A deep-learning algorithm can convert low-dose CT scans into high-quality images, which may be even superior to images processed with iterative reconstruction

A deep-learning model can progressively reduce noise on low-dose CT images up to the optimal level; final images are either superior or comparable to images processed with iterative reconstruction and require a shorter processing time. This new approach can represent a serious competitor to the commercial reconstruction algorithms, effectively combining improved image quality with reduced radiation exposure.