Stochastic parameter prediction in cardiovascular problems

High-fidelity cardiovascular flow modeling requires accurate velocity boundary data, which strongly affects wall shear stress estimates used to predict diseases such as atherosclerosis. However, in-vivo data like 4D flow MRI are often noisy and low in resolution. We propose a stochastic data assimilation approach combining CFD with an Ensemble Kalman filter to refine boundary estimates in real time. Tested on two- and three-dimensional vascular models, our method reduced errors to below 3% in 2D and about 7% in 3D cases. This improved boundary accuracy enhances patient-specific wall shear stress predictions, supporting more reliable cardiovascular diagnostics and treatments.