Classification of 3b alzheimer's mr images using voxel values in volumetric loss regions

Abstract

Alzheimer's Disease is a deadly neurological disease that begins with cognitive disorders and forgetfulness. Volumetric changes caused by the disease in the brain can be monitored with high resolution magnetic resonance images. In this study, volumetric losses occurring in gray matter and white matter regions were mapped by voxel-based morphometry method using 3D T1-weighted magnetic resonance images taken form OASIS database, and a decision support system was designed that classifies alzheimer's and normal magnetic resonance images with significant voxel values in these regions. SPM8, MRIcro programs and VBM8 library were used for inter-group voxel-based morphometry on magnetic resonance images. Gray matter and white matter regions were masked with binary masks obtained from volumetric loss maps. Significant data sets were created with voxel values corresponding to the same coordinate points from the areas under the mask in each gray matter and white matter image. With feature ranking methods, the data were ranked from the most meaningful feature to the most meaningless feature. The ranked features were given as input to the support vector machine using linear and rbf kernel with 10 fold cross validation. As a result of the experiments, the highest accuracy rates were found as 92.857% in gray matter classification and 79.286% in white matter classification with linear support vector machines based on t-test feature ranking. © 2020, TUBITAK. All rights reserved.