By 2030, agreed a roomful of radiologists in Chicago this week, generative AI will be ubiquitous in their written work.
Medical imaging already leads the way in the clinical application of artificial intelligence: Algorithms that help to analyze CT scans, MRIs, and X-rays account for more than three-quarters of AI-based devices authorized by the Food and Drug Administration. But at the annual meeting of the Radiological Society of North America this week, the next generation of AI — large language models — was at the center of attention. Products using LLMs to streamline radiology documentation dominated the demonstration hall, and several sessions centered on health technology’s latest poster child.
“What we see in the hall has changed dramatically over a year,” said Weill Cornell radiologist George Lee Shih in a session debating the future of LLMs in radiology reports, the text that radiologists use to communicate their read of an image to the referring clinician and patients. “I think that this kind of tool will be so powerful that essentially it’s going to read our minds at some point,” said Shih, who consults for Open AI and an AI radiology reporting and annotation company called MD.ai.
This article is exclusive to STAT+ subscribers
Unlock this article — and get additional analysis of the technologies disrupting health care — by subscribing to STAT+.
Already have an account? Log in