Comparison Between Bench-Top and Computational Modelling of Cerebral Thromboembolism in Ventricular Assist Device Circulation.

Despite improvements in ventricular assist devices (VAD) design, VAD-induced stroke rates remain remarkably high at 14-47%. We previously employed computational fluid dynamics (CFD) to propose adjustment of VAD outflow graft (VAD-OG) implantation to reduce stoke. Herein, we present an in-vitro model of cerebral vessel embolization in VAD-assisted circulation, and compare benchtop results to CFD predictions. The benchtop flow-loop consists of a 3D printed aortic bed using Accura 60 polymer driven by a continuous-flow pump. Three hundred spherical particles simulating thrombi of 2, 3.5, and 5 mm diameters were injected at the mock VAD-OG inlet. A water and glycerin mixture (3.8 cP viscosity) synthetically mimicked blood. The flowrate was adjusted to match the CFD Reynolds number. Catch cans were used to capture and count particles reaching cerebral vessels. VAD-OG geometries were evaluated using comparison of means Z-score range of -1.96 ≤ Z ≤ 1.96 to demonstrate overall agreement between computational and in-vitro techniques. Z-scores were: (i) Z = -1.05 for perpendicular (0°), (ii) Z = 0.32 for intermediate (30°), and (iii) Z = -0.52 for shallow (60°) anastomosis and confirmed agreement for all geometries. This study confirmed added benefits of using a left carotid artery bypass-graft with percent embolization reduction: 22.6% for perpendicular, 21.2% for intermediate, and 11.9% for shallow anastomoses. The shallow anastomosis demonstrated lower degrees of aortic arch flow recirculation, consistent with steady-flow computations. Quantitatively and qualitatively, contemporary steady-flow computational models for predicting VAD-induced cerebral embolization can be achieved in-vitro to validate the CFD equivalent.