3D segmentation of vessels by incremental implicit polynomial fitting and convex optimization

Robust and accurate segmentation of blood vessels is important for treatment and diagnosis of cardiovascular diseases. Here, we introduce a new approach for 3D segmentation of vessels which is formulated as a convex parameter estimation problem and combined with an incremental tracking approach. Parameter values are determined as global optimum of a semidefinite program and admissible shape variations are imposed by convex constraints. The performance of the approach has been evaluated using 3D synthetic images and clinical 3D CTA images of the aorta including pathologies.