Tensor factorization-based classification of Alzheimer's disease vs healthy controls

Abstract

So far there is no accurate diagnosis and effective treatment for Alzheimer's disease (AD), which is the most common neurodegenerative illness. Earlier diagnosis and prevention for AD are important goals for many researchers. In this paper, we propose a computational model, based on Tensor decomposition such as the Non-negative Multi-way Factorization (NMWF), to deal with structural magnetic resonance images (sMRI) data of 126 AD patients and 179 healthy controls (HC) at two time points of baseline and one-year. The NMWF algorithm is applied to sMRI data tensor with five factors of 3D tensor-sMRI, patient, and time to find the sMRI representation basis. The fourth and fifth factors then are fed into a classifier based on support vector machine to classify MR images of AD and HC. The experimental results on the sMRI data showed that NMWF-based method has well feasibility for discerning AD from HC from ADNI database.