Uncertainty quantification and sensitivity analysis of left ventricular function during the full cardiac cycle.

Area of Science:

• Computational mechanics
• Biomedical engineering
• Cardiac modeling

Background:

• Patient-specific computer simulations offer potential in clinical diagnostics and treatment development.
• Reliability of these cardiac simulations is paramount for practical application.
• Model construction involves uncertainties in parameters, geometry, and fiber orientation, often due to semi-manual processes.

Purpose of the Study:

• To perform uncertainty quantification and sensitivity analyses on cardiac simulations.
• To assess the variability in key quantities of interest (QoI) for left ventricle function.
• To identify parameters that significantly contribute to prediction variability.

Main Methods:

• Incorporated uncertainties in multiple model parameters: regional wall thickness, fiber orientation, passive material properties, active stress, and circulatory model.
• Performed uncertainty quantification and sensitivity analyses on left ventricle function simulations.
• Analyzed clinical quantities for overall variability and key contributors.

Main Results:

• Quantities of interest (QoI) demonstrated high sensitivity to active stress, wall thickness, and fiber direction.
• Ejection fraction and ventricular torsion were identified as the most impacted outputs.
• Variability in QoI was significantly influenced by the chosen parameters.

Conclusions:

• Improving the precision of cardiac mechanics models requires addressing uncertainties in geometrical reconstruction, active stress estimation, and fiber orientation.
• Reducing errors in these semi-manual processes is essential for enhancing model reliability.
• Further research into robust uncertainty quantification methods is needed for clinical translation.