Quantifying Structure-Function Coupling in the Human Brain using Variational Graph Contrastive Learning.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Graph Theory

Background:

• Understanding brain function requires analyzing the relationship between structural and functional connectivity.
• Existing methods may not adequately capture the dynamic coupling between brain structure and function at a regional level.

Purpose of the Study:

• To develop and validate a novel computational framework for quantifying regional structure-function coupling (SFC) in the human brain.
• To investigate the distribution patterns of SFC across different brain networks and their relation to cognitive principles.

Main Methods:

• Utilized whole-cortex structural connectivity data and resting-state functional MRI (rs-fMRI) data from the WU-Minn Human Connectome Project (HCP).
• Developed a dual-branch variational graph convolutional network employing contrastive learning to align structural and functional brain region representations.
• Defined structure-function coupling (SFC) based on the Gaussian kernel output of aligned latent representations.

Main Results:

• Demonstrated that SFC is highest in visual networks and lowest in orbito-affective networks.
• Observed that the distribution of SFC aligns with established principles of cognitive hierarchy.
• Ablation and single-run experiments confirmed the model's effectiveness and the robustness of the SFC metric.

Conclusions:

• The proposed variational graph contrastive learning method provides a robust framework for quantifying brain structure-function coupling.
• The identified SFC patterns offer insights into the organizational principles of brain networks and cognitive function.
• This approach facilitates the extraction of stable multimodal coupling features from complex brain network data.