An image-based approach for reducing metal artifacts in CT

Various strategies have been developed to reduce metal artifacts in CT images, yet reduction of artifacts is successful to varying degrees. We proposed an image-based metal artifact reduction (MAR) approach incorporating an inpainting step, extraction of metal artifact-corrupted sinogram, and acquisition of a metal artifact-only image to reduce metal artifacts in an uncorrected image. In this work, a simulation study was conducted using a numerical pelvic phantom with bilateral metal inserts to demonstrate the effectiveness of the proposed method. For comparison, a conventional or standard sinogram inpainting method was also implemented. The proposed method produced visually comparable image quality to standard inpainting method. In addition, we particularly investigated the effects of the hard tissues, i.e. bones, that surround the metal on the MAR performance. We found that treating the surrounding bones together with the metals in the context of the proposed MAR scheme drastically reduces the metal artifacts that would otherwise overwhelm the image in conventional approaches.