Creation of high-fidelity CAD model of hemodialysis filters by applying “reverse engineering” approach and CFD simulation of devices performance

Kidney diseases, consisting in a total or partial loss of renal function for the patients, are rapidly increasing over the years. Nowadays, the most common treatment to replace kidney’s functionality is hemodialysis, which filter waste, salts and fluid from patient’s blood by an extracorporeal machine. The core element of the hemodialysis machine is the dialyzer: a filter in which a porous film (i.e., a bundle of fibers) separates blood from the dialysate fluid (with which blood exchanges solutes and toxins). The dialysate flows outside the bundle of fibers while the blood flows inside the fibers. Because the efficacy of hemodialysis therapy depends on the fluid dynamic inside the dialyzer, several computational models were proposed to study the hollow fiber membrane-based hemodialysis process, but only a few used a 3D approach, and none of them reproduced high-fidelity geometry of the filter. In this scenario, the aim of this study was (1) to create high-fidelity CAD models of two clinically used dialyzers (i.e., SOLACEA 21H - Nipro, and CorDiax FX100 - Fresenius Medical Care) by applying a “reverse engineering” approach and (2) to characterize the hemodynamics inside their caps under several working conditions by using computational fluid dynamics (CFD). As for the first goal, the 3D geometry of each component of the two investigating dialyzers was created by using SolidWorks based on measurement acquired in the experimental laboratory on the physical samples. Caliber was used for macroscopic measurements, while the Scanning Electron Microscope was used for the microscopic ones. The final high-fidelity models consisted of about 13000 fibers each. As for the caps’ hemodynamic characterization, two CFD approaches were followed: the Fibers and the Porous model approaches. The first consisted in modelling each dialyzer’s fiber individually considering a reduced number of fibers (N=217) uniformly distributed along the filter cross-section. In the second case the 13000 fibers were simulated by means of a porous media, with estimated porosity, and viscous and inertial resistances. Ansys Fluent was used to perform unsteady-state hemodynamic simulations inside each dialyzer’s cap model at 200, 350 and 500 ml/min, using both CFD approaches. Additionally, the urea transport inside the dialyzers’ geometry was simulated. A vortex ring clearly emerged in the cap of SOLACEA 21H with an extension which slightly decreases with the increase of inflow rate. Additionally, the distance of the vortex ring from the cap center increases with the increase of blood flow rate. More complex and asymmetric flow patterns resulted for the CorDiax FX100. These qualitative observations were independent of the simulated flow regime. Fibers- and the Porous-based velocity distributions were in good agreement, showing the same regions with low, mid and high values inside the models’ caps. However, the Fibers approach resulted in a significant underestimation of blood velocity magnitude inside the dialyzers’ cap. Overall, urea filled the dialyzer in comparable times among all the considered designs and CFD approaches. In conclusion, the findings from this work suggest that the two proposed CFD approaches may provide a detailed characterization of dialyzer’s hemodynamics. The developed high-fidelity CFD dialyzers’ models might be instrumental in benchmarking innovative designs of dialyzer’s cap or as optimization tool for the existing solutions.