Real-Time Epileptic Seizure Detection using Machine Learning Techniques

Abstract

Accurate real-time detection of epileptic seizures carries paramount importance with far reaching ramifications for epileptic patients to get timely and proper medical interventions. An epileptic patient can derive confidence from the fact that effective treatment can be administered to them efficiently. Acquiring real-time data of ictal and inter-ictal states of epileptic patients is crucial for real-time epileptic seizure detection. To meet this requirement, an indigenous system is developed in this research work. This system can gather data of epileptic patients as well as that of healthy patients, thereby, enabling the researcher to distinguish between the pre-ictal, inter-ictal, ictal, and healthy states. The real-time Electroencephalography signals of healthy persons and data of ictal and inter-ictal states of epileptic patients are gathered from the indigenous system. Using machine learning techniques, the data is extracted and classified to achieve high accuracy in the detection of epileptic seizure in real-time. Using the Discrete Wavelet Transform technique, the Mean and Standard Deviation are extracted from the last decomposition level of the frequency bands. The first component of Principal Component Analysis is used as a feature extraction process. Then, using a Quadratic Discriminant Classifier, the data of healthy persons and the data of the ictal state and inter-ictal state of epileptic patients is classified. The proposed methodology achieves a real-time epileptic seizure detection rate of 99%.