Multi-scale Graph Convolutional Network for Mild Cognitive Impairment Detection

Abstract

Mild cognitive impairment (MCI) is an early stage of Alzheimer’s disease (AD), which is also the best time for treatment. However, existing methods only consider neuroimaging features learned from group relationships instead of the subjects’ individual features. Such methods ignore demographic relationships (i.e., non-image information). In this paper, we propose a novel method based on multi-scale graph convolutional network (MS-GCN) via inception module, which combines image and non-image information for MCI detection. Specifically, since the brain has the characteristics of high-order interactions, we first analyze the dynamic high-order features of resting functional magnetic resonance imaging (rs-fMRI) time series and construct a dynamic high-order brain functional connectivity network (DH-FCN). To get more effective features and further improve the detection performance, we extract the local weighted clustering coefficients from the original DH-FCN. Then, gender and age information are combined with the neuroimaging data to build a graph. Finally, we perform the detection using the MS-GCN, and validate the proposed method on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. The experimental results demonstrate that our proposed method can achieve remarkable MCI detection performance.

This work was supported partly by National Natural Science Foundation of China (Nos. 61871274, 61801305 and 81571758), National Natural Science Foundation of Guangdong Province (No. 2017A030313377), Guangdong Pearl River Talents Plan (2016ZT06S220), Shenzhen Peacock Plan (Nos. KQTD2016053112051497 and KQTD2015033016 104926), and Shenzhen Key Basic Research Project (Nos. JCYJ2017 0413152804728, JCYJ20180507184647636, JCYJ20170818142347251 and JCYJ20170818094109846).