Uncertainty Quantification and Sensitivity Analysis for Non‐invasive Model‐Based Instantaneous Wave‐Free Ratio Prediction

The main objectives of this work are to validate a 1D-0D unsteady solver with a distributed stenosis model for the patient-specific estimation of resting haemodynamic indices and to assess the sensitivity of instantaneous wave-free ratio (iFR) predictions to uncertainties in input parameters. We considered 52 patients with stable coronary artery disease, for which 81 invasive iFR measurements were available. We validated the performance of our solver compared to 3D steady-state and transient results and invasive measurements. Next, we used a polynomial chaos approach to characterise the uncertainty in iFR predictions based on the inputs associated with boundary conditions (coronary flow, compliance and aortic/left ventricular pressures) and vascular geometry (radius). Agreement between iFR and the ratio between cardiac cycle averaged distal and aortic pressure waveforms (resting <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:semantics> <mml:mrow><mml:msub><mml:mi>P</mml:mi> <mml:mi>d</mml:mi></mml:msub> <mml:mo>/</mml:mo> <mml:msub><mml:mi>P</mml:mi> <mml:mi>a</mml:mi></mml:msub> </mml:mrow> <mml:annotation>$$ {P}_d/{P}_a $$</mml:annotation></mml:semantics> </mml:math> ) obtained through 1D-0D and 3D models was satisfactory, with a bias of 0.0-0.005 (±0.016-0.026). The sensitivity analysis showed that iFR estimation is mostly affected by uncertainties in vascular geometry and coronary flow (steady-state parameters). In particular, our 1D-0D method overestimates invasive iFR measurements, with a bias of -0.036 (±0.101), indicating that better flow estimates could significantly improve our modelling pipeline. Conversely, we showed that standard pressure waveforms could be used for simulations, since the impact of uncertainties related to inlet-pressure waveforms on iFR prediction is negligible. Furthermore, while compliance is the most relevant transient parameter, its effect on iFR estimates is negligible compared to that of vascular geometry and flow. Finally, we observed a strong correlation between iFR and resting <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:semantics> <mml:mrow><mml:msub><mml:mi>P</mml:mi> <mml:mi>d</mml:mi></mml:msub> <mml:mo>/</mml:mo> <mml:msub><mml:mi>P</mml:mi> <mml:mi>a</mml:mi></mml:msub> </mml:mrow> <mml:annotation>$$ {P}_d/{P}_a $$</mml:annotation></mml:semantics> </mml:math> , suggesting that steady-state simulations could replace unsteady simulations for iFR prediction.