Alibaba’s DAMO RADAR AI is now available for cancer diagnosis using CT scans
Chinese technology company Alibaba, through its research institute DAMO Academy, has open-sourced the medical AI model DAMO RADAR. The neural network analyzes contrast-enhanced abdominal CT scans and can detect 146 different pathologies in a single analysis, including malignant tumors of the liver, pancreas, stomach, and colon.
Researchers published the model’s evaluation results in the scientific journal Science. On a dataset of nearly 40,000 real-world examinations, the neural network achieved an average AUC of 0.913. In a head-to-head evaluation, the AI outperformed 23 of 26 practicing radiologists from various hospitals.
Using the system as a clinical assistant reduced missed diagnoses by approximately 10%, while cutting the analysis time for a single scan by more than 30%. Engineers trained the model on a dataset of 400,000 CT scans paired with existing clinical reports, eliminating the need for extensive manual annotation.
The developers released the neural network’s code under the Apache 2.0 open-source license, while the model weights are available under the non-commercial Creative Commons BY-NC-SA 4.0 license on GitHub and Hugging Face. Medical institutions can run the model locally, allowing patient data to remain on-premises rather than being transferred to cloud storage. The system could help accelerate medical imaging analysis, particularly in regions facing a shortage of qualified radiologists.
