Alibaba's Damo lab has released an open-source model that reads injected abdominal scans and covers 18 organs, with an average AUC of 0.913 across 146 clinical signs.
On September 18, 2026, Alibaba's Damo Academy released a medical imaging model called Damo Radar, presented as a generalist rather than a specialized detector. It analyzes abdominal CT scans with contrast agent and covers eighteen organs, enough to report nearly one hundred and fifty different conditions, including cancers.
The results published in Science show an area under the average curve of 0.913 out of 146 clinical signs, measured from about forty thousand examinations from real practice. For the record, a perfect score is 1.0. The authors say that the model does better than the majority of radiologists on this range of signs.
The main interest lies less in the raw performance than in the availability of the weights: until now, most imaging models of this caliber have remained captive to a supplier. The fact remains that of a precise protocol — the injected CT scan of the abdomen — and there is no indication of how these figures would behave on other machines or other patient populations.
Read the full article on the original source
Go to the source →