Granger causality in cardiovascular variability series: comparison between model-based and model-free approaches

Area of Science:

• Physiology
• Biomedical Engineering
• Nonlinear Dynamics

Background:

• Granger causality is crucial for understanding physiological system interactions.
• Linear model-based (MB) and nonlinear model-free (MF) methods offer different approaches to causality assessment.
• Comparing these methods is essential for accurate physiological signal analysis.

Purpose of the Study:

• To compare the effectiveness of linear model-based (MB) and nonlinear model-free (MF) Granger causality methods.
• To evaluate causality and predictability in cardiovascular and respiratory signals during controlled and spontaneous breathing.
• To determine if nonlinear methods offer advantages over linear methods in this context.

Main Methods:

• Linear model-based (MB) approach using multivariate linear regression and least-squares.
• Nonlinear model-free (MF) approach utilizing local prediction and k-nearest neighbors.
• Both methods optimized multivariate embedding dimension; MF was more parsimonious.

Main Results:

• Both MB and MF methods identified increased heart period (HP) predictability during controlled respiration (RC15).
• Causality from systolic arterial pressure (SAP) to HP and from respiration (R) to HP remained unchanged during RC15.
• No significant superiority of nonlinear methods over linear methods was observed for predictability and causality assessment.

Conclusions:

• Linear model-based methods are as effective as nonlinear model-free methods for assessing Granger causality in cardiovascular and respiratory signals.
• Controlled breathing increases heart period predictability without altering causal influences from SAP and R.
• Findings suggest linear approaches are sufficient for analyzing physiological causality in healthy humans.