Fig. 1
From: Learning stable and predictive network-based patterns of schizophrenia and its clinical symptoms
A stable subset of statistically significant (Bonferroni-surviving) site-standardized (ss) log-degree features, across 95 data subsets corresponding to leave-subject-out CV. For each data subset, a two-sample t-test was performed to select the subset of features surviving Bonferroni correction. The intersection of all such subsets contains 426 “stable Bonferroni” voxels (vs. 700 that survived Bonferroni on the whole dataset but were not necessarily stable). Note that mean values of the site-standardized log-degree, as well as the corresponding log-degree features, at all highlighted voxels, were higher in schizophrenic group than in the control group (see Supplementary Fig. S3). The numbered arrows in the figure point to the most-significant (smallest p-value) largest clusters, which included: (1) left BA6 (precentral gyrus, inferior frontal gyrus); (2) left BA6 (middle frontal gyrus); (3) left BA39 (lateral occipital cortex, superior division); (4) left BA30 (cingulate gyrus, posterior division); and (5) left BA7 (precuneus). Also, see Supplementary Table S5 for the MNI coordinates of the clusters’ local maxima). For visualization purposes, the original-resolution statistical maps are upsampled (to 2 × 2 × 2 × mm), thresholded based on intensity and cluster size, and smoothed using a Gaussian kernel (5 mm FWHH) in bspmview.28 The original statistical p-map is provided in supplementary information below Supplementary Table S5
