Computational fluid dynamics (CFD) is increasingly being used to develop blood-contacting medical devices. However, the lack of standardized methods for validating CFD simulations and blood damage predictions limits its use in the safety evaluation of devices. Through a U.S. Food and Drug Administration (FDA) initiative, two benchmark models of typical device flow geometries (nozzle and centrifugal blood pump) were tested in multiple laboratories to provide experimental velocities, pressures, and hemolysis data to support CFD validation. In addition, computational simulations were performed by more than 20 independent groups to assess current CFD techniques. The primary goal of this article is to summarize the FDA initiative and to report recent findings from the benchmark blood pump model study. Discrepancies between CFD predicted velocities and those measured using particle image velocimetry most often occurred in regions of flow separation (e.g., downstream of the nozzle throat, and in the pump exit diffuser). For the six pump test conditions, 57% of the CFD predictions of pressure head were within one standard deviation of the mean measured values. Notably, only 37% of all CFD submissions contained hemolysis predictions. This project aided in the development of an FDA Guidance Document on factors to consider when reporting computational studies in medical device regulatory submissions. There is an accompanying podcast available for this article. Please visit the journal's Web site (www.asaiojournal.com) to listen.

No abstract

A computational methodology for simulating virtual inferior vena cava (IVC) filter placement and IVC hemodynamics was developed and demonstrated in two patient

Inferior vena cava (IVC) filters have been used for nearly half a century to prevent pulmonary embolism in at-risk patients. However, complications with IVC filters remain common. In this study, we investigate the importance of considering the hemorheological and morphological effects on IVC hemodynamics by simulating Newtonian and non-Newtonian blood flow in three IVC models with varying levels of geometric idealization. Partial occlusion by an IVC filter and a thrombus is also considered. More than 99% of the infrarenal IVC volume is found to contain flow in the nonlinear region of the shear rate-viscosity curve for blood (less than 100 s) in the unoccluded IVCs. Newtonian simulations performed using the asymptotic viscosity for blood over-predict the non-Newtonian Reynolds numbers by more than a factor of two and under-predict the mean wall shear stress (WSS) by 28-54%. Agreement with the non-Newtonian simulations is better using a characteristic viscosity, but local WSS errors are still large (up to 50%) in the partially occluded cases. Secondary flow patterns in the IVC also depend on the viscosity model and IVC morphological complexity. Non-Newtonian simulations required only a marginal increase in computational expense compared with the Newtonian simulations. We recommend that future studies of IVC hemodynamics consider the effects of hemorheology and IVC morphology when accurate predictions of WSS and secondary flow features are desired.