Résumé
Biomaterial surfaces used in medical devices initiate blood coagulation that may lead to device malfunction or thromboembolism. A numerical strategy is proposed where a boundary condition relates the species wall flux to the surface reaction rate of factor XII activation. This contrasts with conventional in-vivo models in which coagulation is initialized at user-defined injury sites where tissue factor is present. The approach makes it possible to predict the coagulation initiated by the contact activation system but it relies on the coagulation scheme proposed by Chatterjee et al. (2010) which involves as many as 63 reaction rates and is thus hardly tractable for the analysis of complex geometries. A reduced order scheme for the thrombin generation triggered by the contact activation system is thus proposed, based on the outcome of the Morris sensitivity analysis performed to identify the most sensitive parameters involved in the original scheme. The best possible values of the 9 remaining reaction rates are obtained by Bayesian inference, using data from thrombin generation tests as a reference. The resulting optimized reduced coagulation scheme is then coupled to a platelet-based thrombus growth model (Taylor et al., 2016) and implemented in the in-house YALES2BIO solver (www.math.univ-montp2.fr/\textasciitilde yales2bio/); the results obtained for the backward-facing step experiment of Taylor et al. (2014) compare favorably with the measurements.