How the Model Was Tested and Validated
Alibaba’s Damo Academy has made its medical artificial intelligence model publicly available, enabling detection of cancer and close to 150 other health conditions from medical imaging. The model was evaluated on almost 40,000 real-world diagnostic exams and demonstrated performance exceeding that of most human radiologists in a study published in the journal Science. This release marks a significant step in expanding access to advanced diagnostic tools through open collaboration.
Breaking news
Canonical’s Web Presence Hit by Second DDoS Assault This Year
Loops AI Secures $3.5 Million to Power a New AI Operating System for E‑Commerce
Winverse AI Secures Tens of Millions in Seed Funding from DeepSeek and Long‑Z
Google Adds Pixel 9 Series to Certified Refurbished StoreThe AI system analyzes radiological images such as X-rays, CT scans, and MRIs to identify patterns associated with diseases ranging from lung and breast cancer to cardiovascular and neurological disorders. Trained on diverse datasets from hospitals across China, the model was designed to assist clinicians in under-resourced settings where specialist shortages delay diagnosis. By open-sourcing the technology, Alibaba aims to accelerate innovation and allow researchers worldwide to adapt and improve the model for local healthcare needs.
Can Open-Source AI Improve Global Health Equity?
In the study published in Science, researchers tested the AI on 39,872 real-world examinations collected from multiple hospitals, comparing its diagnostic accuracy to that of practicing radiologists. The model achieved higher sensitivity in detecting early-stage tumors and reduced false negatives in several categories, particularly in lung and breast imaging. Researchers noted that while the AI does not replace doctors, it can serve as a reliable second reader, especially in high-volume or time-sensitive environments. The validation process included blind reviews by senior radiologists to ensure unbiased comparison.
By making the model freely available, Alibaba seeks to lower barriers for medical institutions that lack the resources to develop proprietary AI systems. Experts suggest that such tools could help address disparities in diagnostic access, particularly in rural or developing regions where radiologists are scarce. However, challenges remain regarding data privacy, local regulatory approval, and the need for clinical integration support. The success of this initiative will depend on how effectively global partners can adapt the model to diverse populations and healthcare infrastructures while maintaining safety and accuracy standards.
What types of conditions can the AI detect? The model is designed to identify nearly 150 conditions, including various cancers, cardiovascular abnormalities, lung infections, and neurological issues, based on patterns in medical imaging data.
Frequently Asked Questions
Is the AI intended to replace radiologists? No, the model is meant to assist healthcare professionals by providing rapid, accurate preliminary readings, particularly in settings with limited access to specialists, not to substitute clinical judgment.
How was the model’s performance measured against human doctors? In the study, the AI’s diagnostic results were compared to those of radiologists on nearly 40,000 real-world exams, showing superior sensitivity in detecting certain diseases and fewer missed cases in key categories.
