On September 18, Alibaba’s DAMO Academy, in partnership with institutions including the First Affiliated Hospital of Zhejiang University School of Medicine, unveiled a general medical imaging AI model named DAMO RADAR. This groundbreaking model was featured in Science, a renowned international academic journal. Specifically designed for abdominal contrast-enhanced CT diagnosis, DAMO RADAR can detect over 146 conditions in a single scan, marking the first time an AI model has achieved accuracy on par with that of expert radiologists. The model is now officially available as open-source.

DAMO RADAR utilizes a visual-language learning approach and introduces, for the first time, an innovative “organ-level fine-grained alignment” strategy. In nearly 40,000 real-world examinations, it demonstrated an impressive AUC (Area Under the Curve) of 0.913. When benchmarked against 26 doctors, its average accuracy surpassed that of 23 of them. With AI-assisted support, doctors’ sensitivity in disease identification increased by 10%, while diagnosis time was reduced by over 30%, significantly enhancing diagnostic efficiency and accuracy.