Treatment-naïve first episode depression classification based on high-order brain functional network

Area of Science:

• Neuroimaging
• Computational Psychiatry
• Brain Networks

Background:

• Resting-state functional MRI (rs-fMRI) shows promise for studying major depressive disorder (MDD).
• Individualized diagnosis of MDD using rs-fMRI remains challenging.
• Existing studies often focus on low-order functional connectivity (FC).

Purpose of the Study:

• To develop and evaluate a computer-aided diagnosis framework for classifying first-episode depression (FED).
• To compare the diagnostic performance of low-order (LON) versus high-order (HON) FC network features.
• To explore the potential of integrated classification models for enhanced FED diagnosis.

Main Methods:

• Utilized a computer-aided diagnosis framework for FED classification.
• Extracted features from both traditional low-order networks (LON) and dynamic high-order networks (HON).
• Compared classification accuracy of HON features (C_HON) against LON features (C_LON) and an integrated model.

Main Results:

• High-order network features significantly improved diagnostic accuracy (82.47%) compared to low-order features (67.53%).
• An integrated classification model achieved 83.77% accuracy.
• Key diagnostic regions involved high-order cognitive function networks, including the cerebellum (vermis and crus II).

Conclusions:

• High-order functional connectivity features enhance the classification of major depressive disorder.
• Cerebro-cerebellar interactions are crucial in the pathophysiology of MDD.
• The findings suggest novel imaging biomarkers for MDD diagnosis.