Uncertainty quantification for patient-specific cardiovascular simulations in high-performance computing

Cardiovascular blood flow simulations solve the incompressible Navier-Stokes equations in patient-specific geometries constructed from image data with physiologic boundary conditions. To assess variabilities of simulation predictions under uncertainties in clinical measurements, we perform uncertainty quantification. In our computational model, we account for fluid-structure interaction, coupling blood flow with deformable vessel walls via an Arbitrary-Lagrangian-Eulerian framework, and open-loop lumped parameter boundary conditions to model the peripheral circulation. We consider multiple sources of input uncertainty, including uncertainties in inflow waveform, material properties, and resistance boundary conditions. We introduce several strategies to reduce computational cost for forward uncertainty propagation using Kalhunen-Loeve expansion and a sub-modeling approach. We report statistics on quantities of interest including flow rate, pressure at the distal vessel branches, time-averaged wall shear stress, and wall displacement. Comparisons of several forward uncertainty propagation methodologies will be discussed including Monte Carlo and a recently proposed multiresolution framework.