Alibaba's damo radar model achieves expert-level accuracy in medical imaging

On September 18, 2026, Alibaba's DAMO Academy announced the release of DAMO RADAR, a general-purpose medical imaging AI model capable of identifying over 146 diseases in a single scan. The model, developed in collaboration with Zhejiang University Affiliated First Hospital, has been published in the journal Science. DAMO RADAR uses a vision-language learning approach to align medical images with corresponding text reports, achieving expert-level accuracy in abdominal CT scans. The model's performance was evaluated across 18 organs, with an AUC of 0.913 in nearly 40,000 real-world tests. Zhang Jianpeng, a senior algorithm expert at DAMO Academy, noted that the model's 'organ-level fine-grained alignment' strategy improved diagnostic accuracy without additional manual labeling. In human-machine comparisons, DAMO RADAR outperformed 23 out of 26 radiologists, enhancing doctors' sensitivity by 10% and reducing diagnostic time by over 30%.