A novel spatial-spectra dynamics-based ranking model for sorting time-varying functional networks from single subject FMRI data

Accumulating evidence suggests that the brain state has time-varying transitions, potentially implying that the brain functional networks (BFNs) have spatial variability and power-spectra dynamics over time. Recently, ICA-based BFNs tracking models, i.e., SliTICA, real-time ICA, Quasi-GICA, etc., have been gained wide attention. However, how to distinguish the neurobiological BFNs from those representing noise and artifacts is not trivial in tracking process due to the random order of components generated by ICA. In this study, combining with our previous BFNs tracking model, i.e., Quasi-GICA, we proposed a novel spatial-spectra dynamics-based ranking method for sorting time-varying BFNs, called weighted BFNs ranking, which was based on the dynamical properties in both spatial and spectral domains of each BFN. This proposed weighted BFNs ranking model mainly consisted of two steps: first, the dynamic spatial reproducibility (DSR) and dynamic fraction of amplitude low-frequency fluctuations (DFALFF) for each BFN were calculated; then a weighted coefficients-based ranking strategy for merging the DSR and DFALFF of each BFN was proposed, to make the meaningful dynamic BFNs rank ahead. We showed the effective results by this ranking model on the simulated and real data, suggesting that the meaningful dynamical BFNs with both strong properties of DSR and DFALFF across the tracking process were ranked at the top.