Who needs experts?

Upper GI endoscopy screening is important for detecting early stage esophageal and gastric cancer. In this study, over one million upper GI endoscopy image sets from nearly 85K patients were used to train and validate GRAIDS, the Gastrointestinal Artificial Intelligence Diagnostic System. Using only standard white-light images, GRAIDS was able to achieve a diagnostic accuracy of 93%, falling slightly below that of expert (97%) and competent (96%) endoscopists but above that of trainees (89%). What’s more, GRAIDS sensitivity for malignant lesions was just as good as that of the experts, which is good enough for China where GRAIDS is currently being used for real-time AI diagnostic support during upper GI endoscopy. TBL: Love it or hate it, artificial intelligence could ease the burden of providing effective cancer screening to the masses. | Luo, Lancet Oncol 2019