Abstract
Dynamic functional connectivity (DFC) analysis can capture time-varying properties of connectivity and may provide further information about transdiagnostic psychopathology across major psychiatric disorders. In this study, we used resting state functional MRI and a sliding-window method to study DFC in 150 schizophrenia (SZ), 100 bipolar disorder(BD), 150 major depressive disorder (MDD), and 210 healthy controls (HC). DFC were clustered into two functional connectivity states. Significant 4-group differences in DFC were found only in state 2. Post hoc analyses showed that transdiagnostic dysconnectivity among there disorders featured decreased connectivity within visual, somatomotor, salience and frontoparietal networks. Our results suggest that decreased connectivity within both lower-order (visual and somatomotor) and higher-order (salience and frontoparietal) networks may serve as transdiagnostic marker of these disorders, and that these dysconnectivity is state-dependent. Targeting these dysconnectivity may improve assessment and treatment for patients that having more than one of these disorders at the same time.
Footnotes
Revison 1: We have reprocess all our fmri data. This time we did not do the despiking. We used a rigorous method to control the head motion: subjects with greater head motion were excluded from this study (Please see Head motion control section of the revised manuscript). We discarded participants if they had mean framewise displacement (FD) values > 0.2 mm, if the outliers accounted for > 30% of all volumes (190 volumes), or if head motion exceeded 3 mm or 3 degree. According to these criteria, we excluded 41 healthy controls, 30 patients with schizophrenia, 18 patients with bipolar disorder and 25 patients with major depressive disorder. Besides, we treated the age, sex and mean FD as covariates, when we performed ANCOVA. Revison 2: This asymmetry is because we only extract the lower triangular matrix in the process of data analysis, and when we get results, we do not mirror the data of the lower triangular matrix into the upper triangular matrix. In this way, when sorting nodes and edges according to their brain network index, this asymmetry emerged. In the revised version, we mirror the lower triangular matrix into the upper triangular matrix (using the MATLAB command M = M + M'; M is the lower triangular matrix), which solves this bug. Revison 3: We reclustered the dynamic functional connectivity: "we used the Manhattan distance (L1 distance) as a similarity measure in clustering, as it has been demonstrated to be the most effective measure for high dimensional data. To reduce the computational demands and to diminish redundancy between windows, we first used the subject exemplars as a subset of windows with local maxima in functional connectivity variance to perform kmeans clustering with varying numbers of clusters k (from 2 to 10). The optimal number of clusters k = 2 was determined based on the silhouette criterion, a cluster validity index that reflects how similar a point is to other points in its own cluster compared to points in other clusters." Revison 4: According to this suggestion, we additionally showed group differences without statistical thresholding (Supplementary_material Figure S1), so that the interpretation is not completely driven by the choice of statistical threshold. Revison 5: We have presented the head motion information in Table 1. After discarded those participants with greater head motion, there was indeed no statistical difference between the four groups (ANOVA p < 0.05).