Metastability, fractal scaling, and synergistic information processing: What phase relationships reveal about intrinsic brain activity

Dynamic functional connectivity (dFC) in resting-state fMRI is promising for clinical biomarkers, but its reliability and interpretability are debated. This study combined phase-based dFC metrics from dynamical systems, stochastic processes, and information dynamics to assess their interrelationships and reliability. Novel relationships between metrics allowed building a predictive model for integrated information. Global metastability, reflecting simultaneous coupling and decoupling tendencies, was the most representative and stable metric in brain parcellations including cerebellar regions. Spatiotemporal patterns of phase-locking changed slowly and continuously over time. The findings suggest that most resting-state fMRI dynamics reflect an interrelated complexity profile unique to each acquisition, challenging cross-sectional designs for neuromarker discovery and indicating individual life-trajectories may be more informative.

Study at a glance

Abstract

Dynamic functional connectivity (dFC) in resting-state fMRI holds promise to deliver candidate biomarkers for clinical applications. However, the reliability and interpretability of dFC metrics remain contested. Despite a myriad of methodologies and resulting measures, few studies have combined metrics derived from different conceptualizations of brain functioning within the same analysis - perhaps missing an opportunity for improved interpretability. Using a complexity-science approach, we assessed the reliability and interrelationships of a battery of phase-based dFC metrics including tools originating from dynamical systems, stochastic processes, and information dynamics approaches. Our analysis revealed novel relationships between these metrics, which allowed us to build a predictive model for integrated information using metrics from dynamical systems and information theory. Furthermore, global metastability - a metric reflecting simultaneous tendencies for coupling and decoupling - was found to be the most representative and stable metric in brain parcellations that included cerebellar regions. Additionally, spatiotemporal patterns of phase-locking were found to change in a slow, non-random, continuous manner over time. Taken together, our findings show that the majority of characteristics of resting-state fMRI dynamics reflect an interrelated dynamical and informational complexity profile, which is unique to each acquisition. This finding challenges the interpretation of results from cross-sectional designs for brain neuromarker discovery, suggesting that individual life-trajectories may be more informative than sample means.