Alzheimer's disease diagnosis in MR images using statistical methods

Alzheimer's disease is one of the neurodegenerative dementia which results in dramatic health consequences and has socio-economic implications. Early and accurate diagnosis of Alzheimer's disease is necessary to provide proper treatment to the patients. Recently, several Computed-Aided-Diagnosis systems has been developed for Alzheimer's disease detection from MR images. Although univariate approaches are most widely used, recent investigations focuses on multivariate approaches which deals with whole image as one observation. In the proposed method, detection is done by efficiently segmenting the gray matter region using Gaussian Mixture Models and then extracting the score vectors based on Partial Least Squares. This relieves the small sample size problem and then classification is done by Support Vector Machine, a supervised learning mechanism. The efficiency of the proposed method is compared with existing methods and is found superior.