The influence of vessel segmentation threshold on the accuracy of patient-specific coronary blood flow simulations

Recently, the potential applicability of Computational Fluid Dynamics (CFD) methods in image-based 3D blood flow simulations in both diagnosis and prognosis of Coronary Artery Disease (CAD) was verified in a number of clinical studies. Strong correlations were demonstrated between the invasively measured and computed FFR (fractional flow reserve) as well as between variations in the computed wall shear stress and the sites of atherosclerotic plaque formation. However, there is no globally established guidelines for 3D patient-specific coronary blood flow simulations. In addition, the existing modelling approaches are generally characterised by relatively high levels of uncertainty and instability due to a number of unknown factors and required modelling assumptions. Analysis and comparison of the impact of various modelling assumptions has the potential to reduce this uncertainty and avoid erroneous results. This paper presents a quantitative analysis of the impact of the vessel lumen segmentation threshold, which, being one of the geometry-related factors, defines the physical boundaries of the 3D computational domain thus directly affecting the computed haemodynamic fields. In the investigated example, a series of blood flow simulations were performed for the patient-specific model of left coronary artery (LCA) reconstructed from Coronary Computed Tomography Angiography (CCTA) volume under different threshold values.