چكيده به لاتين
It is known that some impairments due to brain disorder can be observed through DTI image processing from structural aspect. Therefore, improvements in Diffusion Tensor Imaging (DTI) images and brain connectivity knowledge can be very helpful for prognosis of some diseases such as Alzeimer. Despite medical improvements, there is no exact diagnosis and treatment for Alzeimer’s disease, therefore the importance of finding a proper classifier for prognosis can be figured out. In this study, the ExploreDTI software was used for preprocessing of 60 data (20 equal male and female data for each controlled (CNs), Mild Cognitive Impairment (MCI) and Alzeimer (ADs) group) extracted from the ADNI Database (adni.loni.usc.edu). Data analyses were carried out on certain features viz. Fractional Anisotropy (FA), Mean Diffusivity (MD), Radial Diffusivity (RD) and Brain Volume between the ADs and the CNs, and also between males and females on each group separately; on parietal lobe on ADs and CNs by along-tract analysis; and on the three graph features extracted from Brain Connectivity Tool (BCT) based on the graph theory. Normality test done by Kolmogrov-Smirnov revealed non-normality of the data. Wilcoxon Signed Rank test and Kruskal-Wallis tests were used to consider statistical significance between the two and the three groups, respectively. Finally, through the main purpose of the project and based on the selected features of the brain network, some Support Vector Machine (SVM) classifiers were instructed with the accuracies of 63.33%, 65.00% and 65.83%, comparing with 61.9% mentioned in the literature. These classifiers may discriminate well among the three groups of CNs, MCI and ADs simultaneously. A flow chart of the algorithm and the necessary recommendations for the future investigations have been considered in the study.
Keywords: image processing, DTI, brain network, BCT, graph theory, SVM classifier, MCI, ADs
