Alibaba DAMO Academy's Medical AI Model DAMO RADAR Published in Science, Detects 146 Abdominal Conditions and Goes Open Source

Alibaba DAMO Academy, together with the First Affiliated Hospital of Zhejiang University School of Medicine and other institutions, has developed a general medical imaging AI model called DAMO RADAR, which was recently published in the top international academic journal Science. The model is designed for abdominal contrast-enhanced CT diagnosis and can identify over 146 conditions at once, reaching expert-level accuracy for the first time. It is now officially open source.

DAMO Academy used a vision-language learning approach, letting the model directly learn the links between medical images and corresponding report text. The research was the first internationally to use an "organ-level fine-grained alignment" strategy, breaking CT data into anatomical units and precisely aligning images with report text at the tissue and organ level. This allows scalable, multi-purpose, and interpretable diagnosis without extra manual labeling.

DAMO RADAR covers a wide range of abdominal conditions including malignant tumors, evaluating 146 diseases across 18 organs. In nearly 40,000 real-world exams, it achieved an AUC of 0.913, becoming the first general medical imaging AI model to reach expert level.

In a human-machine comparison test with 26 radiologists, its average accuracy exceeded that of 23 of them. With AI assistance, doctors' sensitivity in detecting diseases improved by 10%, time spent decreased by over 30%, and junior doctors reached the diagnostic level of senior doctors.