Alibaba Group Holding's research division, Damo Academy, has open-sourced Damo Radar, a vision-language model trained to identify nearly 150 abdominal conditions from computed tomography scans. Detailed in a study published in Science, the system processes contrast-enhanced CT imagery across 18 abdominal organs, targeting pathology detection such as malignant tumors without requiring proprietary medical software suites.
During validation across nearly 40,000 real-world clinical examinations, the architecture evaluated 146 distinct findings. According to data reported in Science, the model outperformed most human radiologists participating in the benchmark evaluations.
Benchmark Performance and Imaging Expansion
Performance metrics published in the study establish an average area under the curve of 0.913 across the 146 clinical findings, benchmarked against an AUC of 1.0 for perfect diagnostic accuracy.
"the world’s first expert-level generalist medical imaging model"
The research team at Damo Academy used this designation while indicating that the underlying training methodology could eventually adapt to imaging modalities beyond contrast-enhanced abdominal CT scans.
Releasing open-weights that match expert radiologist benchmarks hands clinical engineering teams a validated diagnostic asset, fundamentally undercutting the pricing power of proprietary medical software vendors. While bridging the gap between an open-source repository and daily hospital integration remains a deployment hurdle, the availability of high-performance diagnostic weights lowers the barrier to entry for regional healthcare providers building local diagnostic tooling.