P78: Risk Assessment of Thrombus Formation in the left Ventricle during VAD Support using computational Models. “How simple is too simple?”

Abstract

Background: Thrombus formation is a major complication in LVAD treatment, and is influenced by intraventricular blood flow dynamics. Many numerical studies have investigated left ventricular (LV) blood flow patterns with LVAD support using various assumptions which are needed to simplify the highly complex physiological processes for a computational environment. Until now, these various assumptions have not been investigated systematically. This study aimed to assess the effect of modelling assumptions on the risk assessment of thrombus formation in the LV during VAD support. Methods: A patient-specific LV and left atrium (LA) was modelled from CT data of a VAD patient. The impact of the turbulence model (laminar vs turbulent), fluid model (Newtonian vs Carreau approach vs Multiphase approach), model geometry (cylindrical LA vs segmented LA, absent MV vs present MV) and boundary conditions (continuous inflow vs periodic inflow) were evaluated systematically, resulting in eleven simulation set ups in total (Figure 1). They were compared quantitatively over 5 s simulation time by the difference of averaged velocities, residence time, washout, turbulent kinetic energy and computing time for the various simulation models. Results: Geometrical LA changes resulted in the biggest quantitative impact, with a 30.7% lower residence time with a cylindrical LA compared to the segmented LA. The comparison of a laminar and a turbulent flow showed a 26.8 % lower residence time in the laminar flow model compared to a model with turbulent flow. The most significant difference in computational time was found when comparing a Newtonian fluid model and a multiphase approach: The simulation with a multiphase approach took 81 core hours longer than the model with a Newtonian fluid model. Conclusion: The recommended model for achieving a low computing time with minimized loss of accuracy compared to current numerical models assumes a turbulent viscous model with Newtonian-modelled blood, a segmented left atrium, a mitral valve geometry, and pulsatile inflow.Figure 1. Simplification study: Steps and result.