A novel Machine Learning approach for epilepsy diagnosis using EEG signals based on Correlation Dimension

Abstract

Epilepsy is a chronic brain disease, it causes seizures that damage patients’ lives. The main examination of epilepsy diagnosis is encephalography since its acquisition simplicity, low cost and significant results. However, the exploitation of electroencephalogram (EEG) signals require a neurological consultation, which is very time-consuming and it is not possible at all times. Recently, several published research propose different machine learning systems for EEG signals classification to detect patients’ brain states, healthy, epileptic during seizures free intervals and epileptic during seizures. Developed systems deal with real-time applications requirement, in addition to program accuracy, run time must be as short as possible and uses the lowest storage memory. So, proposed systems must be based on a simple model using few features that are fast calculated and classified by a simple method. In this paper, we propose a novel approach for EEG features extraction based on a new method for the determination of the Correlation Dimension (CD). The experimental test through a benchmark database shows the efficiency of our method in comparison to other published works. In terms of reliability, we achieve an accuracy of 100 % for almost all classification problems with specific combinations of subsets. In terms of program simplicity, we use few features and fast running model.