Computational Modeling of Oxygen Transfer in Artificial Lungs

Under physiological conditions, up to 97% of the oxygen in blood that is transported from lungs to tissue is bound to hemoglobin. To predict oxygen transfer in artificial lungs on a membrane fiber level with computational fluid dynamics (CFD), previous investigators have incorporated the hemoglobin-oxygen interaction into an effective diffusivity coefficient to modify the convection-diffusion equation. Based on our own simulations and experiments, these approaches tend to significantly overestimate the oxygen transfer. The present study introduces a novel approach to model the oxygen transfer in blood on a fiber level with CFD. Plasma and red blood cells were implemented as two phases and the reaction of hemoglobin and oxygen to oxyhemoglobin was included in the convection-diffusion equation in form of a source term. The model was implemented with the commercial software Ansys CFX 18.1. CFD simulations were compared with in vitro experiments on three micro oxygenators with a staggered fiber configuration under multiple blood flow conditions. To calibrate the model, a reaction rate R₀ was introduced and experimental data was fitted to a blood flow of 50 mL/h. Our model approximated the oxygen transfer rates with a difference, relative to in vitro results, of -23.7 and +6.3% for blood flows of 20 and 90 mL/h, respectively. The effective diffusivity model, used by previous authors, was implemented for comparison and approximated oxygen transfer rates with a difference, relative to in vitro data, of +13.7, +68.8, and +121.0% for blood flows of 20, 50, and 90 mL/h, respectively. A well-established numerical mass transfer correlation approximated the gas transfer with a difference, referenced on the average in vitro data, of 31.8, 13.1, and 5.0% for blood flows of 20, 50, and 90 mL/h, respectively. Even though results are promising, a thorough validation of the model will require extensive CFD and in vitro studies of multiple fiber arrangements, fiber diameters, and therefore fiber bundle porosities in the future. This article should be understood as a first feasibility study to evaluate the potential of the novel oxygen transfer model.