A quantitative analysis of the impact of different flow rate estimation techniques in computational hemodynamics models of coronary arteries

Computational Fluid Dynamics (CFD) has become, in the last few decades, even more an important tool to investigate local hemodynamics and its possible association with atherosclerosis genesis and progressions, to plan patient-specific interventions and to design and evaluate vascular devices. The combination of CFD and medical imaging allows, in fact, to quantify a variety of descriptors that would be difficult to be measured in vivo, with a spatial and temporal resolution unreachable with other clinical methodology in a patient specific manner. However, the accuracy of CFD predictions is strongly influenced by modelling assumptions which introduce sources of uncertainty. An example is represented by the inflow boundary conditions (BCs) which starts from in vivo flow rate estimation depending on the method used and may provide different flow rate values. The aim of this thesis is to investigate how the method used to extract the inflow rate value from the patient influenced the CFD results, in terms of distribution and values of the main wall shear stress (WSS) based descriptors. To do that, from 14 patients, subjected to invasive coronary cineangiography, the three-dimensional models were extracted and unsteady simulations were performed applying four different inflow BCs, generated by flow rate values estimated from the patients with four different techniques: (1) thermodilution, (2) Doppler, (3) frame count, and (4) van der Giessen’s scaling law. The main WSS-based descriptors associated to the “disturbed flow” were computed, i.e., Time Averaged Wall Shear Stress (TAWSS), Oscillatory Shear Index (OSI), Relative Residence Time (RRT), transverse WSS (transWSS), Cross Flow Index (CFI), Axial TAWSS (TAWSSax), Secondary TAWSS (TAWSSsc), Wall Shear Stress ratio (WSSratio) and Topological Shear Variation Index (TSVI). The averaged values and the percentage of luminal area exposed to “disturbed flow” for each WSS-based descriptor was quantified. Statistically significant differences between inlet BCs, were compared by means of Wilcoxon test, while correlations were computed using the Spearman test. Regarding the percentages of area subjected to “disturbed flow” the four BCs lead to results that can be considered not significantly different, except to WSSratio. Concerning the mean values only the couples thermodilution-frame count and doppler-van der Giessen shows result that cannot be considered significantly different. Good correlation values were found between the BCs results, especially for the “disturbed flow” areas, but values were in general not always statistically significant. In fact, only for TSVI the BCs results show high and statistically significant (p<0.05) values of correlation for both “disturbed flow” areas and averaged values while others WSS-based descriptors present significant correlations only for certain couple of BCs. In conclusion the four different techniques and methods chosen to measure the patient specific flow rate, used to derive the inlet BCs, lead not significantly different percentages of area subjected to “disturbed flow”, while this is not generally true concerning the mean values of the several hemodynamic indexes. Despite of that, statistically significant correlations between the all the BCs were found only for the TSVI results, while for the other index, especially those that are more influenced by the WSS magnitude, the different BCs lead to result that are not always statistically correlated.