Analysis of the parameter space of a metric for registering 3D vascular images

Abstract

We present a new metricfor registering 3D images of vasculature, and we analyze the rigid-body transformation parameter space of that metric and its derivatives. To quantify and direct a source image’s alignment with a target image, this new vascular-image registration system models the vessels in the source image and makes measurements in the target image at a sparse set of transformed points from the centerlines of those models. The system is fast and effective because the measures made at the transformed centerline points incorporate the general geometric properties of tubes and specific model-quality information calculated during the vessel model generation process. Additionally, by adjusting the sample density or scaling the centerline point measures, coarse-tofine registration strategies are directly enabled. We present visualizations of the metrica nd its derivatives over a range of mis-registrations given different sample densities and different measure scalings using magnetic resonance angiograms, x-ray computed tomography images, and 3D ultrasound images.