Early adoption and perceived clinical impact of breast cancer detection artificial intelligence tools: Survey of society of breast imaging radiologists.
Authors
Affiliations (3)
Affiliations (3)
- Department of Radiology, Breast Imaging Division, UC San Diego Health, San Diego, La Jolla, CA, USA.
- Altman Clinical and Translational Research Institute, University of California, San Diego, La Jolla, CA, USA.
- Department of Radiology, Breast Imaging Division, UC San Diego Health, San Diego, La Jolla, CA, USA; Lucy Curci Cancer Center, Eisenhower Medical Center, Rancho Mirage, CA, USA. Electronic address: [email protected].
Abstract
To evaluate radiologists' early adoption, perceptions, and radiologist-reported clinical impact of breast cancer detection artificial intelligence (AI) tools in breast imaging practices. An online survey was distributed to members of the Society of Breast Imaging to assess perceived clinical impact of FDA-cleared AI tools for mammographic breast cancer detection. Respondents were categorized as AI users or non-users. Responses were summarized descriptively, and differences were analyzed using Fisher's exact test. A total of 215 radiologists responded. Of these, 47% had implemented breast cancer detection AI tools (AI users), 11.2% are planning to implement them, and 41.8% had not implemented diagnostic AI tools (non-AI users). Among non-AI users, the most reported barrier to adoption was the implementation cost (53.3%). The most frequently perceived appropriate use of diagnostic AI tools was as a second reader. Non-AI users more often anticipated reductions in recall rates (59.3%) compared with AI users reporting reductions (34.7%; p = 0.003). A similar pattern was observed for biopsy rates (36.4% vs 9.1%; p < 0.001). Non-AI users also more frequently anticipated reduced burnout than AI users (56.0% vs 29.4%; p < 0.001). Perceptions of patient outcomes were similar between groups (p = 0.87). Although early adoption of breast cancer detection AI tools is notable, only a minority of AI users reported meaningful clinical benefit across measured outcomes, falling short of the benefits anticipated by non-AI users. Continued validation and refinement of AI tools is needed to ensure meaningful clinical impact.