Modeling sensitivity and uncertainties in platelet activation models applied on centrifugal pumps for extracorporeal life support.

Two platelet activation models were studied with respect to uncertainties of model parameters and variables. The sensitivity was assessed using two direct/deterministic approaches as well as the statistical Monte Carlo method. The first two, are linear in character whereas the latter is non-linear. The platelet activation models were applied on platelets moving within an extracorporeal centrifugal blood pump. The phenomenological, Lagrangian stress- and time-based power law-based models under consideration, have experimentally calibrated parameters and the stress expressed in a scalar form. The sensitivity of the model with respect to model parameters and the expression of the scalar stress was examined focusing on a smaller group of platelets associated with an elevated risk of activation. The results showed a high disparity between the models in terms of platelet activation state, found to depend on the platelets' trajectory in the pump and the expression used for the scalar stress. Monte Carlo statistics was applied to the platelets at risk for activation and not to the entire platelet population. The method reveals the non-linear sensitivity of the activation models. The results imply that power-law based models have a restricted range of validity. The conclusions of this study apply to both platelet activation and hemolysis models.