Using machine learning to predict epileptic seizures from EEG data

Abstract

With the recent Artificial Intelligence (AI) progress, especially Machine Learning (ML), researchers aim to apply techniques for improving and automating certain clinical practice facets. This paper proposes and investigates the efficiency of several machine learning approaches to predict epileptic seizure onsets accurately to prepare patients for recurrent convulsion episodes, enhancing their quality of life. The study was performed in collaboration with industry and analyses the non-invasive scalp Electroencephalography (EEG) signals. The feature space is extracted using statistical, and wavelet transforms. The results from K-Nearest Neighbour (KNN), Support Vector Machines (SVM), and an Ensemble Classifier are compared. The proposed techniques are evaluated using one of the most extensive seizure EEG datasets, the CHB-MIT dataset, which includes 192 seizure readings from 22 patients suffering from intractable seizures. It is among the first seizure prediction studies and shows that the three methods perform similar, although the Ensemble Classifier achieves a higher specificity, sensitivity and accuracy.