A Bayesian Time-Varying Psychophysiological Interaction Model

Area of Science:

• Neuroscience
• Brain Imaging
• Computational Biology

Background:

• Functional connectivity analysis is crucial for understanding brain coordination.
• The standard Psychophysiological Interaction (PPI) model has limitations due to confounding effects.
• Existing methods for inferring task-dependent functional connectivity require refinement.

Purpose of the Study:

• To develop a more robust and dynamic method for analyzing functional connectivity.
• To address confounding effects inherent in the standard PPI model.
• To introduce a Bayesian extension of the PPI model for time-varying connectivity estimation.

Main Methods:

• Utilizing partial correlations derived from Gaussian Graphical Models (GGMs) to correct for confounding.
• Implementing a Bayesian extension to the PPI model for dynamic functional connectivity analysis.
• Employing scale-mixture shrinkage priors for sparsity and a Bayesian decision-theoretic framework for identifying structural zeros.

Main Results:

• The proposed method demonstrates superior performance compared to the standard PPI model using simulated data.
• The framework successfully identified dynamic functional connectivity patterns in human fMRI data.
• Partial correlations provide a more accurate measure of functional connectivity than PPI regression coefficients.

Conclusions:

• The novel Bayesian framework offers a more robust and dynamic approach to functional connectivity analysis.
• This method enhances the understanding of task-dependent brain region coordination.
• The findings have significant implications for neuroimaging research and the interpretation of brain activity.