MULTIFIDELITY ESTIMATORS FOR CORONARY CIRCULATION MODELS UNDER CLINICALLY INFORMED DATA UNCERTAINTY.

Area of Science:

• Computational fluid dynamics
• Biomedical engineering
• Cardiovascular research

Background:

• Numerical models are vital for diagnosing coronary artery disease (CAD) and planning treatments.
• Current deterministic models lack quantitative assessment of simulation output variability due to input uncertainties.
• Accurate patient-specific models require precise parameters like aortic pressure waveform and intramyocardial pressure.

Purpose of the Study:

• To quantify the impact of input parameter uncertainty on clinically relevant outputs in coronary circulation models.
• To develop and validate a computational framework for uncertainty quantification in patient-specific coronary models.
• To reduce the computational cost of uncertainty propagation in complex hemodynamics simulations.

Main Methods:

• Developed a deformable model of the left coronary artery using an arbitrary-Lagrangian-Eulerian framework for fluid-structure interaction.
• Estimated random input uncertainty from repeated intracoronary catheterization measurements and literature data.
• Employed multifidelity Monte Carlo estimators, including 0D lumped parameter models, to reduce computational cost and improve variance estimation.

Main Results:

• Multifidelity Monte Carlo estimators significantly reduced variance and improved accuracy compared to traditional Monte Carlo methods.
• Combining 3D hemodynamics simulations with 0D lumped parameter network models yielded the most accurate results.
• The computational overhead for the combined approach was negligible (less than 1%).

Conclusions:

• Uncertainty quantification is essential for reliable patient-specific coronary circulation modeling in clinical practice.
• Multifidelity Monte Carlo methods offer an efficient and accurate approach for uncertainty propagation in complex cardiovascular simulations.
• The integration of low-fidelity models with high-fidelity simulations provides a powerful tool for advancing noninvasive diagnosis and treatment planning in CAD.