Structural and Functional Brain Covariance in Healthy Controls

We were interested in how covariance within and between for major networks differed between V and NV. For that aim, the functional connectivity toolbox CONN was used (61 (link)). CONN’s standard network atlas was based on an independent component analysis of the functional resting state data of a large sample of healthy adults (61 (link), 62 (link)). Although variances in the brain structure are expected in healthy controls and patient groups, applying the atlas information based on healthy adults for the investigation of patient groups is considered valid because previous studies have shown differences in the DNM, SN, and FPN based on different whole brain nodes and seeds [for a meta-analysis see Koch et al. (63 (link))] suggesting robust group differences in these networks despite of potential structural differences. The atlas provides an established brain parcellation that divided the DMN, SN, DAN, and FPN into 19 spatially distinct network nodes, which were parts of the brain networks (Figure 1). The DMN covered the medial prefrontal cortex (MPFC), the bilateral lateral parietal cortex (LPCs), and the precuneus (PCUN). The SN included the anterior cingulate cortex (ACC) as well as the bilateral anterior insula (AIs), the rostral prefrontal cortex (RPFCs), and the supramarginal gyrus (SMGs). The DAN consisted bilaterally of frontal eye fields (FEFs) and the intraparietal sulci (IPSs). The FPN comprised both the right and left lateral prefrontal cortex (LPFCs) and the posterior parietal cortex (PPCs). The 19 investigated network nodes served as ROIs and were used to extract structural and functional brain information from individuals in the V and NV groups. For each participant, brain data was averaged across all voxels belonging to a particular ROI. This yielded individual average GM volumes, average GM density and average functional resting state time series for each ROI. The extracted GM volumes, densities and time series were z-standardized individually. This z-standardization mainly served two purposes in the following analyses: (i) to ensure the comparability of ROIs, and (ii) to enable the interpretation of covariance measures as correlation (= normalized covariance). To avoid potential confounding effects in the brain data, we accounted for sex, age, MINI diagnosis, antidepressants, and the number of other psychotropic drugs. In the structural analyses, we also accounted for total intracranial volume. Numerical confounds were z-standardized, while scale confounds were dummy encoded. Deconfounding on the group level was performed for time series in CONN, while for GM volumes and densities it was done with NiftiMapsMasker from nilearn package (64 ). The extracted network information served as input for the estimation of structural covariance and functional connectivity matrices in each group.