Research examining whether radiologists and large language models (LLMs) can identify AI-generated x-rays has received the 2026 Alexander R. Margulis Award for best original scientific article published in Radiology.
In the study, led by Mickael Tordjman, MD, and Murat Yuce, MD, of the Icahn School of Medicine at Mount Sinai in New York City, 17 radiologists from 12 centers in six countries reviewed authentic and synthetic radiographs generated with GPT-4o and the RoentGen model.
Only seven of the 17 radiologists spontaneously noticed that AI-generated images were present. When told that synthetic images were included, their mean accuracy in identifying them was 74.8% for the GPT-4o dataset and 70% for the RoentGen dataset, with years of experience offering no advantage. None of the four multimodal LLMs tested detected every synthetic image, the researchers found.
The award will be presented to Tordjman and Yuce during RSNA 2026 in Chicago on November 29.
