Computational blood flow modelling in realistic arterial geometries

grantor: University of TorontoComputational blood flow modelling in patient-specific arterial geometries is a powerful tool for examining the correlation between haemodynamics and vascular diseases. A popular method for obtaining such patient-specific geometries is medical imaging. Haemodynamic modelling in arterial geometries constructed from such images is a relatively new technique, and there are questions regarding its accuracy and utility. This thesis addresses several of these questions. A methodology for reconstructing 3-D finite element models from medical images was developed. The methodology was validated using a straight tube and a carotid bifurcation flow-through phantom, each of which was imaged, computer reconstructed, and used for steady and unsteady flow modelling. In the second phase of this work, an anatomically realistic end-to-side anastomosis geometry was reconstructed and then successively idealized so as to examine the effects of model idealization on flow and wall shear stress fields. Although quantitative wall shear stress was very sensitive to local geometry, substantial simplification of out-of-plane curvature, surface features and local arterial calibre changes could be made to a model before 'significant' qualitative changes in wall shear stress was seen. In the third phase of the work, a comparison was made between arterial models of rabbit aorto-iliac bifurcations obtained from 'in vivo' MR images and models obtained from two common post-mortem tissue processing techniques: vascular casting and pressure fixation. These experiments indicated that different methods of obtaining a description of the same artery can produce significantly different model geometries and wall shear stress fields. ' In vivo' MR provides a more reliable description of overall arterial geometry than the postmortem techniques; however, specific arterial dimensions may not be as accurately replicated by reconstruction of 'in vivo' MR images. These results suggest that, while patient-specific modelling of ' in vivo' haemodynamics is possible, there are important potential inaccuracies. Further, in the absence of patient-specific disease localization, idealized models provide sufficient haemodynamic information for correlating disease sites with flow features.Ph.D