Simultaneous Estimation of Low- and High-Order Functional Connectivity for Identifying Mild Cognitive Impairment

Abstract

Functional connectivity (FC) network has been becoming an increasingly useful tool for understanding the cerebral working mechanism and mining sensitive biomarkers for neural/mental disease diagnosis. Currently, Pearson's Correlation (PC) is the simplest and most commonly used scheme in FC estimation. Despite its empirical effectiveness, PC only encodes the low-order (i.e., second-order) statistics by calculating the pairwise correlations between network nodes (brain regions), which fails to capture the high-order information involved in FC (e.g., the correlations among different edges in a network). To address this issue, we propose a novel FC estimation method based on Matrix Variate Normal Distribution (MVND), which can capture both low- and high-order correlations simultaneously with a clear mathematical interpretability. Specifically, we first generate a set of BOLD subseries by the sliding window scheme, and for each subseries we construct a temporal FC network by PC. Then, we employ the constructed FC networks as samples to estimate the final low- and high-order FC networks by maximizing the likelihood of MVND. To illustrate the effectiveness of the proposed method, we conduct experiments to identify subjects with Mild Cognitive Impairment (MCI) from Normal Controls (NCs). Experimental results show that the fusion of low- and high-order FCs can generally help to improve the final classification performance, even though the high-order FC may contain less discriminative information than its low-order counterpart. Importantly, the proposed method for simultaneous estimation of low- and high-order FCs can achieve better classification performance than the two baseline methods, i.e., the original PC method and a recent high-order FC estimation method.

A toolbox for brain network construction and classification (BrainNetClass)

“…In SR, the BOLD signal of a brain region is represented by a linear combination of the signals from a few of other regions. To measure higher‐level relationship between two ROIs, “high‐order” FC (HOFC) was proposed to define inter‐regional relationship by not measuring “low‐level” features (i.e., BOLD signals) but various “high‐level” features, which provides complementary information to the traditional “low‐order” (PC‐based) brain networks and indicates improved performance in disease diagnosis (Zhang, Chen, et al, ; Zhang, Chen, et al, ; Zhang, Shen, & Lin, ; Zhou, Zhang, Teng, Qiao, & Shen, ; Zhou, Qiao, Li, Zhang, & Shen, ). On the other hand, to avoid too sparse (thus may miss disease‐related FC alterations) brain networks derived from SR and respect inherent structures in the brain network, recent research has been designing new regularization terms to build more biologically meaningful brain networks, resulting in many variants of the SR methods (Qiao et al, ; Yu et al, ; Zhang, Zhang, et al, ).…”

Towards a Better Estimation of Functional Brain Network for Mild Cognitive Impairment Identification: A Transfer Learning View

“…In fact, many recently-proposed FBN estimation models [35][36][37][38] can be unified under this regularized framework with different design of the two terms in Eq. (5).…”

Group Similarity Constraint Functional Brain Network Estimation for Mild Congititive Impairment Classification

“…In particular, 137 participants, including 68 MCIs and 69 NCs, are adopted in this experiment, which is also similar as (Zhou et al 2018). The scanning parameter includes: TR/TE = 3000/30mm, flip angle = 80, imaging matrix=64×64, 48 slices, 140 volumes, and voxel thickness = 3.3mm.…”

“…At last, we put these time series into a data matrix X ∈ R 137×116 . For more details, please refer to (Zhou et al 2018).…”

“…In fact, most of the recently-proposed FBN estimation models (Higgins, Kundu and Guo 2018, Li, Zhu and Fan 2018, Zhou et al 2018) can be unified under this regularized framework with different design of the two terms in Eq. (5).…”