Localisation et détection de stents sur des images rayons-X bas contraste

Abstract

Stents are widely used to treat stenosis. Stent expansion is suspected to be an important factor of restenosis. The problem is that, in day to day practice, correct stent expansion is hard to assess. In X-ray images, stents are visible only with a low contrast. We propose to locate and detect them automatically in these images by applying dedicated algorithms. We characterize the problem of detecting stents in X-ray images as a crease detection problem. We compare classical image processing approaches. We propose two specific algorithms, one belonging to gray-level values methods and the other one based on adapted filtering. Comparison is made on the false detection results. Results have been tested on a hundred image database. In a second time, we propose a multilocal fuzzy algorithm to detect struts in a more accurate way.