High-resolution flow field investigations in membrane lungs, considering the complex blood rheology

Despite major improvements over the last years, coagulative disorders and clotting phenomena in membrane lungs (MLs) are still considerable complications in extracorporeal membrane oxygenation (ECMO). ECMO is an increasingly used treatment for patients with severe respiratory failure or cardiac arrest [1]. For both, evaluation of therapeutic decisions and fundamental research on patient specific intra-device clotting phenomena, the direct visualization and analysis of clot formation in combination with a detailed flow field correlation is highly desirable and therefore an intensively followed research topic. Modelling blood flow and shear induced coagulation in MLs is challenging. The relevant geometry of oxygenator fibers and chaining threads is complex and spans several length scales. In relevant scales and regimes, blood shows several significant non-Newtonian effects. Viscosity impacts shear rate, which is important in several coagulation mechanisms. Additionally, coagulation processes are influencing fluid properties and geometry significantly. Existing approaches of previous research work are only able to consider some, but not all relevant effects and geometrical details. Due to the enormous size of the discretized geometries, highly detailed viscosity and coagulations models are not applicable. Our goal is to develop a model for combined viscosity and coagulation properties of blood flow in MLs. In our work, we compare the influence of different levels of detail of the ML geometry as well as the influence of considering realistic blood flow behavior (viscosity change by considering the local hematocrit distribution within the Fåhraeus-Lindqvist-Effect) on the resulting flow field in relevant subsections of a ML. High-resolution micro-CT geometry reconstructions [1] are compared to idealized generic fiber representations. For realistic blood flow modelling, Newtonian representation is compared to the established Carreau-Yasuda and a multiphase Euler-Euler approach. Results are presented for relevant subsections as well as for the complete ML.