Novel Classification Method of Spikes Morphology in EEG Signal Using Machine Learning

Abstract Since its invention in 1929 by Hans Berger, the electroencephalography (EEG) is the subject of several researches by its importance in the understanding of epilepsy in general and particularly in the diagnosis but especially in the near-surgical evaluation of the disease. EEG is a signal acquisition tool from cerebral electrical discharges. Recently Khouma [1] has proposed a tool to detect the Interictical Paroxystic Events (IPE) or spikes in EEG signals. In this paper, we propose a new classification method of spikes morphology based on the Support Vector Machines (SVM). The SVM is a supervised classification method using kernel functions. It is a powerful technique and particularly useful for data whose distribution is unknown (EEG signals). We apply this technique to identify the different spikes morphologies in EEG signals. Different kernel functions (linear, polynomial, radial and sigmoidal) are used for experimental. Automatic treatment for identification spikes morphology could improve the diagnosis of epilepsy.