Hierarchical Synchronization Estimation of Low- and High-Order Functional Connectivity Based on Sub-Network Division for the Diagnosis of Autism Spectrum Disorder.

Feng Zhao,Xiaobo Chen,Zhongwei Han,Dapeng Cheng,Deming Fan,Yuan Li,Ning Mao,Peiqiang Liu

doi:10.3389/fnins.2021.810431

Abstract

Functional connectivity network (FCN) calculated by resting-state functional magnetic resonance imaging (rs-fMRI) plays an increasingly important role in the exploration of neurologic and mental diseases. Among the presented researches, the method of constructing FCN based on Matrix Variate Normal Distribution (MVND) theory provides a novel perspective, which can capture both low- and high-order correlations simultaneously with a clear mathematical interpretability. However, when fitting MVND model, the dimension of the parameters (i.e., population mean and population covariance) to be estimated is too high, but the number of samples is relatively quite small, which is insufficient to achieve accurate fitting. To address the issue, we divide the brain network into several sub-networks, and then the MVND based FCN construction algorithm is implemented in each sub-network, thus the spatial dimension of MVND is reduced and more accurate estimates of low- and high-order FCNs is obtained. Furthermore, for making up the functional connectivity which is lost because of the sub-network division, the rs-fMRI mean series of all sub-networks are calculated, and the low- and high-order FCN across sub-networks are estimated with the MVND based FCN construction method. In order to prove the superiority and effectiveness of this method, we design and conduct classification experiments on ASD patients and normal controls. The experimental results show that the classification accuracy of “hierarchical sub-network method” is greatly improved, and the sub-network found most related to ASD in our experiment is consistent with other related medical researches.

Highlights

We compare the performance of the “hierarchical sub-network method” and the “fully network Functional connectivity network (FCN) method” in the autism spectrum disorder (ASD) classification experiment
In order to analyze the influence of different types of features in two compared methods on the experimental results, we carry out separate experiments on intra-sub-network high-order features, intra-sub-network loworder features, inter-sub-network high-order features, inter-subnetwork low-order features, fully network high-order features and fully network low-order features
This paper proposes a new strategy for mental illness diagnosis based on FCN

Summary

Introduction

Functional connectivity networks (FCN), usually calculated from resting-state functional magnetic resonance imaging (rsfMRI), using blood oxygenation level dependent (BOLD) signals as neurophysiological indicators, are playing an increasingly important role in exploring the working mechanism of the brain and investigating the brain’s functional variations of some mental disorders, such as autism spectrum disorder (ASD) (Felouat and Oukid-Khouas, 2020; Sun et al, 2021), major depressive disorder (Mousavian et al, 2020), Alzheimer’s disease (Jones et al, 2012; Wang et al, 2017), and its early stage, i.e., mild cognitive impairment (Chen et al, 2016; Zhang et al, 2020), et al.FCN is a weighted network based on the graph theory, which takes the regions of interest (ROIs) in the brain as the nodes, the correlation of the rs-fMRI time series between different ROIs as the functional connectivity (FC) and the FC strength as the weight of the edge (Smith et al, 2013). Based on the sliding window strategy, the rs-fMRI time series are divided into a set of short time series fragments, and the conventional FCN is constructed on each fragment This can capture dynamic FC changes over time to a certain extent (Kudela et al, 2017). The second class of FCN model is the so-called “high-order FCN” (Song et al, 2020) that can capture deeper brain information by designing FC model of multiple ROIs. For example, on the basis of dynamic FCN, Chen et al (2016) and Zhao et al (2018) took each FC time series as the network node and the correlation coefficient of FC time series of each ROI pair as the edge weight to construct a high-order FCN, which fills the interaction between paired ROI and other ROI pairs. Since this paper mainly focuses on the applications of high-order FCNs for diagnosis, detailed discussion about general biological meanings of high-order FC networks can be found in these published works

Methods

Results

Discussion

Conclusion