Geometrical reconstruction from medical images, classification and modelling of arterial by-pass grafts

Abstract

This thesis presents a non-invasive procedure, based on magnetic resonance imaging (MRI) and computational fluid dynamics (CFD), to characterise the geometry and the flow environment of the distal anastomosis of peripheral by-pass grafts in vivo. A technique to reconstruct three-dimensional models from medical images is described. This technique exploits implicit functions to interpolate the lumen of blood vessels. Algorithms to smooth the reconstructions are discussed together with procedures to represent the reconstructions by parametric surfaces. Reconstructions obtained by the proposed method are accurate and reproducible within the resolution of the images. The technique is automatic and does not rely on topological information known a priori. Hence, its application is not limited to the modelling of vascular structures only. The reconstructions of the anastomoses of 24 patients are classified according to their planarity and to the angles between blood vessels. These parameters are computed with minimum user intervention from the medial lines of the vessels obtained by a three-dimensional thinning algorithm. The results show that the anastomotic geometry depends on the surgical procedure followed to construct the graft. Anastomoses of superficial grafts show a wider angle between graft and proximal host vessel and are less planar. The steady blood flowfield is computed at the patient-specific Reynolds number in four anastomoses using a spectral//ip element method. The distribution of wall shear stress and the transport of fluid particles are influenced by the anastomotic geometry. Larger vortical structures are observed in anastomoses whose angle between graft and proximal host vessel is wider, or whose configuration is less planar; wall shear stresses are higher in anastomoses whose vessels are partially stenosed. Coupling MRI and CFD allows to assess the effects of the in vivo geometry on blood flow and to investigate which anastomotic configurations induce a haemodynamic environment that may delay the progression of disease.