Membrane oxygenator analysis and design using computational fluid dynamics

Kenneth Gage, University of Pittsburgh

Long-term patient support with oxygenators results in hematologic complications as blood elements are exposed to unnatural flow regimes and artificial surfaces. Platelet consumption and activation is significantly increased, potentially resulting in hemorrhagic and thrombotic complications. Thrombotic deposition (blood clotting) within the device may lead to stroke or result in oxygenator failure as the pressure drop across the device increases. These complications may be minimized by eliminating areas of slow and recirculating flow, and by reduction of the overall oxygenator size. Oxygenator design improvement is currently an iterative process requiring fabrication of prototypes and extensive testing. Computational fluid dynamics (CFD) offers an investigator the potential to interpret the effect of oxygenator geometry upon blood flow characteristics, and indirectly, oxygenator hematologic biocompatibility.


Blood flow and oxygen transfer in a commercial hollow fiber membrane oxygenator has been modeled using a widely available CFD package. A 3D geometric model and surface mesh representing the Medtronic Maxima Plus was generated using I-DEAS (SDRC, Milford, OH), a commercial computer-aided design and engineering (CAD/CAE) package. The volumetric mesh was created in TGRID (Fluent, Inc., Lebanon, NH) and the governing equations were solved in FLUENT/UNS (Fluent, Inc., Lebanon, NH). Blood was modeled as a continuum with constant density and viscosity applicable at the flow rates under investigation. Since modeling blood flow around thousands of individual fibers in the oxygenator is currently an intractable problem, the fiber bundle was described as a porous region using a modified Ergun relation to approximate the flow resistance.


The oxygen flux to blood within the oxygenator was modeled using a mass transfer coefficient and the oxygen gradient between the fibers and the surrounding fluid. Oxygen transfer rates were computed on an element by element basis. The effect of local velocities and blood saturation was thus incorporated into the determination of local oxygen transfer. The variable mass transfer coefficient was determined by utilization of a correlation between the local Sherwood number and the local Reynolds and Schmidt numbers. Oxygen uptake by hemoglobin affected the oxygen diffusivity and flux and was accounted for by use of the Hill equation. The blood flow and oxygen transfer characteristics of the Medtronic Maxima Plus at clinical flowrates were computed using the analysis described above. Regions of low velocity flow were qualitatively correlated with clinically observed areas of thrombus formation within the oxygenator. In conclusion, CFD offers a methodology to investigate alternative oxygenator geometries which may lead to more efficient and biocompatible devices in the future.

Abstract Author(s): Kenneth Gage