Alibaba’s research arm, Damo Academy, announced the open‑source release of a new artificial‑intelligence model that can read contrast‑enhanced CT scans and identify nearly 150 abdominal conditions, including multiple cancers. Dubbed Damo Radar, the model marks a significant milestone in the firm’s push to bring expert‑level imaging tools to clinicians worldwide.
What happened
Damo Radar is a vision‑language model trained on about 40,000 real‑world CT examinations paired with their corresponding clinical reports. The system analyses scans of 18 abdominal organs and flags 146 distinct clinical findings. In validation tests, it achieved an average area under the curve (AUC) of 0.913, a metric where 1.0 denotes perfect diagnostic accuracy.
A comparative study involving 26 radiologists from several hospitals showed the model’s average accuracy surpassing that of 23 participants. When radiologists worked with Damo Radar, they reduced missed diagnoses by 10 % and cut interpretation time by more than 30 %. The research was published in Science and involved collaborators such as a hospital affiliated with Zhejiang University.
The model’s creators describe it as “the world’s first expert‑level generalist medical imaging model,” emphasizing its ability to handle a wide spectrum of diseases rather than a single target. Damo Academy plans to extend the training approach to other imaging modalities in the future.
Why it matters
Early and accurate detection of abdominal diseases—particularly cancers of the pancreas, stomach, and colon—remains a critical challenge in healthcare. Radiologists must interpret complex, high‑resolution scans, a process that can be time‑consuming and prone to human error. Damo Radar’s performance suggests AI can not only match but exceed human expertise in many cases, offering a safety net that catches subtle abnormalities.
The reported 10 % reduction in missed diagnoses could translate into earlier interventions for patients, potentially improving survival rates. Faster scan interpretation also eases radiology department workloads, allowing clinicians to focus on treatment planning rather than image triage.
Open‑sourcing the model amplifies these benefits. Researchers, hospitals, and developers worldwide can access the code, adapt it to local data, and integrate it into existing workflows without waiting for commercial licensing. This openness may accelerate validation studies, encourage transparency, and foster collaborative improvements.
The bigger picture
Damo Radar builds on a series of medical‑AI projects from Alibaba’s Damo Academy. Earlier tools targeted specific cancers—such as the Coca AI model released in April, which was claimed to be more sensitive than radiologists at spotting early‑stage colorectal cancer. The new model expands the scope from single‑disease detection to a comprehensive diagnostic assistant covering a broad array of abdominal pathologies.
Alibaba’s move reflects a broader trend among Chinese technology giants to apply cutting‑edge AI in healthcare. By investing in large‑scale data collection, model training, and partnerships with hospitals, firms aim to create AI‑driven screening tools that can be deployed across the country’s vast medical system. Open‑sourcing also aligns with global calls for greater transparency in AI research, especially in high‑stakes domains like medicine.
What happens next
The Damo Academy team indicated that the training methodology behind Damo Radar could be extended to other imaging types, suggesting future models for MRI, X‑ray, or ultrasound may follow. While the current release is open source, the researchers have not announced a commercial product rollout; instead, they appear to be inviting the broader community to test, refine, and potentially integrate the model into clinical practice.
The comparative study’s findings hint that collaborative workflows—where radiologists use the AI as a decision‑support tool—could become standard practice. If hospitals adopt Damo Radar widely, the reported efficiency gains and diagnostic improvements may reshape radiology department operations.
In the months ahead, the focus will likely shift to external validation of the model on diverse patient populations, integration testing with hospital information systems, and regulatory review where required. The open‑source nature of Damo Radar means that the pace of these developments will depend on the engagement of the global research and medical communities.
The release of Damo Radar underscores Alibaba’s ambition to position AI as a partner in medical diagnosis, offering a powerful, openly accessible tool that could enhance both accuracy and efficiency in abdominal imaging.
