Altered brain network topology in adolescents with major depressive disorder and bipolar disorder: A resting-state fMRI graph-theoretical and machine learning study.

BACKGROUND: Adolescents with major depressive disorder (MDD) and bipolar disorder (BD) share substantial clinical overlap and elevated suicide risk, yet the neurobiological distinctions between these disorders and their associations with suicidality remain incompletely understood. This study investigated functional connectome differences between adolescent MDD and BD and examined associations with suicide attempts (SA). METHODS: Resting-state fMRI data were acquired from 125 adolescents aged 12-19 years (48 MDD, 36 BD, 41 healthy controls). We used graph-theoretical analysis to investigate group differences in functional brain networks, and machine learning models were applied to functional network data to distinguish between MDD and BD. RESULTS: Compared with MDD and healthy controls, adolescents with BD exhibited a global shift toward network randomization, characterized by a lower clustering coefficient and widespread reductions in nodal centrality across hubs of the DMN, SN, and CEN. In contrast, MDD was characterized by preserved global topology but focal nodal alterations. Within the MDD group, greater suicidal-ideation severity was associated with lower nodal efficiency in the insula and supramarginal gyrus. A support vector machine classifier distinguished MDD from BD with 88.24% (p < 0.001) accuracy, with features from the insula and cingulate gyrus being highly informative. CONCLUSIONS: Adolescent MDD and BD showed distinct patterns of functional network disruption, with BD showing global network disorganization and MDD showing more localized disruptions. Alterations involving the insula and supramarginal gyrus may be relevant to suicidality in adolescent MDD, and network-based features may aid in distinguishing MDD from BD.