Two approaches to selecting set of voxels for the diagnosis of Alzheimerʼs disease using brain SPECT images

Abstract

This paper presents a computer-aided diagnosis technique for improving the accuracy of the early diagnosis of the Alzheimer type dementia. The first proposed methodology is based on the selection of those voxels which present a greater difference between Controls and Alzheimer type dementia patients. The mean value of the intensities of the selected voxels is used as features for different classifiers. The proposed methodology reaches an accuracy of 89% in the classification task. A second criterion is chosen to select voxels based on considering those which present not only greater overall difference between both modalities (Controls and Alzheimer) but also present lower dispersion. The classification accuracy using this second condition increases to 94%.