Detection of Epileptic Seizures with Different Machine Learning Algorithms Using EEG Signals in Daily Life

Today, Electroencephalography (EEG) is commonly used as a diagnostic tool for epilepsy. In this study, an effective method for diagnosing epileptic seizures in non-clinical settings is proposed. To evaluate the performance of this method, EEG data from 7 pediatric patients at Boston Children's Hospital were analyzed using Decision Tree (DT), Linear Discriminant (LD), Naive Bayes (NB), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN). The time and frequency characteristics of the EEG signals were compared. Experimental results show that epileptic seizures can be determined effectively with 100% accuracy by using only 3 channels (FP1-F7, FP2-F4 and T8-P8) with mean amplitude, mean frequency, median frequency and variance features with SVM, KNN or DT.