Practical identifiability and uncertainty quantification of a pulsatile cardiovascular model

Area of Science:

• Cardiovascular Physiology
• Mathematical Modeling
• Computational Biology

Background:

• Mathematical models are crucial for understanding cardiovascular homeostasis.
• Model accuracy, limited by uncertainty quantification (UQ), often assumes model identifiability.
• A challenge in UQ is the assumption of a one-to-one mapping between model parameters and outputs.

Purpose of the Study:

• To present a novel workflow for calibrating a lumped-parameter cardiovascular model.
• To address the challenge of parameter identifiability in uncertainty quantification.
• To improve the predictive accuracy of cardiovascular models using experimental data.

Main Methods:

• Literature review and data analysis for nominal parameter values.
• Sensitivity analysis and subset selection for identifying estimable parameters.
• Optimization techniques for parameter point estimation.
• Frequentist and Bayesian uncertainty quantification for predictive assessment.

Main Results:

• A workflow was developed to calibrate a lumped-parameter model to left ventricular pressure and volume data.
• Five identifiable model parameters were successfully determined from experimental data of three rats.
• The computed uncertainty quantification intervals effectively captured measurement and model errors.

Conclusions:

• The proposed workflow enables robust calibration of cardiovascular models even with identifiability challenges.
• Accurate estimation of identifiable parameters enhances the reliability of cardiovascular model predictions.
• This approach improves the utility of mathematical models in cardiovascular research through effective UQ.