Metal artifact reduction in filtered backprojection reconstruction with iterative sinogram updates

Purpose These days, machine learning methods seem to outperform other methods in many applications. However, machine learning methods require large amount of training data, which may not be available for many applications. This paper proposes an effective metal artifact reduction method for X-ray computed tomography using the efficient filtered backprojection algorithm. Design/methodology/approach The proposed method is a not a machine-learning-based method. It is not a classic sinogram inpainting method either, because our proposed method does not estimate the metal corrupted data based on the neighboring data. Findings The proposed method is applied to some patient data with artificially introduced metals, and the metal artifacts are significantly reduced. Research limitations/implications In implementation of the proposed algorithm, care must be taken to use a fraction of the maximum image value to segment the metals so that the metal-affected sinogram regions can be identified. Practical implications One unique feature of the proposed method is that the metal objects are correctly estimated. Originality/value The novelty of the proposed method is to use the energy of the image artifacts as an objective function and to convert this objective function calculation from the image domain into the sinogram domain.