Introduction:
To improve the treatment of pulmonary arterial hypertension, personalized cardiovascular models have been proposed. However, detailed four-chamber cardiovascular models depend on a large number of parameters, which cannot be reliably estimated from routinely available clinical data. These parameters are typically fixed to nominal literature-based values, which could introduce a new type of modelling bias.
Methodology:
We construct a reduced-complexity two-ventricle model and investigate whether it improves practical identifiability of its parameters while maintaining predictive dynamics. We show that nine out of thirteen model parameters are influential using global sensitivity analysis and that seven of the influential parameters are identifiable from synthetic clinical indices using profile likelihood analysis. Actual calibration is done using a Bayesian inference framework and a Markov chain Monte Carlo method.
Results:
We show convergence of the parameter estimation during calibration to synthetic data generated by the two-ventricle model itself. When calibrated to data generated by a four-chamber model, our model closely reproduces the overall cardiovascular dynamics. However, it does not fully capture atrial filling effects or ventricular relaxation, and predictions in the venous compartments remain uncertain. These results characterise the trade-off between parameter identifiability and physiological detail and show which dynamics can and cannot be recovered with our model.
Conclusion:
The two-chamber model is a promising complementary approach to four-chamber-based workflows, retaining clinical relevance for well-defined tasks while improving parameter identifiability.