Abstract
BACKGROUND: Predicting hemolysis numerically based on the power-law model using idealized coefficients obtained from simplified devices yields a large variability in hemolysis index predictions. A computational fluid dynamics (CFD)-based Kriging surrogate modeling approach, developed by Craven et al. at the US Food & Drug Administration (FDA), was applied to a Fontan cavopulmonary assist device (CPAD) to generate device-specific hemolysis power-law coefficients. METHODS: The hemolysis index of a CPAD was measured using tests in a mock loop and simulated using CFD. The Kriging surrogate modeling approach was employed for the Lagrangian and Eulerian formulations of the stress-based hemolysis power-law model. The CPAD-specific power-law coefficients obtained from one design of the CPAD were used in predicting the Modified Index of Hemolysis (MIH) for an alternate design of the CPAD. RESULTS: The MIH CFD predictions with the CPAD-specific coefficients deviate by 16%-20% using the Eulerian approach, and 7%-15% using the Lagrangian approach, compared with experimental results for the alternate design. This vastly improves over the use of idealized empirical coefficients, which yield variation in MIH predictions up to two orders of magnitude. CONCLUSION: The presented power-law approach shows good correlation between CFD and tests in predicting MIH for CPAD design modifications. The hemolysis power-law coefficients obtained in this study may be useful in predicting hemolysis in similar rotary blood pumps.