A hybrid metaheuristic framework for epileptic seizure detection in healthcare decision support systems

Area of Science:

• Computational intelligence and machine learning applied to biomedical signal processing.
• Healthcare informatics and the development of automated diagnostic tools.

Background:

• Epileptic seizure detection is vital for patient care, requiring precise and reliable methods.
• Automated seizure detection systems benefit from computational intelligence and optimization techniques.

Purpose of the Study:

• To develop a novel hybrid approach for patient-specific epileptic seizure detection.
• To enhance the performance of automated seizure detection using optimization algorithms.

Main Methods:

• A hybrid approach combining Ant Colony Optimization (ACO) for feature selection and Gray Wolf Optimization (GWO) for Random Forest (RF) hyperparameter tuning.
• Evaluation using multichannel EEG data from the CHB-MIT and Seina datasets.
• Assessment of performance using accuracy, sensitivity, and specificity metrics.

Main Results:

• The ACO-GWO-RF pipeline achieved high performance on the CHB-MIT dataset (96.70% accuracy, 92.66% sensitivity, 99.24% specificity).
• The Seina dataset yielded 93.01% accuracy, 89.82% sensitivity, and 96.26% specificity.
• The hybrid method demonstrated robustness in handling complex EEG data and outperformed existing approaches.

Conclusions:

• The hybrid metaheuristic approach effectively optimizes EEG data processing and classification for seizure detection.
• The method's strong, patient-specific performance suggests potential for interactive health applications and personalized epilepsy management.