Authors: Reza Boostani; Malihe Sabeti
Addresses: Department of CSE & IT, Electrical and Computer College of Engineering, Shiraz University, Shiraz, Iran ' Department of Computer Engineering, College of Engineering, Shiraz branch, Islamic Azad University, Shiraz, Iran
Abstract: The current criterion for the diagnosis of schizophrenia is qualitative; consequently, other psychotic disorders such as schizoaffective or delusional disorder, which have similar clinical manifestations, might be misdiagnosed as schizophrenia. To overcome this drawback, a quantitative diagnosis tool, in the form of a novel brain map, is proposed to reveal the schizophrenic-dependent changes which are spatially distributed over the brain of these patients. In this study, electroencephalogram (EEG) signals from 20 schizophrenic and 20 control subjects were acquired and then the energy in the five standard EEG bands were elicited from each channel. Discriminative bands were selected using genetic algorithm, particle swarm optimisation and ant colony optimisation. The selected features were then fed to Fisher linear discriminant analysis for classifying the two groups. Experimental results provided 83.74%, 81.41% and 81.06% classification accuracy for PSO, ACO and GA feature selectors, respectively. According to the selected band at each channel, a brain map was constructed and grand average brain maps for patients and control subjects using GA, ACO and PSO algorithms were separately sketched. Among the proposed brain maps, the one optimised by PSO revealed all differences which previously observed between their PET, fMRI and CT images.
Keywords: brain map; GA; PSO; ACO; EEG; schizophrenic; band power.
International Journal of Biomedical Engineering and Technology, 2018 Vol.28 No.2, pp.105 - 119
Received: 19 Oct 2016
Accepted: 13 Dec 2016
Published online: 30 Aug 2018 *