Identification Of Depression Subtypes And Related Brain Magnetic Resonance Biomarkers

Objective:Data-driven functional network connectivity(FNC)data and multi-modal neuroimaging data were used to perform subtyping analysis on first-episode and medication-na(?)ve major depressive disorder(MDD)to determine subtypes of depression with distinct clinical features and related neuroimaging features.Materials and Methods:We recruited 115 first-episode and medication-na(?)ve MDD patients and 129 matched healthy controls.First,independent component analysis(ICA)was used to obtain FNC data,and then Fresurfer,CONN,and MRtrix3 software were used to obtain multimodal neuroimaging data including structural magnetic resonance imaging(sMRI)data,functional magnetic resonance imaging(fMRI)data and diffusional magnetic resonance imaging(dMRI)data.For FNC data,we used regularized canonical correlation analysis(r CCA)methods to obtain low-dimensional data that can reflect the association between FNC and clinical symptoms.For multimodal neuroimaging data,we use principal component analysis(PCA)combined with r CCA method to obtain low-dimensional data that can reflect the association between multimodal neuroimaging and clinical symptoms.Then,these low-dimensional data were analyzed by unsupervised clustering method for MDD subtypes.Finally,for FNC data,we compared the differences in clinical symptoms and imaging characteristics between the subtypes and analyzed the correlation between each other.For multimodal neuroimaging data,we used extreme gradient boosting(XGBoost)method to obtain the more important imaging features that distinguish each subgroup and analyzed the correlation between these neuroimaging features and clinical features.Results:For the FNC data,we identified two subtypes of MDD: insomnia-dominated subtype 1 and melancholia-dominated subtype 2.Subtype 1 was characterized by abnormal hyperconnectivity within ventral attention network(VAN),which was associated with poor sleep maintenance.Subtype 2 was characterized by abnormal hypoconnectivity between subcortical network(SC)and the dorsal attention network(DAN).For multimodal neuroimaging data,we also identified two subtypes of MDD:somatic-dominated subtype 1 and subtype 2 with typical depressive symptoms.In subtype 1,fears and suicide were negatively related with the connectivity between language network(LAN)and cingulo-opercular network(CON),between sensorimotor network(SMN)and CON,respectively.In subtype 2,anxiety(psychol)and autonomic were positively related to the connectivity between CON and secondary visual network 2(VIS2),between default mode network(DMN)and primary visual network 1(VIS1),while somatic-sensory was negatively correlated with connectivity between SMN and CON.Conlusion:Subtyping analysis studies based on single-modality(FNC)and multimodal neuroimaging data have identified two MDD subtypes with specific clinical symptoms and neuroimaging characteristics.These findings promote understanding of the clinical and biological heterogeneity of MDD,offering a pathway for defining categorical subtypes of illness via consideration of both clinical and biological features.