A probabilistic reduced-order modeling framework for patient-specific cardio-mechanical analysis.

Area of Science:

• Computational mechanics
• Biomedical engineering
• Medical imaging

Background:

• Cardiac models offer valuable clinical insights but are computationally intensive, limiting their real-world application.
• Existing models face challenges in balancing accuracy with computational efficiency for patient-specific analyses.

Purpose of the Study:

• To develop a probabilistic reduced-order modeling (ROM) framework to decrease computational effort in cardiac simulations.
• To provide a credibility interval for model predictions, enhancing clinical decision-making trustworthiness.

Main Methods:

• Developed a generalized one-fiber model with correction factors for patient-specific attributes.
• Employed Bayesian inference and a Gaussian process for calibrating and predicting correction factors using a full-order model (FOM).
• Validated the framework on idealized and scan-based left-ventricle geometries.

Main Results:

• The ROM framework demonstrated accurate online predictions when sufficient FOM training data was available.
• The framework successfully emulated patient-specific attributes like local geometry variations.
• Uncertainty bands provided insights into prediction trustworthiness, indicating areas for further data collection.

Conclusions:

• The probabilistic ROM framework significantly reduces computational burden for cardiac models.
• The approach enables faster, patient-specific simulations with reliable uncertainty quantification.
• This method holds promise for enhancing clinical decision-making through efficient and trustworthy computational tools.