A Probabilistic Variational Framework for Blood Flow Simulation in Patient-specific Arterial Geometries: Non-Newtonian Blood Flow Models and Optimization of Ventricular Assist Devices (VADs)

Arif Masud, University of Illinois at Urbana-Champaign

Usage Details

We have developed a Probabilistic Variational Framework for blood flow simulation in patient-specific arterial models. Our objective is to model arterial thrombosis in Carotid artery and the increased risk due to reduced pulsatility induced by Axial-Flow Ventricular Assist Devices (VADs) in patients with acute heart conditions. This project extends our variational multiscale method for non-Newtonian blood flow modeling to a computational probabilistic framework that accounts for material and geometrical uncertainty that is always there in biological systems. The current developments are aimed at predicting post-surgical risks in hearttransplant patients who have received a Ventricular Assist Device (VAD). Since medical science and clinical diagnostics are statistical in nature, uncertainty in the estimation of various patient specific parameters (e.g., blood viscosity, blood pressure, inflow  conditions, etc.) results in a relatively large number of possible events that need to be considered to develop reliability envelopes for the computed response. With our previous Blue Waters allocation, we were successful in developing a highly parallel code, and we plan to employ it within the newly developed probabilistic framework. The Blue Waters platform is ideally suited for this class of problems not only because of massive parallel problems that need to be executed, but also because of the unique features of our code that are ideally suited for the Blue Waters hardware architecture. We expect that this development will help clinicians in making reliable diagnostics as the computational probabilistic method will provide reliable response envelopes encompassing all the probable critical parameters and conditions related to the patients.