Alibaba open-sources a free AI model that reads CT scans for 150 diseases

Damo Academy's new model, called Damo Radar, flags abdominal cancers and other conditions on CT scans, and Alibaba is releasing it for free.

What happened: Alibaba's research arm, Damo Academy, has open-sourced an AI model called Damo Radar that reads contrast-enhanced CT scans of the abdomen. It's designed to spot nearly 150 conditions, including cancers, across 18 organs. In tests on almost 40,000 real-world scans, it hit an average accuracy score (AUC) of 0.913 across 146 clinical findings, where 1.0 would be a perfect score.

Why it matters: Reading CT scans is slow, specialist work, and caseloads keep growing. A tool that can flag likely problems across dozens of organs at once could give radiologists a faster second opinion, especially useful where specialist coverage is thin. Because Alibaba is releasing it openly rather than selling it as a locked product, any hospital or research team can inspect, test, or adapt it without a licensing deal.

How it works, plainly: Damo Radar is a vision-language model, meaning it was trained on CT scans paired with the written clinical reports doctors produced from them. That let it learn the visual patterns linked to each diagnosis. When it reads a new scan, it outputs likelihood scores for each of the 146 conditions it was trained to recognize, acting like a checklist a radiologist can weigh against their own read.

The rollout: Damo Academy calls it the first 'expert-level generalist' medical imaging model and says the same training approach could later be applied to other types of scans beyond abdominal CT. The model is now open for outside researchers and institutions to download and experiment with. What isn't addressed is whether, or how, it would be validated and approved for actual bedside use anywhere.