Evaluating dynamic bivariate correlations in resting-state fMRI: a comparison study and a new approach

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Network Dynamics

Background:

• Functional connectivity (FC) in fMRI is often assumed constant, but dynamic changes offer insights into brain function.
• Existing methods like sliding-window techniques have limitations in capturing temporal FC variations.

Purpose of the Study:

• To evaluate methods for estimating dynamic pair-wise correlations in brain region time series.
• To introduce and assess the Dynamic Conditional Correlation (DCC) model for neuroimaging applications.

Main Methods:

• Critique of sliding-window techniques for fMRI functional connectivity analysis.
• Application of financial volatility models, specifically DCC, to neuroimaging data.
• Simulation studies to compare DCC with other methods for detecting dynamic correlations.

Main Results:

• The Dynamic Conditional Correlation (DCC) model demonstrates superior balance in sensitivity and specificity for detecting dynamic FC changes.
• DCC shows scalability for analyzing correlations among multiple brain regions.
• Successful application of DCC to test-retest resting-state fMRI data.

Conclusions:

• The DCC model is a robust and effective tool for estimating dynamic functional connectivity in fMRI.
• This approach advances the understanding of brain network dynamics beyond static FC assumptions.
• DCC offers a promising avenue for analyzing complex temporal patterns in brain activity.