Ecologically-Valid Emotion Signatures Enhance Mood Disorder Diagnostics

Mood disorders, including Major Depressive Disorder (MDD) and Bipolar Disorder (BD), are highly prevalent conditions. These disorders are characterized by persistent emotional dysregulation and substantial functional impairments. Despite extensive neuroimaging research, reliable neurofunctional markers remains elusive. To address this gap, we propose a novel approach that utilizes Divergent Emotional Functional Networks (DEFN), derived from functional magnetic resonance imaging (fMRI) in naturalistic contexts.By integrating naturalistic emotion induction, dynamic functional connectivity (dFC), and machine learning, we identified emotion-specific functional patterns in healthy individuals with an accuracy of 83.99%. The DEFN was subsequently validated in clinical datasets, including a multi-site MDD cohort (Hiroshima University: MDDs = 63, HCs = 111; University of Tokyo: MDDs = 62, HCs = 96) and an independently BD cohort (BDs = 59, HCs = 50). Using static functional connectivity (sFC) and nested 10-fold cross-validation, DEFN-based models (MDD: 70.33%, BD: 75.18%) significantly outperformed baseline models in classifying patients and HCs (MDD: 70.33% vs. 57.58%; BD: 75.18% vs. 63.18%). Additionally, DEFN demonstrates highly reproducibility across age and sex, supporting the robustness of DEFN model. In conclusion, the DEFN approach presents a promising, reproducible, and clinically relevant neural marker for diagnosing, offering potential for more effective and timely interventions.