Direct Estimation of General Pulmonary Vascular Models from Occlusion Experiments

Abstract

Occlusion experiments yield time–pressure and time–flow curves which are related to the longitudinal distribution of compliances and resistances in the pulmonary circulation. The standard approach to the analysis of these curves involves the observation of relevant features of their graphs, which may directly reflect model parameter values. The present work considers five possible models of pulmonary vascular pressure dynamics and the relative (nonlinear) least-squares parameter estimation from experimental data, making simultaneous use of all available information. In situ isolated perfused and ventilated pig lung preparations were used, and pressure and flow changes during arterial, double, and venous occlusion maneuvers were measured. The five models considered included two linear models without inductance units, one linear model with inductance units, one nonlinear model with variable resistance, and one nonlinear model with variable compliance. In all cases parameter estimation for the numerically integrated model was performed by unweighted least squares, using a variable-metric minimization technique. Comparisons between competing models were based on parameter identifiability and on the Akaike Information Criterion (AIC), concluding that significant nonlinearity in the response of the pig lung resistance to variations in pressure is present for the analyzed data sets.