RSNA Knee Abnormality Detection AI Challenge (2026)
This year's AI challenge engages researchers to develop models that detect and classify knee abnormalities.
The challenge is the first to use both images and text from radiology reports to train AI models. The dataset includes more than 5,000 knee MRI exams from 19 sites worldwide and associated radiology reports written in a dozen different languages.
How Can AI Help Evaluate Knee Injuries?
While MRI reveals damage to joint structures, including ligaments, menisci and cartilage loss, interpretation varies and access to specialty-trained musculoskeletal radiologists is limited. By training AI models on a large dataset of knee MRIs paired with their original multilingual reports, this challenge aims to develop tools that can reliably identify abnormalities and support more consistent, timely care.
Challenge Timeline and Recognition for Top Performing Teams
The competition is open now and runs until October. The top performing teams will share a total of $77,000 in prize money. Winners will be announced in November and the top performing teams will be recognized during RSNA 2026 (Nov. 29 – Dec. 3) at McCormick Place in Chicago.
Acknowledgments
RSNA would like to thank all who made this challenge possible.
View Acknowledgments
Contact Us
For more information, contact us at informatics@rsna.org.