Patient perceptions and attitudes towards the use of artificial intelligence in the symptomatic breast unit

Comprehensive Summary

Patients attending a symptomatic breast clinic were surveyed about their comfort with artificial intelligence in breast imaging, especially for mammogram interpretation. This was an observational cross-sectional survey done at Beaumont Breast Centre/Beaumont Hospital in Ireland using an anonymous paper questionnaire. Data were collected from July 08 2024, to October 04 2024, and 1534 participants were included. The main question was whether patients would accept AI assisting a radiologist rather than AI reading mammograms alone. Most responses supported AI as a tool, but not as a replacement. 61.0% agreed (n=935) with radiologist + AI double reading, while 66.9% disagreed with AI acting as the sole reader. Even if AI was described as more efficient, 75.4% still preferred a radiologist (n=1156). The same pattern emerged when AI was described as more accurate, with 66.1% still preferring a radiologist (n=1014). Overall, this suggests that patients still want human oversight, even when AI is framed as better. RRRs with 95% confidence intervals were used. Age 50–69 was associated with greater agreement with radiologist+AI readings than with neutral responses (adjusted RRR 1.55, 95% CI 1.04–2.32). Education also mattered. Patients with a bachelor’s degree or higher were more likely to agree with AI+radiologist reading (adjusted RRR 2.33, 95% CI 1.75–3.12), but also more likely to disagree with AI-only reading (adjusted RRR 2.10, 95% CI 1.58–2.80). The study is limited because the survey was only offered in English, and the demographics were not fully representative of the Irish population. Fairness analysis was not detailed. Since participation was voluntary and responses were mostly tick-box, selection bias is possible. Overall, patients seemed open to AI support, but not AI replacing radiologists.

Outcomes and Implications

Patients seem willing to accept AI in breast imaging when it clearly supports a radiologist rather than replacing them. Even when AI is described as more accurate or more efficient, most respondents still preferred human involvement, which suggests that technical performance alone may not be enough for adoption in the symptomatic setting. If AI tools are introduced into symptomatic breast clinics, they will likely need to be framed as a second reader or decision-support tool. A rollout that pushes AI as an independent interpreter may face strong patient resistance. The findings also highlight that trust and accountability are still major concerns, so patient education and clear communication would be important before AI is integrated into routine care. Overall, this study suggests that AI may improve workflow behind the scenes, but at the bedside, patients still want reassurance that a radiologist is making the final call.