On software tools and variability in medical image-based computational fluid dynamics

Abstract

Intracranial aneurysms are balloon-shaped blobs on blood vessels inside the skull and are more brittle than the surrounding blood vessels, and can rupture, causing a stroke. To predict if an aneurysm is going to rupture, we are interested in the local blood flow inside the aneurysm. However, it is challenging to measure blood flow in the brain with good enough resolution. Progress in numerical simulations, combined with medical images of patients’ blood vessels, has provided new opportunities to model blood flow instead. However, the field has had several contentious areas, some of which could be attributed to the multitude of options and assumptions when performing such simulations. In this thesis I have therefore measured the real-world variability, going from a medical image to final blood flow simulation and rupture prediction, by comparing results across many research groups. I have also developed a new software tool that allows for parametric alteration of blood vessel geometry to study how it impacts blood flow. In medical image-based blood flow modeling for aneurysms, better reproducibility and replicability may help advance blood flow modeling as a robust research tool, which ultimately could assist clinicians in improving treatment strategy and patient outcomes.