T-test feature ranking based 3D MR classification with VBM mask

The monitoring of volumetric changes in the brain during neurological diseases is done with three dimensional structural MR images. Numerical methods are needed to evaluate the volumetric difference between healthy and diseased MR image groups. Voxel based morphometry is a numerical method used to perform inter-group and intra-group analysis of MR images. With this method, the volume differences between the groups can be analyzed in the coordinate dimension. In this study, volumetric differences between Alzheimer and Normal MR images were investigated by voxel based morphometry. From these differences, three dimensional volume loss masks of groups were obtained. The mask of loss was masked with the gray matter area where the basic volume losses of Alzheimer's begin. In each gray matter image, the voxel values corresponding to the same coordinates as the loss mask were transformed into a vector. The voxel values corresponding to the same point as the loss point were ranked by t-test and the most significant coordinate values were ranked. After this ranking, classification was done with support vector machines using voxel values. As a result of this model, the number of voxel coordinates giving the best classification was determined and the performance of the method was measured.