Inverse Uncertainty Quantification for Personalized Biomechanical Modeling: Application to Pulmonary Poromechanical...

Area of Science:

• Biomechanics
• Computational modeling
• Uncertainty quantification

Background:

• Personalized biomechanical models require accurate parameter estimation for clinical applications.
• Current methods often lack information on estimation accuracy and parameter identifiability.
• Robustness to measurement and model errors is critical for reliable model personalization.

Purpose of the Study:

• To develop a general inverse uncertainty quantification pipeline for personalized biomechanical models.
• To quantitatively assess parameter identifiability and estimation robustness.
• To improve the reliability of digital twins in clinical settings.

Main Methods:

• Generation of synthetic data with known ground-truth parameters under varying noise and model error levels.
• Performing parameter estimation across numerous realizations of errors and initializations.
• Analysis of estimated parameter error distributions to determine identifiability and confidence intervals.

Main Results:

• The proposed pipeline provides quantitative insights into parameter identifiability.
• Confidence intervals were successfully retrieved for estimated parameters in a poromechanical lung model.
• The method demonstrates robustness to different noise and model error levels.

Conclusions:

• The developed inverse uncertainty quantification pipeline enhances the reliability of personalized biomechanical models.
• Quantitative assessment of parameter identifiability is crucial for clinical applications of digital twins.
• This approach offers valuable information for diagnosis and prognosis using personalized model parameters.