Predicting conversion from MCI to AD by integrating rs-fMRI and structural MRI

Abstract

Structural MRI (sMRI) and resting-state functional MRI (rs-fMRI) have provided promising results in the diagnosis of Alzheimer's disease (AD), though the utility of integrating sMRI with rs-fMRI has not been explored thoroughly. We investigated the performances of rs-fMRI and sMRI in single modality and multi-modality approaches for classifying patients with mild cognitive impairment (MCI) who progress to probable AD-MCI converter (MCI-C) from those with MCI who do not progress to probable AD-MCI non-converter (MCI-NC). The cortical and subcortical measurements, e.g. cortical thickness, extracted from sMRI and graph measures extracted from rs-fMRI functional connectivity were used as features in our algorithm. We trained and tested a support vector machine to classify MCI-C from MCI-NC using rs-fMRI and sMRI features. Our algorithm for classifying MCI-C and MCI-NC utilized a small number of optimal features and achieved accuracies of 89% for sMRI, 93% for rs-fMRI, and 97% for the combination of sMRI with rs-fMRI. To our knowledge, this is the first study that investigated integration of rs-fMRI and sMRI for identification of the early stage of AD. Our findings shed light on integration of sMRI with rs-fMRI for identification of the early stages of AD.

Keywords: Alzheimer's disease (AD); Graph theory; Machine learning approach; Mild cognitive impairment (MCI); Resting-state fMRI; Structural MRI.