Calibration of a fully closed-loop lumped-parameter cardiovascular model with baroreflex in response to induced central hypovolemia

Calibration of cardiovascular models is essential to their clinical utility. We provide a full description of the calibration of a non-linear, dynamic model using global optimisation methods. Specifically, we consider an experimental cohort response in a closed-loop, baro-regulated cardiovascular model, in the setting of induced central hypovolemia.We propose a framework for simplifying calibration of dynamic, regulated models. We perform global sensitivity analysis and orthogonality analysis of a 58-parameter non-linear cardiovascular model including the baroreflex, to select an optimal subset of parameters for model calibration. We fit the model to relative Lower Body Negative Pressure data, using global optimisation (Differential Evolution method), searching a reduced input parameter space. Global sensitivity analysis and orthogonality analysis revealed multiple higher-order interactions within the model, as well as areas of similar influence of input parameters, on the components of the objective function. The model was fitted using only baroreflex parameters. Relative errors in the estimation of heart rate, mean arterial pressure, and stroke volume were reduced from −15.03%, 1.60% and 42.06% before calibration to 0.42%, −1.37% and 0.61% after the fitting, respectively.