Verification and uncertainty quantification of clinical CFD software for patient-specific coronary hemodynamics assessment

Patient-specific coronary hemodynamics, based on medical imaging and Computational Fluid Dynamics (CFD) simulations, plays a critical role in cardiovascular research, as approximately one-third of cardiovascular-related deaths result from coronary artery stenosis. Growing evidence demonstrates that Wall Shear Stress (WSS) has a significant prognostic value in predicting lesions potentially responsible for myocardial infarction in stenotic coronary arteries. Although CFD simulations are considered the optimal approach to obtain a realistic assessment of WSS, their high computational cost combined with the need for specialized users has motivated the development of clinical software able to perform CFD simulations in times compatible with clinical practice, directly in the cath lab by clinicians. The aim of this thesis was to perform verification and uncertainty quantification of the clinical prototype software to evaluate its reliability. Starting from angiographic images and using two three-dimensional reconstruction techniques (Bifurcation, where the actual coronary bifurcation is reconstructed, and SD, where the side branch is represented by a cylindrical extension in the position of the actual bifurcation), models of the coronary arteries were obtained. Transient CFD simulations, previously carried out with the clinical software, were then also performed using the Finite Volume Method (FVM) based commercial software Ansys Fluent. Blood was considered to be an incompressible, homogeneous, Newtonian fluid. As for boundary conditions, two different waveforms typical of Left Circumflex arteries (LCX) and Left Anterior Descending arteries (LAD), and scaled on the basis of the diameter of the inlet section were prescribed at the inlet, as parabolic velocity profiles. At the outlets, flow rate portions of inflow, determined according to a diameter-based scaling law were imposed. The arterial wall was assumed to be rigid and with no-slip conditions. The verification was conducted through comparative analysis of WSS-based descriptors, spatially averaged over coronary artery lesions, focusing on: Time-Averaged Wall Shear Stress (TAWSS), Oscillatory Shear Index (OSI), transverse Wall Shear Stress (transWSS) and Topological Shear Variation Index (TSVI). Statistically significant differences were observed between the lesion-averaged values of all hemodynamic descriptors considered in the analysis. The difference between the software and Fluent, for the Bifurcation technique, is statistically significant for OSI (p = 0.0039), for transWSS (p = 0.0195) and for TSVI (p = 0.0195), but not for TAWSS (p = 0.5703), suggesting that magnitude is more robust to meshing and numerical setup than directional metrics. As for the SD technique, significant differences between the software and Fluent were observed for all descriptors with the exception of TSVI (p = 0.1704). Importantly, despite the statistically significant differences, a strong significant correlation was observed between the lesion-averaged values of all descriptors evaluated on the basis of WSS (r &#8805; 0.8 and p < 0.05), thereby confirming the robustness of the results obtained with clinical software. The results obtained indicate that the clinical prototype software provides a reliable and faster computation of WSS, with times compatible with clinical practice, representing an efficient alternative to traditional CFD simulations. This may promote the widespread use of WSS as a prognostic indicator in clinical settings.