Optimization-Driven EEG-Based Seizure Detection in the Healthcare Internet of Things.
Abstract
Objective
This work proposes an automated system for the detection of epileptic seizures based on electroencephalogram (EEG) signals, integrating numerous metaheuristic optimization algorithms. This study aims to improve classification accuracy and computational efficiency by optimizing feature selection and classification in a machine learning-based detection system.
Methods
The four feature extraction strategies were employed:(i) the Discrete Wavelet Transform (DWT), (ii) the DWT with the Hjorth parameters, (iii) the DWT with the statistical features, and (iv) the DWT with the statistical and Hjorth parameters. The metaheuristic- based optimization methods, Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Bee Algorithm (BA), and Genetic Algorithm (GA) were used to select the optimal feature from each feature set. Optimized features were then classified using a Support Vector Machine (SVM) classifier.
Results
DWT + PSO + SVM model showed the highest level of classification of 97.8%, which is significantly higher compared to the other combinations of feature extraction and optimization. The feature optimization using Genetic Algorithms showed better computational efficiency, whereas ACO and BA yielded performance improvement within a given feature configuration.
Discussion
The better performance of the meta-heuristic-optimized models, especially the DWT-PSO hybrid, demonstrates the effectiveness and robustness of the proposed model for automated seizure detection. These findings demonstrate the appropriateness of meta-heuristic optimization for dealing with the complex feature space of EEG and enhancing detection reliability. The results are particularly meaningful to real-time seizure monitoring systems in resource-limited systems, e.g., Healthcare Internet of Medical Things (IoMT) systems.
Conclusion
The findings indicate the high potential of meta-heuristic optimization tools for improving the accuracy of epileptic seizure detection and the overall performance of the system. The suggested architecture provides a clear pathway for developing effective, reliable, and real- time EEG-based seizure-detection systems.