Discriminating bipolar depression from major depressive disorder using functional and microstructural signatures of...

Bipolar depression (BD-d) and major depressive disorder (MDD) share overlapping clinical symptoms, posing a significant diagnostic challenge. The inferior fronto-occipital fasciculus (IFOF) is a critical long-range tract that integrates cognitive and affective processes across brain networks. However, it remains unexplored whether bilateral IFOF abnormalities differ between BD-d and MDD and have utility for machine learning classification.

Processed data from 47 BD-d, 56 MDD, and 44 healthy controls (HCs) were analyzed for structural and functional alterations in bilateral IFOF metrics (including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), and amplitude of low-frequency fluctuations (ALFF)) were assessed at both the tract and node levels. Five machine learning models were then employed to evaluate the diagnostic utility of significant features.

Tract-level analysis revealed significant alterations in the right IFOF (rIFOF), characterized by decreased FA and AD alongside increased RD among patient groups. Functional ALFF demonstrated opposite changes, being elevated in BD-d but reduced in MDD. Node-wise analysis revealed that structural and functional alterations co-localized within temporal segments. Besides, RD in the rIFOF correlated with HAMD scores in MDD, linking structure to clinical severity, while a k-nearest neighbor (KNN) classifier based on tract-level metrics distinguished BD-d from MDD with 86% accuracy.

This multimodal study reveals both shared and distinct segmental alterations in the rIFOF within BD-d and MDD, highlighting a microstructure-function dissociation pattern. These alterations correlate with clinical severity and enable high-accuracy diagnostic differentiation, providing objective biomarkers for understanding depression pathophysiology and improving differential diagnosis.