By Melissa Silverberg
As AI continues to expand across radiology, new research presented on Wednesday shows that patients are generally positive about its use, but they overwhelmingly prefer that radiologists remain closely involved in interpreting their medical images. The survey, led by Hayley Briody, MBBCh, radiology trainee at Royal College of Surgeons in Ireland, is one of the largest to date examining patient attitudes toward AI in a general radiology population.
“While there is an abundance of literature highlighting the potential applications of AI, we noticed a relative paucity of literature focusing on patient attitudes,” Dr. Briody said. “This study adds significantly to the current conversation.”
The team distributed a voluntary 16-item questionnaire to patients attending a tertiary radiology department, gathering 1,041 total responses. Participants were asked about their demographics, general views on AI in health care, expectations for its use in imaging and willingness to share anonymized data for AI research.
More than half of respondents, 55.1%, agreed or strongly agreed that AI in health care is a good idea. Patients with higher levels of education showed significantly greater interest in AI, while those over age 70 were more likely to respond, “I don’t know.”
“A higher level of education may lend a better understanding of what the umbrella term ‘AI’ captures,” Dr. Briody said. “This in turn generates increased curiosity and interest.”
“Patients are showing us that the source of an interpretation matters. As AI evolves, we need to ensure that implementation maintains the trust patients presently have in radiologists.”
Hayley Briody, MBBCh
One of the clearest themes was a strong preference for shared human-AI interpretation. More than two-thirds (67.3%) were comfortable with a radiologist and AI reading their images together.
“Human oversight is the real takeaway message,” Dr. Briody said. “Patients are more comfortable with a combined approach because it’s closer to our current reality—an added layer rather than a total overhaul.”
Responsibility was also viewed as shared: when images were interpreted by both AI and a radiologist, only 20% of patients felt the radiologist alone should be accountable for errors, while 63% assigned responsibility to both.
By contrast, only 20.4% of respondents would accept AI alone as the sole reader, even if it was quicker or more accurate. “The main cause of concern for patients is standalone AI,” Dr. Briody said. “These findings suggest that while patients believe AI has a role, they do not necessarily agree that it should replace radiologists.”
More than half of patients, or 54% wanted to be asked for written consent if AI were used to interpret their imaging, despite the practical challenges such a requirement would pose.
“We anticipated that posters or leaflets might be acceptable, but less than 10% selected those options,” Dr. Briody said. “The preference for written consent was surprising and may reflect a fear that AI could be used as a standalone reader without their explicit approval.”
Despite their caution about standalone AI, most respondents supported the development of future tools, with 70.9% indicating they were willing to share anonymized images for AI training.
“This level of willingness is really encouraging,” Dr. Briody said. “AI development depends on large training datasets, and patient support can significantly impact this field.”
Dr. Briody said the findings help radiology departments shape their AI rollout with patients in mind. “Patients are showing us that the source of an interpretation matters,” she said. “As AI evolves, we need to ensure that implementation maintains the trust patients presently have in radiologists.”
The study authors noted limitations, including a predominantly white Irish population and an English-only survey. Dr. Briody said she hopes to explore these questions more in further studies that allow free text response to better understand patient concerns about AI.
Access the presentation, “Patient Perceptions of the Use of AI in a Tertiary Referral Radiology Department,” (W1-SSNPM03-2) on demand at RSNA.org/MeetingCentral.
© 2025 RSNA.
The RSNA 2025 Daily Bulletin is the official publication of the 110th Scientific Assembly and Annual Meeting of the Radiological Society of North America. Published online Sunday, November 30 — Thursday, December 4.
The RSNA 2025 Daily Bulletin is owned and published by the Radiological Society of North America, Inc., 820 Jorie Blvd., Suite 200, Oak Brook, IL 60523.
