Sensitivity of Hostile Hemodynamics to Intracranial Aneurysm Geometry

An intracranial aneurysm (IA) is an abnormal outpouching of an artery in the brain and is thought to be present in about 1 in 30 adults. Rupture of an IA is a devastating event leading to death or disability in most cases. To supplement clinical treatment decisions, researchers have turned to hemodynamics, which are thought to contribute to the growth and rupture of IAs. Simulations using medical-image-based computational fluid dynamics (CFD) yield patient-specific risk parameters, but these parameters are often highly uncertain due to modelling limitations, limiting their clinical utility. Recent CFD studies have shown IA rupture to be associated with wall shear stress (WSS) and jet impingement, but these studies may be overlooking turbulent-like flow instabilities. This thesis investigates the nature of these flow instabilities and risk parameters in the context of uncertainty due to geometry. To characterize these instabilities, this thesis highlights the novel use of spectrograms as an interpretable summary of the flow field (Chapter 2) and demonstrates their variation in a cohort of 50 IAs (Chapter 3). Based on these spectrograms, I developed a novel marker to quantify harmonic flow features and demonstrated that this marker was a better predictor of rupture than popular WSS-based metrics in our cohort of 50 high-fidelity IA simulations (Chapter 4). Overestimation of the IA neck is a widespread issue among IA CFD researchers, which could selectively promote or inhibit the development of flow instabilities. To overcome this issue, I devel- oped a novel technique for improving IA segmentation from 3D imaging (Chapter 5) and showed that hemodynamic parameters including flow instabilities can be sensitive to these common segmentation errors, potentially muddling efforts toward rupture prediction (Chapter 6). Hemodynamic parameters are ultimately only useful if they are robust to modelling error and provide useful diagnostic information that is not more-easily extracted from geometry alone. Lever- aging a recent deep neural network for unsupervised 3D shape interpolation, I created plausible interpolations between patient-derived geometries and demonstrated the hemodynamic sensitivity to these broad changes in geometry (Chapter 7). This research exposes limitations inherent to reduced risk indices and could help researchers identify better risk markers.

Department dc:contributor.department

Mechanical and Industrial Engineering

Year dc:date.issued

2022

Author dc:creator

• MacDonald, Daniel Edward

Advisor dc:contributor.advisor

• Steinman, David A

• aneurysm
• cfd
• fluids
• geometry
• imaging

Handle dc:identifier.uri

http://hdl.handle.net/1807/125005

OAI identifier oai:identifier

oai:utoronto.scholaris.ca:1807/125005