Nature 577, 89–94 (2020)

An artificial intelligence system that can identify breast tumors from mammograms outperforms humans at the same task.

Early detection of breast cancer can considerably improve outcomes, and screening mammography helps make this possible. Currently, mammograms are interpreted by trained experts, but there are high rates of both false positives and false negatives.

McKinney and colleagues developed a deep learning–based model for identifying breast cancer in screening mammograms and tested it on two datasets from the UK and USA. They found that the artificial intelligence performance was better than that of radiologists and show how it could be integrated into clinical workflows in the future.