Abstract:Background Neurobiological heterogeneity in depression remains largely unknown, leading to inconsistent neuroimaging findings. Methods Here, we adopted a novel proposed machine learning method ground on gray matter volumes (GMVs) to investigate neuroanatomical subtypes of first-episode treatment-naïve depression. GMVs were obtained from high-resolution T1-weighted images of 195 patients with first-episode, treatment-naïve depression and 78 matched healthy controls (HCs). Then we explored distinct subtypes of depression by employing heterogeneity through discriminative analysis (HYDRA) with regional GMVs as features. Results Two prominently divergent subtypes of first-episode depression were identified, exhibiting opposite structural alterations compared with HCs but no different demographic features. Subtype 1 presented widespread increased GMVs mainly located in frontal, parietal, temporal cortex and partially located in limbic system. Subtype 2 presented widespread decreased GMVs mainly located in thalamus, cerebellum, limbic system and partially located in frontal, parietal, temporal cortex. Subtype 2 had smaller TIV and longer illness duration than Subtype 1. And TIV in Subtype 1 was positively correlated with age of onset while not in Subtype 2, probably implying the different potential neuropathological mechanisms. Limitations Despite results obtained in this study were validated by employing another brain atlas, the conclusions were acquired from a single dataset. Conclusions This study revealed two distinguishing neuroanatomical subtypes of first-episode depression, which provides new insights into underlying biological mechanisms of the heterogeneity in depression and might be helpful for accurate clinical diagnosis and future treatment.

Identifying two distinct neuroanatomical subtypes of first-episode depression using heterogeneity through discriminative analysis