Predictive Information Decomposition as a Tool to Quantify Emergent Dynamical Behaviors in Physiological Networks

Area of Science:

• Physiology
• Network Science
• Information Theory

Background:

• Physiological systems exhibit complex dynamics.
• Understanding collective behaviors in these networks is crucial for diagnosing health conditions.

Purpose of the Study:

• Introduce a framework for multivariate time series analysis.
• Detect and quantify collective emerging behaviors in physiological network dynamics.

Main Methods:

• Compute predictive information (PI) between present and past network states.
• Dissect PI into unique, redundant, and synergistic information.
• Quantify emergence as the ratio of synergistic to redundant information.
• Implement the framework using vector autoregressive (VAR) models.

Main Results:

• Emerging behaviors were observed in simulated VAR processes with coexisting causal interactions and internal dynamics.
• Significant net synergy was detected in cardiovascular and respiratory networks during rest and postural stress.
• Synergy modulation correlated with sympathetic nervous system activation.

Conclusions:

• Causal emergence can be efficiently assessed by decomposing PI in network systems using VAR models.
• The synergy/redundancy balance is a key feature of integrated short-term autonomic control.
• Measures of causal emergence offer a practical tool for quantifying causal influence in physiological networks across various conditions.