Peter Rissland

Prosthetic Heart Valve and Abdominal Aortic Aneurism Simulations

A discrete particle dynamics simulation of blood flow through a prosthetic heart valve and a CFD solution of fluid flow of abdominal aortic aneurisms (AAAs) are proposed for the Seawulf supercomputer.
A discrete particle simulation of blood flow through prosthetic heart valves will be utilized in order to develop a multi-scale numerical model with the ability to predict flow induced thrombogenicity. It is aimed at bridging the gap between macroscopic flow scales and cellular scales by incorporating biochemical and cellular reaction kinetics and interactions leading clot formation. All interactions between particles are computed by the Leonard-Jones potential. Current runs on the AMS Galaxy cluster use a 2D representation of the artificial valve containing ~80K particles show that the simulation runs approximately 1 s/step on 16 processors, requiring 100K steps to converge. A full 3D simulation will contain over 10^6 particles, which is beyond the capacity of Galaxy. With 10^6 particles, we estimate 10 s/step on 16 processors, equating to 4444 cpu-hrs. Multiple simulations are needed to check the validity of the simulation based on changes in the force field. In order to validate the simulations, an industry standard CFD package, Fluent, will be installed on Seawulf.
Fluid structure interaction (FSI) simulations of AAA reconstructed from patients CT scans will provide a better diagnostic tool of the severity of the aneurism, its risk of rupture, and give the ability to make a more informed decision regarding the need for surgery. Current policy for surgical intervention is based on the diameter of the AAA, which is a rudimentary and antiquated approach. The ability to quantify stresses in the artery walls of the AAA and indicate areas of stress concentration will provide clinicians with the information necessary to make an effective decision. Current simulations are done on an 8-cpu system in the BME department, running for about 150 hrs (1200 cpu-hrs) using an industry standard package, ADINA. Many AAA simulations need to be run on a variety of geometries to validate numerical models. Having the ADINA and Fluent packages on Seawulf would be valuable for students to use and gain experience using real world CFD packages.
We estimate that the innovative 3D blood flow and AAA simulations will require 10,000 cpu hrs/month for 10 months to obtain preliminary results that will make our planned applications for NIH and NSF funding more competitive.