An expert-level generalist AI for abdominal CT diagnosis | Science

Researchers developed RADAR, a generalist vision-language AI model trained on over 400,000 abdominal CT studies that achieves expert-level diagnostic performance across diverse clinical tasks. Unlike prior supervised approaches limited to narrow use cases, RADAR handles a broad range of abdominal pathologies including detection and characterization of organ lesions. The study, published in Science, demonstrates that a single foundation model can match specialist-level radiology across multiple diagnostic scenarios. This represents a significant step toward clinically deployable general-purpose AI in abdominal imaging, with direct implications for cancer detection in organs such as the liver, pancreas, kidneys, and colon.