FIGURE 1.

Framework for construction of high-order functional connectivity (FC) network. (1) Partition of the RS-fMRI time series into multiple overlapping segments of subseries applying sliding-window technique; (2) collection of low-order FC matrices, one for each subseries; (3) stack of all matrices of all subjects together to obtain correlation time series for each element; (4) application of the clustering algorithm to group all the correlation time series; (5) construction of high-order FC network, considering the mean correlation time series for each cluster as vertex and the pairwise Pearson’s correlation coefficient between each pair of vertices as weight; (6) calculation of local clustering coefficients; (7) selection of a discriminative feature subset from the local clustering coefficients; (8) implementation of support vector machine (SVM) model for classification. RS-fMRI, resting-state functional magnetic resonance imaging; FC, functional connectivity (reproduced with permission from Chen et al., 2016).