Structurally-informed Bayesian functional connectivity analysis.

Area of Science:

• Neuroscience
• Brain Imaging
• Computational Biology

Background:

• Functional connectivity measures covarying activity between brain regions using functional magnetic resonance imaging (fMRI).
• Distinguishing direct from indirect communication is challenging using standard correlation methods on fMRI data.
• The precision matrix, inverse of the covariance matrix, offers insights into direct brain region interactions via partial correlations.

Purpose of the Study:

• To develop a Bayesian model for functional connectivity analysis.
• To estimate posterior density over precision matrices for quantifying uncertainty in partial correlations.
• To leverage structural connectivity from diffusion imaging to inform the precision matrix structure.

Main Methods:

• Proposed a Bayesian statistical model for functional connectivity.
• Utilized diffusion imaging data to estimate the sparsity pattern of the precision matrix.
• Compared the model's performance against graphical lasso analysis on simulated and resting-state fMRI data.

Main Results:

• The Bayesian model successfully estimated posterior densities over precision matrices.
• The approach allowed for quantification of uncertainty in estimated partial correlations.
• Model performance was validated on both simulated and real resting-state fMRI datasets.

Conclusions:

• The proposed Bayesian method provides a robust framework for functional connectivity analysis.
• This approach enhances the understanding of direct brain communication by quantifying uncertainty.
• Integrating structural connectivity aids in more accurate functional connectivity estimation.