Vessel tree segmentation via front propagation and dynamic anisotropic Riemannian metric
Da Chen, Laurent David Cohen · 2016
In this paper, we present a blood vessel segmentation method by front propagation and anisotropic Riemannian metric. The front is defined as the level set of the geodesic distance to a set of given initial source points, with respect to a dynamic anisotropic Riemannian metric. The boundaries of the vessels can be represented by the level set at the given distance threshold. The anisotropic Riemannian metric can be defined using a prior estimate of the vessel orientations and the local intensity difference values, where the vessel orientations are detected by the oriented flux filter. Experimental results demonstrate the proposed vessel detection method indeed outperforms the traditional vesselness based detection method.