Automated Tuning for Parameter Identification and Uncertainty Quantification in Multi-scale Coronary Simulations

Area of Science:

• Cardiovascular Medicine
• Computational Fluid Dynamics
• Biomedical Engineering

Background:

• Atherosclerotic coronary artery disease is a leading cause of death globally, necessitating better diagnostic and prognostic tools.
• Non-invasive assessment of coronary blood flow is crucial for understanding and managing this condition.
• Current computational models for hemodynamics require manual parameter tuning and lack uncertainty quantification, limiting clinical utility.

Purpose of the Study:

• To develop an automated, Bayesian approach for parameter estimation in computational fluid dynamics models of coronary blood flow.
• To integrate non-invasive clinical measurements into these models to improve accuracy and reliability.
• To quantify uncertainty in model parameters and predicted hemodynamic indicators for better clinical decision-making.

Main Methods:

• Implementation of an adaptive Markov chain Monte Carlo sampling technique for parameter estimation.
• Assimilation of routinely acquired non-invasive hemodynamic data into a multiscale computational framework.
• Development of methods to quantify uncertainty in estimated parameters and local hemodynamic predictions.

Main Results:

• Successful development of an automated Bayesian framework for hemodynamic modeling.
• Demonstrated ability to assimilate non-invasive clinical data for parameter estimation.
• Quantification of uncertainty in model parameters and hemodynamic indicators.

Conclusions:

• The proposed automated Bayesian approach offers a robust method for personalized hemodynamic analysis in coronary artery disease.
• This technique addresses limitations of manual parameter tuning and uncertainty in current computational models.
• Improved hemodynamic insights can aid in the prevention, understanding, and treatment of atherosclerotic coronary artery disease.