Efficient Parallel Computation to Simulate Blood Flow

Hemodynamic simulation is becoming an important tool for the understanding of endovascular diseases. The goal of this PhD dissertation is to provide a parallel, fast, and robust numerical solver of blood flow dynamic that interfaces easily with medical imaging. One of the necessary components of this project is to guarantee reliability, performance and portability of the blood flow simulation code. The core idea is to provide a fast hemodynamic simulation of a given patient close to clinical condition in order to help the endovascular surgeons to get faster results and analysis for treatment planning and to ensure better therapy. This simulator will provide paramount information for cardiovascular diseases, such as velocity, pressure and shear stress. The key factor in this process is the numerical simulation with the implementation of an interface gathering the state of the art solver libraries choosing the fastest solver depending on the grid size and the computing environment. It also integrates a fast prototyping of a Navier-Stokes flow simulation based on the L2 penalty method in order to deal with complex geometry. Finally, a domain decomposition called AitkenSchwarz has been incorporated to take advantage of the parallel environment.