Publication

Abstract : Schizophrenia (SZ) is a brain disorder, which affects verbal, memory, perception, and decision-making functions of the brain. The anatomical and morphological abnormalities in human brain regions are analyzed using magnetic resonance imaging (MRI). In this work, the pattern variation in the ventricle region from schizophrenic MR brain images was analyzed using Hu moments with random forest classifier. The images are obtained from Centers of Biomedical Research Excellence (COBRE) database. Initially, the images are exposed to simultaneous bias correction and segmentation of ventricle region using multiplicative intrinsic component optimization method. Hu moments are extracted from the segmented ventricle region and are subjected to the random forest classifier to identify the disease condition. The outcomes of the analysis results show that the energy minimization method is capable to segment ventricle with high correlation index. The similarity measures such as Dice and Jaccard coefficient are high (0.967 and 0.936). Hu moment features could classify the normal and schizophrenia subjects better with an accuracy of 82.73% using random forest classifier compared to conventional classifiers. The area under region of convergence is found to be 0.827. The Hu moments extracted from the ventricle along with random forest classifier seem to be significant; hence, it may be clinically supportive in the diagnosis of schizophrenic subjects.