of the brain.The accurate detection and classification of seizures help to provide an optimal solution to diagnose the patient.In this research, a hyperparameter tuning with Zebra Optimization Algorithm (ZOA) is proposed for the fine tuning of features from EEG signals.The EEG signals are taken from the three standard datasets such as Temple University Hospital (TUH) at a rate of sampling signal of 250Hz, Bonn University (BU) at a 173.61Hz sampling rate, and Bern Barcelona (BB) alongside the sampling frequency of 512 Hz.The EEG signals are pre-processed using Butterworth 8th order filtering method to remove the unwanted noise, and de-noised signals are decomposed by the swarm decomposition method.Features like statistic-based features, frequency-dependent features, multi-scale wavelet transformation, entropy features, and power spectral features are extracted from the decomposition of signals.The extracted features then undergo hyperparameter tuning using ZOA followed by feature selection using Enhanced Spatial bound Whale Optimization Algorithm (WOA) with the combination of Salp Swarm Algorithm (SSA) hybridized with Lens Opposition-based Learning (LOBL) mechanism.The features obtained from the selection algorithm are then fed to hyper parameter optimized Long Short-Term Memory (LSTM) classifier to classify the normality and abnormality of seizures.The attained outcomes of the suggested approach exhibit a better classification rate with 98.43% accuracy on BU dataset, 99.71% accuracy on BB dataset, and 98.43% accuracy on TUH dataset.