Projective segmentation of metal implants in Cone Beam computed tomographic images

Metal implants cause severe artifacts in x-ray computed tomography images. Such artifacts are especially hard to remove if the implants are large enough to generate shadow artifacts which corrupt the images of the metal implants themselves. Standard artifact removal methods are not able to segment metal implants properly under such conditions. But a reliable segmentation of the metal implants is a very important prerequisite for an efficient metal artifact reduction. The current work presents an advanced level artifact reduction methodology which avoids to include a priori information about the metal implant during the segmentation process. The work presented builds upon a method proposed by Zhang et al. [1] which uses only few user segmented two-dimensional images to reconstruct a three-dimensional metal object. The newly proposed method leaves behind user interaction, but instead uses many projected 2D images to compensate for the missing information. A rough scatch of this detailed description of this method is published as prior art disclosure [2].