The Extended Brief

Alibaba open-sources AI model that can detect cancer and nearly 150 conditions

Brief by The AI News AI newsroom · Sep 19, 2026, 3:02 AM EDT edition

Original reporting by Hacker News · published Sep 18, 2026, 7:54 PM EDT

A free, open-source model that reportedly reads abdominal CT scans at radiologist level could let any hospital or developer deploy broad cancer screening.

Key points

• Alibaba's Damo Academy open-sourced Damo Radar, a model detecting nearly 150 abdominal conditions, including cancers, in CT scans. source ↗
• In nearly 40,000 real-world exams, the model averaged 0.913 AUC across 146 clinical findings. source ↗
• The model outperformed most radiologists, according to a study published in Science. source ↗
• The vision-language model reads contrast-enhanced CT scans of 18 abdominal organs, trained on scans paired with clinical reports. source ↗
• The research team called it the world's first expert-level generalist medical imaging model. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• A hospital AI team can download Damo Radar and benchmark it against its own abdominal CT archive before paying for commercial radiology tools.
• A radiology department can pilot it as a second reader flagging findings across 146 conditions on contrast-enhanced scans.
• Imaging researchers can study its scan-plus-clinical-report training method for reuse on other modalities, as the team suggested.

Context

AUC, or area under the curve, scores how well a diagnostic model separates positive from negative cases, with 1.0 being perfect. Most medical imaging AI is built for a single disease or organ, so one model covering 146 findings is unusual. Open-sourcing means the model is publicly available for others to inspect, adapt, and deploy.

What to watch

• Independent validation on external datasets would test whether the 0.913 AUC holds outside the published study.
• Regulatory clearance or real hospital deployment would signal clinical adoption; failed replication would unwind the claims.

Related briefs

• New Records Reveal Problems with Medicare’s AI Prior Authorization Experiment
• A Pragmatic Randomized Trial of an EHR-Integrated Generative AI Chart Summarization Tool for Ambulatory Clinicians
• Generative design of novel bacteriophages with genome language models [R]
• Predicting Your Health Arc

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.