Differential Dynamical Pattern of Regional Homogeneity in Bipolar and Unipolar Depression: A Preliminary Resting-State fMRI Study

Background: Bipolar depression (BD) and unipolar depression (UD) are both characterized by depressive moods, which are difficult to distinguish in clinical practice. Human brain activity is time-varying and dynamic. Investigating dynamical pattern alterations of depressed brains can provide deep insights into the pathophysiological features of depression. This study aimed to explore similar and different abnormal dynamic patterns between BD and UD.Methods: Brain resting-state functional magnetic resonance imaging data were acquired from 36 patients with BD type I (BD-I), 38 patients with UD, and 42 healthy controls (HCs). Analysis of covariance was adopted to examine the differential pattern of the dynamical regional homogeneity (dReHo) temporal variability across 3 groups, with gender, age, and education level as covariates. Post-hoc analyses were employed to obtain the different dynamic characteristics between any 2 groups. We further applied the machine-learning methods to classify BD-I from UD by using the detected distinct dReHo pattern.Results: Compared with patients with UD, patients with BD-I demonstrated decreased dReHo variability in the right postcentral gyrus and right parahippocampal gyrus. By using the dReHo variability pattern of these two regions as features, we achieved the 91.89% accuracy and 0.92 area under curve in classifying BD-I from UD. Relative to HCs, patients with UD showed increased dReHo variability in the right postcentral gyrus, while there were no dReHo variability differences in patients with BD-I.Conclusions: The results of this study mainly report the differential dynamic pattern of the regional activity between BD-I and UD, particular in the mesolimbic system, and show its promising potential in assisting the diagnosis of these two depression groups.

Automated summary

This study compared the dynamic regional homogeneity variability among patients with BD-I, UD, and healthy controls using resting-state functional MRI data. The results showed that patients with BD-I had decreased dReHo variability in certain brain regions compared to patients with UD. Machine-learning methods achieved high accuracy in classifying BD-I from UD based on these distinct dReHo patterns. Overall, the study highlights the potential of dynamic patterns in assisting the diagnosis of BD and UD.