Epileptic Seizure Prediction Based on Hybrid Seek Optimization Tuned Ensemble Classifier Using EEG Signals

Area of Science:

• * Neuroscience and Biomedical Engineering
• * Computational Intelligence and Machine Learning

Background:

• * Manual analysis of electroencephalogram (EEG) for seizure detection is time-intensive and complex.
• * Automated methods combining signal processing and machine learning are crucial for efficient epilepsy management.

Purpose of the Study:

• * To develop a hybrid optimization-controlled ensemble classifier for automated epileptic seizure prediction.
• * To enhance the accuracy and efficiency of EEG-based seizure detection.

Main Methods:

• * Pre-processing of EEG signals followed by feature extraction (statistical, wavelet, entropy-based) using a hybrid seek optimization algorithm.
• * An ensemble classifier integrating AdaBoost, Random Forest (RF), and Decision Tree (DT) classifiers, optimized by the hybrid seek optimization technique.
• * Utilizing corvid and gregarious search agent characteristics in the optimization algorithm for parameter evaluation.

Main Results:

• * Achieved 96.61% accuracy, 94.67% sensitivity, and 91.37% specificity on the CHB-MIT database.
• * Demonstrated 95.31% accuracy, 93.18% sensitivity, and 90.07% specificity on the Siena Scalp database.
• * The proposed method shows significant efficacy for early seizure prediction.

Conclusions:

• * The hybrid optimization-controlled ensemble classifier is effective for automated EEG analysis and seizure prediction.
• * The developed technique offers a promising solution for improving early epilepsy detection.
• * The study highlights the potential of advanced machine learning in neurological disorder diagnostics.