A Comparative Analysis of Artificial Intelligence Optimization Algorithms for the Selection of Entropy-based Features in the Early Detection of Epileptic Seizures

Abstract

Epilepsy, a neurological condition colloquially known as a seizure disorder, causes involuntary muscle contractions and cognitive changes through sudden, uncontrolled neuronal discharges in the brain. The recurrent, unpredictable nature of these seizures poses the threat of potentially fatal or irreversible brain damage, underscoring the critical importance of early detection of epilepsy seizures. This study extracts informative features from medical records to improve early epilepsy seizure diagnosis. Employing bio-inspired optimization algorithms, it performs feature selection and constructs two different machine learning models, both equipped with optimization algorithms for epilepsy seizure diagnosis. Evaluation encompasses comprehensive metrics including accuracy, precision, F1 score and computational cost, with convergence graphs highlighting the impact of the optimization algorithms. Encouragingly, the results show that the model, using just five selected features, achieves an impressive 95.28% accuracy in diagnosing epileptic seizures. This highlights the suitability of the proposed model for real-time applications, characterized by its streamlined parameter complexity.