Personalized and uncertainty-aware coronary hemodynamics simulations: From Bayesian estimation to improved...

Area of Science:

• Computational fluid dynamics
• Medical imaging
• Cardiovascular modeling

Background:

• Current non-invasive coronary hemodynamics simulations lack personalization, often using empirical flow distribution that overlooks patient-specific factors and data uncertainty.
• This limitation hinders accurate clinical risk stratification and treatment optimization for coronary artery disease (CAD).
• Existing models fail to fully account for variability in patient data, disease presence, and other crucial clinical factors.

Purpose of the Study:

• To develop an end-to-end, uncertainty-aware pipeline for personalized coronary flow simulations.
• To incorporate patient-specific coronary flows and cardiac function into simulations.
• To predict clinical and biomechanical outcomes with enhanced precision by accounting for data uncertainty.

Main Methods:

• Assimilated patient-specific myocardial blood flow data from CT myocardial perfusion imaging to estimate branch-specific coronary artery flows.
• Employed adaptive Markov Chain Monte Carlo sampling to estimate model parameters, simulating noise in clinical data to capture uncertainty.
• Utilized a novel multi-fidelity Monte Carlo estimation combined with non-linear, data-driven dimensionality reduction to determine posterior predictive distributions.

Main Results:

• The framework accurately reproduced clinically measured cardiac function and branch-specific coronary flows, even with simulated measurement noise.
• Observed significant reductions in confidence intervals for predicted quantities compared to single-fidelity and state-of-the-art multi-fidelity Monte Carlo methods.
• Demonstrated substantial computational cost savings for the proposed multi-fidelity Monte Carlo estimators compared to traditional approaches.

Conclusions:

• The developed pipeline enables personalized and uncertainty-aware predictions of coronary hemodynamics using routine clinical data and advanced CT imaging techniques.
• The approach offers significant improvements in predictive precision and a notable reduction in computational expense.
• This methodology holds promise for advancing risk stratification and treatment planning in coronary artery disease.