Figure 1.

A flowchart illustrating the main analytical process in the current study. Briefly, individual structural images were first segmented into gray matter, white matter and cerebrospinal fluid (A). The gray matter maps (smoothed and nonsmoothed) were then divided into different numbers of regions according to prior brain atlases (AAL and HOA) (B). For each region, the gray matter volume values within it were extracted and used to estimate the probability distribution function (C). Subsequently, the KL divergence‐based similarity was calculated between any pair of regions in their probability distribution functions, resulting in a similarity matrix (D). The resultant similarity matrix was further thresholded into both binary and weighted networks (E), which could be visualized as graphs (F). Finally, several graph‐based network measures were employed to topologically characterize the graphs at both global and nodal levels (G). AAL, Anatomical Automatic Labeling atlas; HOA, Harvard‐Oxford atlas; PCUN, precuneus; TPOmid, temporal pole: middle temporal gyrus; L, left; R, right.