Statistical and Correlation Analysis

For each subject, the dFCD variance and sFCD maps were converted to z-values using Fisher z-transformation to enhance the normality of the data distribution. Then, all normalized images were smoothed with a 6 mm × 6 mm × 6 mm full-width at half-maximum (FWHM) Gaussian kernel. A two-sample t-test was employed on the FCD maps to analyze the differences between the MDD and HCs groups. Age, sex, years of education, and mean FD were used as covariates. In order to control the false positive rate, multiple comparisons were corrected using a Gaussian random field theory (GRF) approach at a voxel-wise threshold of p < 0.005 with the cluster-level threshold set to p < 0.05 (the cluster-level correction refers to family wise error), and cluster-extent threshold was set at 20 voxels.

The brain clusters showing significant global, intra-, and interhemispheric FCD differences between patients with MDD and HCs were defined as regions of interest (ROIs). The values of dFCD variance and sFCD at global and intra- and interhemispheric levels in all ROIs were extracted separately. A Pearson correlation analysis was performed to identify correlations of abnormal dFCD variance and sFCD with depressive symptom severity (p < 0.05, uncorrected).