Segmentation and Quantification of Blood Vessels in 3D Images using a Right Generalized Cylinder State Model

Abstract

We present a vascular segmentation and quantification method based on the right generalized cylinder state model (RGC-sm). The RGC-sm model includes a curvilinear axis associated to a stack of contours. The axis is described by a state vector (local curvature, torsion and rotation). The contours are described by a Fourier series decomposition. The challenge is to automatically adjust this model to 3D vascular data (segmentation). By fitting the synthetic model to the actual medical data, it is possible to get the state model parameters and quantification measures. We present quantitative results on a set of calibrated phantoms and qualitative results on clinical datasets (carotid 3D-CTA and aortic 3D-MRA).