Loading...
Even among expert radiologists, accuracy varies when interpreting screening mammograms. Artificial intelligence (AI) has shown promise in the setting of other medical imaging applications; thus, investigators used large U.K. and U.S. clinical datasets to create a “deep learning” model for identifying breast cancer (biopsy-confirmed cases diagnosed within 3 and 2 years of the index mammogram in U.K. and U.S. women, respectively), then compared the performance of their AI-based system with that of clinical radiologists in either country. In the U.K., screening is triennial, with each image interpreted by two clinicians; U.S. screening is biennial, with each image read by one clinician. To confirm generalizability of AI's utility, an AI system…