Seizure detection using integrated metaheuristic algorithm based ensemble extreme learning machine

Abstract

In biomedical research, the brain signal analysis occupies an important space in recent days. Mostly Epileptic seizure detection is a challenging task for all brain signal researcher. In this paper ensemble approach is considered for seizure classification and detection. It is essential of early detection for the patient to save the life. Initially Wavelet transform is used to extract the relevant features. As the feature dimension is high, features are reduced using linear discriminant analysis (LDA). The metaheuristic algorithms named as Water cycle algorithm and accelerated particle swarm optimization (APSO) are integrated to optimize the weights of ensemble extreme learning machine (EELM) for classification. The features are aligned as input to WCA-APSO based EELM model. To validate the proposed integrated algorithm three benchmark functions are utilized for optimization to exhibit the uniqueness. The data is taken from university of BONN database for experimentation. The performance is measured with the parameters like sensitivity, Specificity, and Accuracy are obtained as 99.32%, 99.68%, and 99.16%, The result found superior as compared to earlier methods and outperforms the classification of Epileptic seizure signals.

Keywords

• Extreme learning machine
• Feature reduction
• Water cycle algorithm
• Wavelet transform
• Epilepsy
• Accelerated particle swarm Optimization