Sensitivity analysis and uncertainty quantification of 1D models of the pulmonary circulation

Abstract

This study combines a one-dimensional (1D) model with micro-CT imaging and hemodynamic data to quantify uncertainty of flow and pressure predictions in the pulmonary arteries in a control and hypoxia induced hypertensive mouse. We use local and global sensitivity and correlation analysis to determine parameters that can be inferred from the model and available data. Least squares optimization is used to estimate mouse specific parameters, and Bayesian as well as asymptotic uncertainty quantification techniques are employed to determine confidence, credible, and prediction intervals for the model parameters and response. These techniques are used to examine the effects of network size and to understand how parameters change with disease (hypertension). Results showed that the peripheral vascular resistance is the most sensitive, and as the network size increases the parameter behavior changes. Correlation analysis revealed that in hypertension large vessel stiffness is correlated with proximal resistance in the boundary. We were able to estimate identifiable parameters using both deterministic and Bayesian techniques (the maxima of the parameter distributions determined using Bayesian analysis aligned with local optima). From these estimates we determined confidence and prediction intervals, which all were within physiological expectation. Analysis of estimated parameter values for the representative mice studied here showed that the hypertensive mouse has stiffer (but larger) vessels and that compliance is decreased both in the proximal and peripheral vasculature.

References in corpus (3)

• Uncertainty quantification of inflow boundary condition and proximal arterial stiffness coupled effect on pulse wave propagation in a vascular network
• Practical Identifiability and Uncertainty Quantification of a Pulsatile Cardiovascular Model
• Numerical method of characteristics for one-dimensional blood flow

Cited by in corpus (3)

• Parameter inference in a computational model of hemodynamics in pulmonary hypertension
• Bayesian Parameter Inference and Uncertainty Quantification for a Computational Pulmonary Hemodynamics Model Using Gaussian Processes
• A multiscale model of vascular function in chronic thromboembolic pulmonary hypertension