An intelligent System for Diagnosis Schizophrenia and Bipolar Diseases based on Support Vector Machine with Different Kernels

M.I. El Gohary., T.A.Al Zohairy, A.M. Eissa., S. El Deghaidy, H.M. Hussein

Bipolar disorder and schizophrenia overlap in symptoms and may share some underlying neural substrates. The discrimination between the two diseases is one of the problems that face psychiatric experts. This paper will propose some solutions to this problem based on the artificial methods. The support vector machine (SVM) is used for discrimination based on measuring of the patient EEG rhythms. The large set of features included in the EEG rhythms is reduced into smaller set of features after Fast Fourier Transform (FFT) segmentation. Different kernels are applied on the SVM which are linear, polynomial, quadratic and radial basis function. The application of SVM with different kernels for the EEG discrimination of the patients suffering from schizophrenia and bipolar diseases is the core of this work. Experimental results have shown that the proposed algorithms will solve the discrimination between the two diseases using EEG rhythms and the support vector machine with linear and quadratic kernels have achieved a high performance rate equal to 98 % and 97.667% respectively compared to the other kernels.