Customized Total Variation Algorithm for Metal Artifact Reduction in Computed Tomography

Metal artifact reduction (MAR) is a challenge for commercial CT systems. The metal objects of high density adversely affect the measurement process and bring difficulties to image reconstruction. Compressed sensing (CS) reconstruction algorithms have been successfully applied in MAR. Ideally, the desired anatomical information can be restored from incomplete projection data. However, in most practical cases, these conventional CS algorithms may instead introduce severe secondary artifacts due to improper prior information. In this paper, we propose a customized total variation (CTV) method to reduce the metal artifacts based on the specific pattern of the artifacts. The gradient operator within the TV norm is redefined according to the distribution of both the metal objects and tissues for each MAR case. We also provide a weighting strategy to further protect the fine details. Experimental results show that the CTV method achieves better performances than those of the conventional methods.