Regression dynamic causal modeling for resting-state fMRI

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Connectomics

Background:

• Resting-state functional magnetic resonance imaging (rs-fMRI) is crucial for studying brain connectivity.
• Current methods are limited to undirected functional connectivity or directed effective connectivity in small networks.
• A gap exists for scalable, directed whole-brain connectivity analysis in rs-fMRI.

Purpose of the Study:

• To adapt regression dynamic causal modeling (rDCM) for rs-fMRI.
• To evaluate rDCM's ability to provide directed, whole-brain connectivity estimates.
• To assess rDCM's computational efficiency and biological plausibility.

Main Methods:

• Simulations were used to validate rDCM parameter recovery across varying signal-to-noise ratios and repetition times.
• Construct validity was tested by comparing rDCM with spectral DCM using rs-fMRI data from ~200 healthy participants.
• rDCM was applied to reconstruct whole-brain networks (>200 areas).

Main Results:

• Simulations confirmed rDCM's accurate parameter recovery.
• rDCM yielded biologically plausible directed connectivity estimates consistent with spectral DCM.
• Whole-brain network reconstruction was achieved within minutes on standard hardware, demonstrating high computational efficiency.

Conclusions:

• Regression dynamic causal modeling (rDCM) is a valid and efficient method for directed whole-brain connectivity analysis in rs-fMRI.
• rDCM overcomes previous limitations, offering scalable and computationally tractable directed connectome estimation.
• This advancement opens new possibilities for connectomics research using resting-state fMRI data.