An Efficient Patient-Independent Epileptic Seizure Assistive Integrated Model in Human Brain-Computer Interface Applications

Abstract

Epileptic seizures are brief disruptions in the electrical activity of the brain. Epilepsy is a central nervous system illness in which people have repeated seizures that happen at random times and usually without warning. Seizures are more likely to cause physical harm and even death in people who have them frequently. Intelligent techniques supporting brain-computer interfaces in seizure control can provide efficient epilepsy detection where seizure events can be identified in the early stages triggering electrical stimulations to be sent to the cortex of the brain. In this paper, a non-patient-specific seizure detection model is introduced. The proposed model integrates wavelet-based electroencephalography EEG brain signal processing with feature extraction to extract combined features from both time and frequency domains. Classification has been applied using different machine learning techniques for efficient early detection of epileptic seizures. In addition, channel selection analysis is implemented to reach an accurate generic model. The experimental comparative study demonstrated that the electroencephalography signals from the frontal lobe channels supply more discriminative features than the other channels, which enhanced the classification accuracy and sensitivity in the proposed model. Experiments have been carried out on the Children's Hospital Boston-Massachusetts Institute of Technology dataset to validate the robustness of the proposed model. The experimental results showed that the proposed model achieved 99.792% accuracy and 99.59% sensitivity in detecting epileptic seizures with improvement in accuracy with rates of 7%, 5%, and 8% compared to [14], [11], and [9], respectively. Experiments showed that the proposed system can detect epileptic seizures effectively, which can give remarkable potential in clinical applications.