| Abstract [eng] |
Artefacts in computed tomography images are a relatively common occurrence. These include beam hardening, motion, ring, noise, metallic and other types of artefacts. Among these, some of the most common are metallic and metal-induced artefacts like beam hardening and Poisson noise. In this paper, several models for metal artefact reduction in computed tomography images are examined. Three models were tested: the artefact disentanglement network, the reused convolutional network, and a reused convolutional network enhanced with sinogram information. The study showed that the models performed similarly quantitatively, as indicated by comparable values in metrics such as peak signal-to-noise ratio and structural similarity index measure. However, in qualitative assessments, it was determined that the reused convolutional network yielded higher-quality images compared to the other models. |
