To help ease some of their workload, researchers presented a new AI assistant called RADAR (Rapid Abdominal Diagnosis with AI and Radiology).
The tool is a generalist AI system that can analyze abdominal CT scans and assist doctors in reaching diagnoses.
It is built on a scalable vision-language framework trained directly on clinical reports covering 424,911 examinations, including 1.5 million image-text pairs and more than 15 million anatomy-specific pairs, without requiring experts to manually tag or label the data beforehand.
When RADAR was tested on real-world hospital examinations, it achieved an average diagnostic accuracy score (AUC) of 0.913, outperforming existing medical vision-language AI models.
The results held up outside the original setting across eight different hospitals, emergency room cases and diverse patient populations.
According to findings published in Science, RADAR accurately identified 18 anatomical structures and 146 imaging signs and diseases, regardless of where or how it was tested.
【MORE】