A novel epilepsy seizure prediction model using deep learning and classification

Abstract

Epilepsy is a common neurological disease where the earlier disease prediction significantly impacts those patients’ lives. In this paper, a novel epilepsy seizure prediction approach is designed using deep learning. The proposed model is applied to the Electroencephalogram (EEG) recordings collected from Children’s Hospital Boston (CHB-MIT). The recording data is grouped into 23 cases, including 17 females and five males of different ages. The recordings are sampled at 256 samples/s of 16-bit resolution. The target is to analyse the brain’s state and evaluate the changes encountered from the interictal state. The earlier prediction process helped in timely disease identification and treatment to rescue the patients. Feeding the raw EEG signals over the feature extractor reduces the computational complexity and execution time. An Adaptive Grey Wolf Optimizer (AGWO) is used for learning the features and promoting those discriminative features to enhance the prediction rate and classification accuracy. To optimize the features integrating the auto-encoder concept with Genetic Algorithm (GA) in an adaptive manner termed as to enhance the prediction rate. The functionality of is tested over the subjects of the CHB-MIT EEG dataset to achieve resourceful outcomes. The proposed attains higher accuracy of 99% and reduces the False Alarm Rate (FAR) with little prediction time. The model’s functionality is evaluated using the MATLAB simulation environment and shows a better trade-off than existing approaches.

Keywords

• Epilepsy
• Deep learning
• Predictor model
• Feature learning
• Electroencephalogram
• Genetic algorithm