Artificial Intelligence‑Based Reconstruction for Chest Computed Tomography and Magnetic Resonance Imaging

Medical imaging modalities, especially computed tomography (CT), magnetic resonance imaging (MRI), X-ray, ultrasound and nuclear medicine, such as single-photon emission computed tomography or positron emission tomography (PET) fused with CT or MRI (PET/CT or PET/MRI) with various radioisotopes, have been used for management of patients with oncologic as well as other diseases and for some specific cancer screening requirements. However, images acquired with these imaging modalities may suffer from low signal-to-noise-ratio and low contrast-to-noise ratio as well as image artifacts. Image reconstruction techniques have, therefore, been developed during the last few decades to deal with these problems and to improve the quality of images for better visual interpretation, understanding, and analysis on CT and MRI. Moreover, deep learning (DL) techniques have been made available and successfully used from the late 2010s to the present as reconstruction methods for not only CT but also MRI. DL is far superior to traditional machine learning methods because it can learn features from raw input data during training. In this review, we describe (1) the basics of DL reconstruction (DLR), (2) clinical applications of DLR, and (3) future trends for DLR, all for CT as well as MRI.