The Second International Conference on Image Formation in X-Ray Computed Tomography

 Abstract—Imaging in the presence of prosthetic implants presents a notoriously difficult challenge to CT reconstrucion. Such hardware is made of alloys that are highly attenuating (e.g., Co-Cr-Mo) and impart severe degradation in image quality due to photon starvation, beam hardening, etc. An important clinical example is in the proliferation of total knee replacement, increasing the need for technologies capable of imaging in the presence of knee prostheses. The usefulness of CT in follow-up to knee replacement surgery is, however, extremely limited due to severe artifacts associated with the implant. Recent developments in likelihood-based CT reconstruction offer a potential solution to the problem. In particular, we exploit the fact that exact models of the shape and composition of prostheses are often available. A framework is proposed that extends earlier work [1,2] on known component reconstruction (KCR) to account for polyenergetic beam hardening and apply to the case of a large, highly attenuating object such as a knee implant. The proposed algorithm uses a polyenergetic object model to simultaneously estimate the unknown background density volume and the position and orientation of the known implant. We test the approach in studies emulating a recently developed, dedicated cone-beam CT scanner for extremities imaging. The results indicate substantial reduction of image artifacts and significant improvements in the visualization of areas adjacent to the implant. The KCR approach is found to outperform traditional filteredbackprojection and penalized-likelihood methods that do not account for the implant model or polyenergetic object attenuation. The method suggests promising new capability to assess implant integrity, loosening, and tissue disease (osteolysis and soft-tissue derangement).