Boundary condition assessment and geometrical accuracy enhancement for computational haemodynamics

Cardiovascular diseases cause over 47 % of all deaths in Europe each year. Computational fluid dynamics provides the research community with a unique opportunity to investigate cardiovascular diseases with the intent of enabling optimised, patient-specific medical therapies. Incorporating physiologically accurate geometries and boundary conditions into computational fluid dynamics simulations can be difficult tasks and are a concern for researchers. This thesis analyses the impact various inlet and outlet boundary conditions can have on the outcome of a simulation. It also presents a novel, semi-automated process that prepares accurate geometrical models for haemodynamic simulations. Firstly, rabbit and human aorta models were used to analyse the impacts of boundary conditions on haemodynamic metrics used for understanding cardiovascular disease pathology. Comparisons were made between traction free, Murray’s Law, three-element Windkessel, and Murray’s Law/in vivo data hybrid outlet boundary conditions. Steady-state, transient, fully-developed and plug-type inlet boundary conditions were also investigated. Results showed that when advanced models such as the three-element Windkessel are unavailable, the Murray’s Law based outlet returns the most physiologically accurate haemodynamics. Results also showed that prescribing a transient simulation and a fully-developed flow at the inlet are not required when the focus is only on the flow within the aorta and around the intercostal branches. Secondly, a sensitivity test was conducted on the simulation of Left Ventricular Assistive Device (LVAD) configurations. The effects of flow ratios between the LVAD and aortic root on haemodynamic metrics were quantified. The general irregular sensitivity of the subclavian and carotid arteries to flow ratios indicates that the perfusion and wall shear stress-based haemodynamic metrics within these arteries cannot be accurately predicted unless the flow ratios are incorporated into the preoperative planning of the optimal LVAD configuration. Finally, a semi-automated reconstruction process combining magnetic resonance angiography and optical coherence tomography data was developed. The process was successful in its ability to create an accurate geometry in a relatively short time. This forms the foundation on which more sophisticated methods can be developed.