Robust image-based estimation of cardiac tissue parameters and their uncertainty from noisy data

Area of Science:

• Computational biology
• Biomedical engineering
• Medical imaging

Background:

• Accurate personalization of computational cardiac models is vital for clinical applications.
• Parameter non-identifiability and data limitations lead to non-unique solutions and uncertainty in model fitting.

Purpose of the Study:

• To develop and validate a stochastic method for estimating cardiac model parameters and quantifying their uncertainty.
• To address the challenge of parameter non-identifiability in image-based electromechanical heart models.

Main Methods:

• Bayesian inference with Markov Chain Monte Carlo (MCMC) sampling to estimate the posterior probability density function (PDF).
• Utilized a fast surrogate model based on Polynomial Chaos Expansion (PCE) for computational efficiency.
• Employed the mean-shift algorithm to identify modes of the PDF and select robust solutions.

Main Results:

• The stochastic method achieved goodness-of-fit comparable to established deterministic methods.
• Successfully demonstrated the non-uniqueness of parameter estimation in cardiac models.
• Provided crucial uncertainty estimates for personalized cardiac models.

Conclusions:

• The proposed stochastic approach effectively quantifies parameter uncertainty in image-based cardiac models.
• The method enhances the reliability of personalized cardiac models for clinical decision-making.
• Addressing parameter uncertainty is essential for robust clinical applications of computational cardiology.