A cerebral blood vessels segmentation method using a flux based second order tensor model

Official URL: http://dx.doi.org/10.1109/SIU.2014.6830437

In this paper, we view the segmentation of cerebral blood vessels from Digital Subtraction Angiography (DSA) and Rotational Angiography (RA) problem from a tensor estimation and tractography perspective as in diffusion tensor imaging (DTI). We have developed a flux based multi-directional cylinder model that fits to a second-order tensor whose principal eigenvector represents the vessel's centerline. This anisotropic tensor inside the vessel drives the segmentation analogously to a tractography approach in DTI analysis starting from a seed point used as initialization.