Support vector machine classification of major depressive disorder using diffusion-weighted neuroimaging and graph...

Area of Science:

• Neuroscience
• Psychiatry
• Computational Biology

Background:

• Major depressive disorder (MDD) research increasingly focuses on brain network alterations.
• Diffusion-weighted imaging (DWI) and graph theory enable the study of brain fiber networks.

Purpose of the Study:

• To differentiate individuals with MDD from healthy controls using whole-brain graph metrics.
• To identify the most important graph metrics for classifying depression.
• To pinpoint specific brain regions with abnormal network connectivity in MDD.

Main Methods:

• Employed a machine learning approach, specifically support vector machines, for classification.
• Analyzed structural brain networks derived from diffusion-weighted imaging (DWI) tractography.
• Conducted global and local graph analyses to assess network properties and regional connectivity.

Main Results:

• Successfully classified depression using whole-brain graph metrics.
• Small-worldness emerged as the most significant metric for differentiating between depressed and healthy individuals.
• Identified abnormal network connectivity in the right pars orbitalis, right inferior parietal cortex, and left rostral anterior cingulate in MDD patients.

Conclusions:

• This study is the first to utilize structural global graph metrics for classifying individuals with depression.
• Findings underscore the potential of network properties in understanding MDD.
• Future research should explore network alterations across imaging modalities and correlate them with clinical factors.