ZOADL: An Optimized Deep Learning Framework for Detecting Abnormal Brain Activity using EEG Signals

Electroencephalography (EEG) is extensively employed for monitoring cerebral activity; nevertheless, the human interpretation of EEG signals is laborious, susceptible to errors, and inadequate for real-time applications.This paper presents a novel Zebra Optimisation Algorithm with a Deep Learning-Assisted Framework (ZOADL) to detect aberrant brain activity in EEG data.The ZOADL system has three fundamental stages: data preprocessing, feature selection, and classification.During the preprocessing phase, raw EEG data are purified to diminish noise and improve signal integrity.The Zebra Optimisation Algorithm (ZOA) is employed for feature selection, pinpointing the most pertinent features that enhance classification efficacy.The system utilises the Long Short-Term Memory (LSTM) neural network to identify aberrant brain activity, as it is particularly adept at capturing temporal relationships in EEG signals.The Chicken Swarm Optimisation (CSO) technique is utilised to optimise the hyperparameters of the LSTM, guaranteeing superior model performance.Experimental results indicate that the ZOADL framework surpasses conventional and cuttingedge methods regarding accuracy, precision, recall, and F1-score, markedly decreasing false positives and enhancing overall detection efficiency.This study emphasises the capabilities of ZOADL as a sophisticated, automated decision-support system, allowing doctors to identify abnormal brain activity with improved precision and speed, hence augmenting patient management and therapeutic results.Subsequent efforts may concentrate on expanding the framework to accommodate larger datasets and realtime applications.