The geometry of conduits derived from in vivo image data is subject to acquisition and reconstruction errors. This results in a degree of uncertainty in defining the bounding geometry for a patient-specific anatomical conduit. The impact of the conduit geometry uncertainty should be considered with respect to haemodynamic clinically relevant measures that may alter the perception and evaluation of prognosis and diagnosis. These are commonly fluid mechanic stresses on or near the wall. Taking an example of a peripheral bypass graft configuration, we examine the effects of image threshold on the geometry. Thresholding approaches are chosen from the existing image segmentation community and are based on clustering schemes. Two novel methods are also introduced. The geometries are reconstructed using a partition-of-unity implicit function approach from the stack of segmented cross-sections that yields a piecewise linear triangulated mesh. Methods to quantify the differences resulting in the virtual model reconstruction from the different thresholding methods are based on the distance between the models and the surface mean curvature.

CEMAT - Center for Computational and Stochastic Mathematics