Nearly 100% of patients surveyed say they’d want to know when AI is used in imaging

Nearly 100% of patients surveyed say they would want to know when artificial intelligence is used in medical imaging, according to new research published Tuesday. 

Previous papers have focused on provider and medical student perceptions about the technology, with literature lacking on consumer views. Researchers in Ireland recently sought to better understand how patients perceive rapidly evolving advancements in AI, sharing their findings in RSNA’s Radiology.

Out of over 1,000 survey respondents, about 96% said they believe patients should be informed when AI is used for reporting on their images. Of those, 53% preferred to give consent in writing versus 34% who would rather do so verbally. 

“Disclosure of adjunctive AI tools for image interpretation is not currently mandatory, and although patients have the right to be informed, obtaining written consent would be time-consuming,” Dr. Hayley Briody, with the Department of Radiology at Beaumont Hospital, Dublin, Ireland, and co-authors wrote June 30. “The Blueprint for an ‘AI Bill of Rights’ emphasizes a ‘right to notice and explanation;’ however, whether this should take the form of standardized notification or informed consent is not clearly defined,” they added. 

Briody et al. distributed the questionnaire to patients in waiting areas of their imaging department from July 2024 to April 2025. Most were between the ages of 50 to 69 and of white Irish ethnicity. When asked who would be at fault if an incorrect result was provided when images were read by AI and a radiologist, 64% selected both the doc and the technology. Briody and co-authors believe this reflects a patient “preference for transparency.” They highlighted previous research on this topic, underlining the importance of involving patient representatives in AI governance. 

The study is limited by several factors including selection bias, potential confounding due to single-language availability, and a lack of data on nonresponders. In addition, the total number of patients invited to participate (and thus the response rate) is unknown. 

“Notwithstanding, we hope this work prompts further research to determine whether attitudes differ when AI tools are used for secondary purposes, such as triaging radiology requests,” the authors concluded. “Overall, despite this study’s limitations, transparency appears important when implementing AI tools in radiology to ensure patient comfort with their use.”