Fast medical image segmentation through an approximation of narrow-band B-spline level-set and multiresolution

We have recently proposed a new level-set formulation, where the level-set is modelled as a continuous parametric function expressed in a B-spline basis. We propose in this paper to adapt this formalism to the class of narrow-band level-set methods, where the implicit function evolves only around its zero-level. For this purpose, we propose to model the interface by two lists of boundary points and we express the level-set evolution into the B-spline framework. We show that the flexibility of the method makes the algorithm well suited to the segmentation of 2-D and 3-D medical images. In particular, we introduce a multiresolution implementation of the method, yielding an efficient algorithm in term of computational time. The behavior of this approach is illustrated on medical images from various fields.