Classification of bipolar disorders using the multilayer modularity in dynamic minimum spanning tree from resting...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Network Science

Background:

• Bipolar disorder (BD) diagnosis relies heavily on clinical observation, often limiting accuracy.
• Objective diagnostic tools for BD are needed to improve early and accurate detection.
• Functional brain connectivity patterns may offer insights into BD pathophysiology.

Purpose of the Study:

• To develop a novel diagnostic framework for bipolar disorder (BD) using multilayer modularity in dynamic minimum spanning trees (MST).
• To assess the accuracy of this framework in discriminating between BD patients and healthy controls (HC).
• To identify key brain networks involved in the classification of BD.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (fMRI) data from 45 un-medicated BD patients and 47 HC.
• Constructed dynamic MSTs using a sliding window approach.
• Applied multilayer modularity analysis to module allegiance for classifier training and explored variations in FC estimators and MST scales.

Main Results:

• Multilayer modularity in dynamic MST is a non-random brain process.
• The developed model achieved an 83.70% accuracy in identifying BD patients.
• The default mode network, subcortical network (SubC), and attention network were crucial for classification.

Conclusions:

• Multilayer modularity in dynamic MST effectively differentiates between BD patients and HC.
• This method presents a promising novel diagnostic tool for bipolar disorder.
• Brain network dynamics, particularly involving the default mode, subcortical, and attention networks, are key to BD diagnosis.