Graph Convolutional Network Analysis for Mild Cognitive Impairment Prediction

Abstract

Mild cognitive impairment (MCI) is an early stage of Alzheimer's disease (AD), which is a neurodegenerative disease. Functional connectivity networks (FCN) provide an effective method for analyzing brain functional regions connectivity. However, most methods only considered the neuroimaging information and focused on group relationship without the subjects' individual features, and ignored the demographic relationship. To handle it, in this paper, we introduce a novel method based on graph convolutional networks (GCN), which combines image and other information for MCI prediction tasks. The proposed model is capable of representing the individual features and data associations among subjects from potentially populations simultaneously. Specifically, we use different collection devices and gender information to build a graph called MCI-graph and modify convolutional neural networks (CNN) to construct GCN for MCI prediction. The experimental results demonstrate that our proposed method has achieved remarkable prediction performance.