High-quality computed tomography (CT) images are key to clinical diagnosis. However, the current quality of an image is limited by reconstruction algorithms and other factors and still needs to be improved. When using CT, a large quantity of imaging data, including intermediate data and final images, that can reflect important physical processes in a statistical sense are accumulated. However, traditional imaging techniques cannot make full use of them. Recently, deep learning, in which the large quantity of imaging data can be utilized and patterns can be learned by a hierarchical structure, has provided new ideas for CT image quality improvement. Many researchers have proposed a large number of deep learning algorithms to improve CT image quality, especially in the field of image postprocessing. This survey reviews these algorithms and identifies future directions.
Keywords: Deep learning; Image denoising; Metal artifact correction; Super-resolution imaging.
© 2022. International Federation for Medical and Biological Engineering.