In computed tomography (CT) metal objects cause nonlinear variations of the acquired Radon data. During image reconstruction the inconsistencies introduce star shaped artefacts around the metal objects extending over the entire image. Due to this data deterioration, the acquired image is often unusable for diagnostic assistance or, at worst, causes a false diagnosis. In this work, metal artefact reduction (MAR) is performed in Radon space, in order to reduce the inconsistencies of the raw data. One-dimensional MAR-methods are considered, where two are based on polynomial interpolation and one uses a non-equispaced fast Fourier transform (NFFT) for interpolation. Since the common CT reconstruction, the filtered backprojection (FBP), is highly susceptible to inconsistent Radon data, a weighted, regularized Maximum Likelihood method can be used as alternative. The results of this reconstruction strategy are presented based on an anthropomorphic torso phantom (assigned with two steel markers). A comparison concerning the enhancement of the image quality and a discussion of principal limitations of the methods is given by means of the relative errors with respect to a ground truth (torso phantom without steel markers).
|Title of host publication||World Congress on Medical Physics and Biomedical Engineering, September 7 - 12, 2009, Munich, Germany|
|Number of pages||4|
|Publisher||Springer Berlin Heidelberg|
|Publication status||Published - 01.12.2009|
|Event||World Congress on Medical Physics and Biomedical Engineering: Diagnostic Imaging - Munich , Germany|
Duration: 07.09.2009 → 12.09.2009
Conference number: 81644