Atenuação de artefato metálico em imagem de tomografia computadorizada com uso de Deep learning

Abstract

Metallic artifacts are serious problems of reconstruction of medical images of CT scans, which may cause significant difficulties in diagnosis, or even render it unfeasible. This research, developed in the Nucleus of Strategic Technologies in Health - NUTES, in partnership with the Regional Hospital of Trauma Dom Luiz Gonzaga Fernandes, generated a convolutional neural network architecture that attenuates the effects caused by metallic artifacts in images produced by computed tomography . The proposed architecture, after training, was able to reconstruct computed tomography images with metal artifact attenuation, with a result measured by qualitative evaluation through forms. The obtained results demonstrated that the technique is promising for the attenuation of metallic artifacts.