Distinct brain functional patterns across the entire disease course between bipolar disorder and major depressive...

Bipolar disorder (BD) and major depressive disorder (MDD) are frequently misdiagnosed, while most neuroimaging studies have focused on single illness stages rather than the entire disease course. This study applied Gaussian mixture model (GMM) clustering to resting-state functional magnetic resonance imaging (rs-fMRI) features across depressive episode, partial remission, and complete remission stages to model the stage related differences of differential patterns between the two affective disorders.

A total of 254 participants (109 BD and 145 MDD) aged 15-50 years underwent rs-fMRI scanning. Rs-fMRI data were preprocessed using DPABI, and dynamic fractional amplitude of low-frequency fluctuations (dfALFF) was calculated. Based on dfALFF features, Gaussian mixture model (GMM) unsupervised clustering was applied to cluster the two affective disorders across different illness stages, followed by clustering performance analysis of stage related differences, feature importance analysis, and construction of connection curves for the two affective disorders.

Clustering accuracies during depressive episode, partial remission, and complete remission stages were 71.83%, 70.69%, and 53.42%, respectively. Discriminative brain regions during depressive episode and partial remission stages mainly involved frontal, temporal, parietal, limbic, cerebellar, subcortical, and occipital regions. No meaningful discriminative brain regions were identified during complete remission because BD and MDD could not be effectively distinguished.

Brain functional patterns are distinguishable during depressive episodes and partial remission but become indistinguishable during complete remission. This study simulated differential diagnostic model patterns across disease course for two disorders, offering a novel perspective.