Heart segmentation with an iterative Chan-Vese algorithm

Olivier Rousseau (1) , Yves Bourgault (1)

1 Department of Mathematics and Statistics [Ottawa]

This paper presents 2D and 3D applications of the Chan-Vese model to heart and trachea segmentation. We improved the multi-phase Chan-Vese model by introducing an iterative method, by choosing an appropriate $L^1$ fidelity term as well as an efficient and prior free initial condition. For 3D applications, the algorithm is parallelized in order to speed up the computations. We provide extensive information on computational details, on the convergence times and on the quality of segmentations obtained. The results of the segmentations are then meshed to be used for finite element simulations.

Keywords: Heart Segmentation; Chan-Vese model; Active Contours; Mesh Generation

Domains: Analysis of PDEs [math.AP]

Olivier Rousseau, Yves Bourgault. Heart segmentation with an iterative Chan-Vese algorithm. 2008. ⟨hal-00403627v2⟩