New Approaches for Metal Artifact Reduction in Computed Tomography

In computed tomography acquisitions, the presence of metal in an object can significantly disrupt the reconstructed image due to artifacts. Over the past few decades, the normalized sinogram interpolation concept has emerged as the established gold standard reference method for mitigating metal artifacts, as documented in the literature. This method involves the combination of linear sinogram interpolation applied to the uncorrected sinogram and a tissue normalization step. Metal artifact reduction frameworks employed in clinical routines typically integrate normalized sinogram interpolation with a frequency split based on image-based Gaussian high-pass filtering. This combination aims to restore anatomical details that may be erased during sinogram interpolation. However, both the sinogram interpolation and the ensuing frequency split present distinct methodological challenges, which are thoroughly addressed in this chapter. These challenges encompass the introduction of an additional regularization step aimed at minimizing interpolation-related artifacts. Moreover, an advanced frequency split is proposed, which facilitates a better restoration of details by suppressing high-frequency streak artifacts. These methodological advancements prove beneficial not only for computed tomography systems utilizing conventional energy-integrating detectors but also for those equipped with modern photon-counting detectors.