Breast cancer screening with AI - detection, tumor types, and breast density

Artificial intelligence (AI) software has potential to partly or fully replace breast radiologists in mammography screen reading. There is however a lack of knowledge of the AI software performance in non-enriched screening materials. The overall aim of this project is to explore the diagnostic accuracy of an AI software compared with radiologists in a retrospective analysis of >96 000 mammography screening examinations. Secondary aims are to compare the diagnostic accuracy in different breast density groups, based on automatic software density assessment, detected tumor types and to explore the ability of AI in reducing interval cancers.