[Analyzing Frequency-Dependent Human Brain Information Flow in Resting-State fMRI Using Multiple Effective...

To investigate the information flow patterns in the human brain across different frequency bands of resting-state functional magnetic resonance imaging (rs-fMRI) using 7 analysis methods to assess effective brain network connectivity.

The high spatio-temporal rs-fMRI data of 60 healthy volunteers (30 males and 30 females) aged between 22 and 35 years were downloaded from the Human Connectome Project (HCP) database. The information flow patterns of different frequency bands, including conventional low-frequency band (0.01-0.08 Hz), high-frequency band (0.08-0.69 Hz), and whole-frequency band (0.01-0.69 Hz), were analyzed by Granger causality analysis (including linear Granger causality model, kernel-based Granger causality model, and non-parametric multiplicative regression Granger causality model), transfer entropy (based on binning, k-nearest neighbors, and permutation), and convergent cross mapping.

Within the low frequency band, the preferred information flow showed similar topologies across all the analysis methods, with the information flow going predominantly from sub-cortical nucleus, limbic lobe, and a few regions of frontal and temporal lobes into occipital and parietal lobes and other regions of frontal and temporal lobes. In contrast, within the high and whole frequency bands, the information flow was in the opposite direction. Additionally, significant negative correlations were found between the preferred information flow direction and the relative power of low- and high-frequency bands, respectively.

The multimodal effective connectivity analysis conducted in the study reveals rs-fMRI frequency-dependent information flow patterns in the human brain, validates the consistency of different methods in assessing the directional information transfer in the brain network, and offers new insights for understanding the regulatory mechanisms of resting-state brain functions.