Abstract:BACKGROUND: The present study aimed to explore the difference in the brain function and structure between patients with major depressive disorder (MDD) and healthy controls (HCs) using two-center and multi-modal MRI data, which would be helpful to investigate the pathogenesis of MDD.METHODS: The subjects were collected from two hospitals. One including 140 patients with MDD and 138 HCs was used as primary cohort. Another one including 29 patients with MDD and 52 HCs was used as validation cohort. Functional and structural magnetic resonance images (MRI) were acquired to extract four types of features: functional connectivity (FC), amplitude of low-frequency fluctuations (ALFF), regional homogeneity (ReHo), and gray matter volume (GMV). Then classifiers using different combinations among the four types of selected features were respectively built to discriminate patients from HCs. Different templates were applied and the results under different templates were compared.RESULTS: The classifier built with the combination of FC, ALFF, and GMV under the AAL template discriminated patients from HCs with the best performance (AUC=0.916, ACC=84.8%). The regions selected in all the different templates were mainly located in the default mode network, affective network, prefrontal cortex.LIMITATIONS: First, the sample size of the validation cohort was limited. Second, diffusion tensor imaging data were not collected.CONCLUSION: The performance of classifier was improved by using multi-modal MRI imaging. Different templates would be suitable for different types of analysis. The regions selected in all the different templates are possibly the core regions to investigate the pathophysiology of MDD.

A two-center radiomic analysis for differentiating major depressive disorder using multi-modality MRI data under different parcellation methods