Assessing Artificial Intelligence in Breast Imaging: A Survey of Breast Radiologists' Insights on Adoption, Benefits, and Challenges.
Authors
Affiliations (5)
Affiliations (5)
- Jefferson Einstein Hospital, Philadelphia, PA, USA.
- Faculty of Medicine, University of Khartoum, Khartoum, Sudan.
- Detroit Medical Center, Detroit, MI, USA.
- Solis Mammography, Addison, TX, USA.
- Temple University Hospital, Philadelphia, PA, USA.
Abstract
To evaluate radiologists' opinions on the clinical applications of artificial intelligence (AI), especially AI-based computer-aided detection (CAD), in breast imaging. An IRB-exempt anonymous survey was distributed to Society of Breast Imaging (SBI) members on May 10, 2024, and May 20, 2024. Survey questions included practice demographics and perspectives on AI. Results were analyzed using descriptive statistics. In all, 7.2% (162/2264) SBI members responded, with 69 (42.6%) respondents being in private practice and 56 (34.6%) respondents in academics. Artificial intelligence-aided CAD was used by 90 (55.6%) respondents. Reported benefits of AI-CAD included improved work efficiency (98, 71.0%), increased cancer detection (90, 65.2%), and reduced recall rates (54, 39.1%). Artificial intelligence-aided CAD was reported as being best for detecting calcifications (92, 61.7%) and architectural distortion (65, 43.6%), although the highest false-positive results were reported in postsurgical scar (104, 75.9%) and benign calcifications (95, 69.3%). Additionally, respondents reported AI tools have great impacts on workflow efficiency (109, 73.2%), patient care (98, 65.8%), reducing burnout (70, 47.0%), turnaround time (70, 47.0%), and addressing radiologist shortages (69, 46.3%). Barriers to AI adoption included cost (113, 71.5%), software integration (98, 62.0%), and lack of trust (100, 63.3%). Only 41 (27.5%) felt AI might threaten job security, and 52 (34.9%) believed it could reduce reimbursement. Most survey respondents believe AI enhances breast cancer detection and workflow efficiency, but false-positive results, particularly in postsurgical scar and benign calcifications, limit its diagnostic utility. Barriers such as cost, integration, and trust hinder AI adoption. Refining AI algorithms to minimize false-positive results and addressing these barriers are crucial for clinical implementation.