ICA-based classification of MCI vs HC

Abstract

Compared to Alzheimer's disease(AD) patients, mild cognitive impairment(MCI) subjects are usually overlooked because of the cryptic features of the occurrence and development of disease. As a result, to as accurately as possible tell MCI subjects from healthy normal persons is of great importance and urgency. In this paper, we proposed a novel method based on independent component analysis (ICA) to analyze structural magnetic resonance imaging(MRI) data of 55 MCI subjects and age-matched 69 healthy controls. First, all these images are preprocessed with atlas adjustment and normalization. Then ICA is applied to extraction of features that are able to differentiate MCI subjects from healthy controls. 9 independent components are estimated using information criteria and finally drawn from the original data. On the basis of that, we construct the features for classification which are composed of both the extracted Independent components and some clinical examination values. And the final averaged classification accuracy was obtained with 82.70%. The experimental results show that the proposed method based on ICA is able to obtain highier classification accuracy of MCI vs HC than only ICs or clinical measures.