Integrating structural and functional brain features to classify major depressive disorder: a multi modal approach

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Major Depressive Disorder (MDD) significantly impacts psychosocial functioning and quality of life.
• Integrating diverse neuroimaging modalities is crucial for understanding complex psychiatric disorders like MDD.

Purpose of the Study:

• Investigate neural differences between MDD patients and healthy controls (HCs).
• Explore the relationship between neuroimaging components and Beck Depression Inventory (BDI-II) scores.
• Develop a generalized predictive model for MDD classification using machine learning.

Main Methods:

• Applied unsupervised Parallel Independent Component Analysis for multimodal data fusion (gray matter, white matter, ReHo).
• Analyzed neuroimaging data from 197 MDD patients and 172 HCs.
• Utilized a supervised Random Forest (RF) classifier for MDD diagnosis.

Main Results:

• Identified gray matter reductions in frontal lobe regions (e.g., anterior cingulate cortex) and white matter increases in the cerebellum and default mode network (DMN).
• Observed enhanced functional activity in dorsomedial prefrontal areas of the DMN.
• Found significant correlations between identified networks and BDI scores.
• Achieved 75.68% accuracy in distinguishing MDD patients from HCs using the RF classifier, highlighting key classification features.

Conclusions:

• Multimodal, data-driven approaches are valuable for uncovering the neural underpinnings of MDD.
• These findings support the development of precision diagnostic tools for psychiatric disorders.