Adaptive Normalized Metal Artifact Reduction (ANMAR) in Computed Tomography

Metal artifacts drastically impair the image quality in CT images and often reduce their diagnostic value. There are many publications on metal artifact reduction (MAR) which regard the metal-affected parts of the rawdata as completely unreliable and therefore replace it. While those sinogram inpainting methods are in general successful in removing metal artifacts, they cannot exactly recover the true values which are replaced and therefore the corrected image will exhibit new artifacts and blurring. In reference [1] we presented a new method, normalized MAR (NMAR), to replace corrupted rawdata while suppressing most of those new artifacts. Still, a complete replacement sheme is used there, which leads to loss of data. For metal implants with an elongated shape, as for example cardiac pacemakers or pedicle screws, the intersection length of the rays with metal strongly depends on the view direction. If the intersection length is rather short, the corresponding part of the rawdata still contains usable information. The replacement of corrupted projection data with NMAR turned out to be a robust solution, but close to implants the result may exhibit blurring due to a loss of data. In this work we introduce the adaptive normalized metal artifact reduction (ANMAR),an algorithm with rawdata-based merging of original and NMAR projections. It ensures that metal artifacts are reduced considerably, while excessive loss of data is avoided. This adaption could also be combined with MAR methods other than NMAR. We applied ANMAR to data sets from a modern clinical CT scanner, including a patient with a cardiac pacemaker, a patient with a fixateur interne, and a patient with a bilateral hip prosthesis. For comparison, they were also corrected with NMAR and with a standard linear interpolation method as described in reference [2]. ANMAR removes severe metal artifacts while it ensures that usable data from the metal trace is taken into account. Therefore, also details close to and even between implants can be better preserved