Influence of Input Data Composition and Measurement Errors on Computational Model-Mediated Assessment of...

The integration of computational cardiovascular models with clinical data through parameter estimation algorithms represents a promising approach for quantitative cardiovascular assessment. However, the sensitivity of assessment to variations in the composition of and measurement errors in input clinical data remains insufficiently understood. In this study, we utilized a cardiovascular model to generate in silico datasets representing diverse virtual subjects. From these datasets, hemodynamic data mimicking noninvasive clinical measurements were extracted as inputs for the parameter estimation algorithm (herein the Levenberg-Marquardt algorithm), while the corresponding values of model parameters representing key cardiovascular properties (e.g., arterial stiffness, cardiac function, and vascular resistance) served as the ground truth for evaluating the accuracy of parameter estimation. The obtained results showed that input datasets comprising brachial arterial blood pressures and flow velocities in the ascending aorta and four peripheral arteries (supplying major organs/tissues) enabled accurate parameter estimation. Cardiovascular models reassigned with the estimated parameter values precisely replicated the input data and predicted other important hemodynamic variables (e.g., aortic blood pressure). While these capabilities were largely maintained upon the omission of flow velocity data in a single peripheral artery, they were significantly compromised when flow velocity data in two peripheral arteries were removed. Furthermore, the introduction of measurement errors into the input data substantially increased errors in both parameter estimation and hemodynamic prediction. In summary, this study underscores the importance of rigorous clinical data selection and high-precision in vivo measurements for enhancing the reliability of computational model-mediated cardiovascular assessment.