Unsupervised segmentation of stents corrupted by artifacts in medical X-ray images

We propose a new methodology for the segmentation of stents in 3D X-ray acquisitions. Such data are often corrupted by strong artifacts around the stent, requiring the development of a robust algorithm: because of the medical application, we need to produce an accurate segmentation. Moreover, we aim at developping a robust technique that can handle heterogeneous data. We propose a two-step, coarse-to-fine approach, that handles the corrupted cases. This approach leads to better results illustrated in the context of metallic artefact reduction.