Information Flow Between Resting-State Networks.

Area of Science:

• Neuroscience
• Brain Imaging
• Network Science

Background:

• Resting-state brain dynamics organize into resting-state networks (RSNs).
• Understanding the interaction patterns between RSNs is crucial but not fully elucidated.
• Current methods for analyzing RSN interactions have limitations.

Purpose of the Study:

• To propose a novel method for computing information flow (IF) between RSNs using resting-state fMRI data.
• To investigate the dimension-dependent nature of IF between RSNs.
• To apply the method for group comparisons, exemplified by Alzheimer's disease (AD).

Main Methods:

• Developed a novel method to compute IF between RSNs from resting-state fMRI.
• Utilized hemodynamic response function blind deconvolution and principal component analysis (PCA) for dimensionality reduction within RSNs.
• Calculated multivariate IF by systematically increasing the number of principal components (k).

Main Results:

• Information flow (IF) among RSNs is dimension-dependent, peaking at k=5 components.
• IF decays to zero for k≥10, suggesting a small number of components capture the interaction patterns.
• Significant differences in inter-RSN IF were observed between Alzheimer's disease patients and controls, particularly at k=2.

Conclusions:

• The proposed method effectively quantifies multivariate IF between RSNs.
• A small number of principal components (around k=5) are sufficient to characterize inter-RSN IF.
• This approach offers a valuable tool for group comparisons in neurological health and disease, as demonstrated in Alzheimer's disease research.