Monday, November 30 | 9:30 a.m.-10:30 a.m. | M3-SSIN02-7 | Room E450B
A large language model (LLM) that drafts report impressions can save radiologists time, and the benefits appear to grow as users become familiar with it, according to this scientific presentation.
Presenter Kailash Singh of the Cleveland Clinic Lerner College of Medicine and colleagues evaluated how LLM-generated impressions affected radiologist efficiency across multiple imaging modalities at a large academic health network.
Using over 2.7 million reports generated from September 2024 through February 2026, the researchers compared how long radiologists spent on impressions with and without LLM assistance, as well as how often reports needed addenda. They examined two periods: a nine-month trial in which selected readers were asked to use the software, followed by a full rollout in which use was voluntary.
CT and MRI showed significant time savings in both periods, while ultrasound showed a benefit after full implementation, the researchers found. Addendum rates didn’t differ during the trial. After the full rollout, however, LLM-assisted reports needed significantly fewer addenda, both overall and across CT, MRI, ultrasound, and x-ray These findings support the possibility that “workflow familiarity may improve both efficiency and downstream report quality,” according to the researchers.
“This study reveals modality-specific benefits of LLM documentation assistance in radiology and highlights opportunities for improved efficiency and accuracy in clinical practice,” the authors wrote.
Learn more by attending this Monday morning session.



