Background: Bipolar disorder (BD) is a heterogeneous psychiatric illness marked by dynamic mood states, including manic (BipM), depressive (BipD), mixed (mBD), and remitted (rBD) episodes. These clinical fluctuations are accompanied by widespread functional disruptions in the brain. However, the shared and individual specific neural mechanisms across distinct episodes of BD remain poorly understood. Methods: We analyzed resting state fMRI data from 190 participants (BD patients in four episodes and healthy controls) using edge centric functional connectomes (eFC), which capture time resolved cofluctuations between brain regions. A common orthogonal basis extraction (COBE) algorithm was applied to decompose individual eFC matrices into shared and individual specific subspaces. We characterized the spatial topology, genetic relevance, and circuit level correlates of the shared component. Dynamic properties (entropy, identifiability) and symptom prediction models were assessed using entropy metrics, intraclass correlation coefficients, and support vector regression. Results: The shared eFC pattern was stable across participants, aligned with the sensory association gradient, and exhibited significant heritability and test retest reliability . Entropy of individual loadings increased with illness duration and was significantly elevated in BD, particularly in mBD. Microcircuit modeling revealed that this shared pattern was inversely related to external input strength , indicating intrinsic network dominance. mBD was associated with globally elevated eFC entropy and markedly reduced fingerprint stability. Symptom severity (HDRS, YMRS, HAMA) was significantly predicted from individual network topographies across BD phases, highlighting clinically meaningful dynamic signatures. Conclusion: Our findings demonstrate that BD episodes are underpinned by a conserved functional scaffold and distinct individual specific neural fingerprints. Edge centric dynamics especially those derived from individual specific decompositions offer robust biomarkers for mood state characterization and symptom severity, and may facilitate future personalized interventions in BD.
Liu et al. (Fri,) studied this question.