Comparison of Patient Specific and General Classification of Epileptic Seizure Prediction

Abstract

Epilepsy is a neurological disorder in which abnormal brain activity occurs, causing seizures. Recent studies have used machine learning techniques to produce a seizure classification system. In this work, two aspects of seizure classification are discussed and compared in terms of accuracy and efficacy. Seizure classification can follow a patient specific or general approach. For a patient specific approach, feature extraction and classification are performed for each patient independently. However, a general approach means data is trained and classified for all patients at once. Results show that AUC of general approach is 0.74 which is higher than that of patient-specific 0.71. Computational time is decreased when using patient-specific approach to 8 hours, while general approach requires 10 hours for training and prediction.