Reconstructing multivariate causal structure between functional brain networks through a Laguerre-Volterra based...

Area of Science:

• Neuroscience
• Brain Connectivity
• Signal Processing

Background:

• Classical Granger causality (GC) for brain functional connectivity relies on autoregressive models.
• Existing methods have limitations in capturing multi-scale interactions due to signal autocorrelation and low model orders.
• Accurate assessment of brain functional connectivity is crucial for understanding neurological disorders.

Purpose of the Study:

• To propose a novel method for effective Granger causality (GC) assessment of brain functional connectivity.
• To overcome limitations of classical autoregressive models in capturing complex brain dynamics.
• To analyze functional connectivity in human brain fMRI data using the new approach.

Main Methods:

• Utilized discrete-time orthogonal Laguerre basis functions within a Wiener-Volterra decomposition of BOLD signals.
• Developed a Laguerre-Volterra based Granger causality (GC) framework for functional connectivity analysis.
• Validated the method using synthetic noisy oscillator networks and experimental fMRI data from the Human Connectome Project (HCP).

Main Results:

• Laguerre-Volterra based GC estimates demonstrated superior accuracy in detecting true causal links and rejecting false ones in synthetic networks compared to classical methods.
• Analysis of human fMRI data revealed novel causal modulations.
• The default mode network was found to causally modulate both the salience network and fronto-temporal circuits.

Conclusions:

• The proposed Laguerre-Volterra method enhances the accuracy of Granger causality (GC) assessments for brain functional connectivity.
• This approach effectively captures complex, non-linear interactions in brain networks.
• The study provides new insights into causal interactions within human brain networks, particularly involving the default mode network.