Deformable boundary detection of stents in angiographic images

In this paper, a procedure is described for deformable boundary detection o f medical tools, called stents, in angiographic images. A stent is a surgic al stainless steel coil that is placed in the artery in order to improve bl ood circulation in regions where a stenosis has appeared. Assuming initiall y a set of three-dimensional (3-D) models of stents and using perspective p rojection of various deformations of the 3-D model of the stent, a large se t of synthetic two-dimensional (2-D) images of stents is constructed. These synthetic images are then used as a training set for deriving a multivaria te Gaussian density estimate based on eigenspace decomposition and formulat ing a maximum-likelihood estimation frame- work in order to reach an initia l rough estimate for automatic object recognition, The silhouette of the de tected stent is then refined by using a 2-D active contour (snake) algorith m integrated with a novel iterative initialization technique, which takes i nto consideration the geometry of the stent. The algorithm is experimentall y evaluated using real angiographic images containing stents.