Clinical Report: AI Utilization in Breast Imaging – European Society Guidelines
Overview
Artificial intelligence (AI) integration in breast imaging, particularly mammography screening, enhances diagnostic accuracy and efficiency while reducing radiologist workload. Despite promising improvements in lesion detection and workflow, robust post-market surveillance and further validation studies are essential before widespread clinical adoption.
Background
Breast cancer screening relies heavily on imaging modalities such as digital mammography (DM), digital breast tomosynthesis (DBT), ultrasound (US), and MRI. AI systems have emerged as tools to augment radiologist performance by improving lesion detection, specificity, and reducing inter-reader variability. However, current evidence primarily supports AI as an aid rather than a replacement for radiologists, with limited data on long-term clinical outcomes and cost-effectiveness. The World Health Organization classifies AI medical devices into four evidence phases, emphasizing the need for continuous post-market surveillance to ensure safety and efficacy.
Data Highlights
Metric
Standard Reading
AI-Supported Reading
AUC
0.87
0.89
Sensitivity
83%
86%
Specificity
77%
79%
Reading Time (seconds)
146
149
Key Findings
AI tools in mammography screening increase AUC from 0.87 to 0.89, sensitivity from 83% to 86%, and specificity from 77% to 79% without prolonging reading time.
AI as a second reader can reduce radiologist workload by up to 88% while maintaining comparable accuracy.
Decision-referral models using AI improve detection of invasive lesions, small tumors, and dense breast tissue cases, outperforming unaided radiologists.
Commercial AI tools for DBT and ultrasound reduce interpretation time and inter-reader variability but lack extensive post-implementation evaluation.
AI applications in breast MRI offer limited added value for experienced radiologists, though advanced AI research shows promise in molecular subtype prediction and treatment response assessment.
Strong clinical recommendations require AI tools to reach WHO phase 4 evidence, involving continuous post-market surveillance.
Clinical Implications
Clinicians should consider AI as an adjunct to radiologist interpretation rather than a standalone diagnostic tool, integrating AI findings with clinical context. Implementation of AI in breast imaging demands rigorous post-market surveillance to ensure safety and efficacy. Adoption of AI-supported workflows may improve cancer detection rates and reduce workload but requires validation in diverse clinical settings.
Conclusion
AI holds significant potential to enhance breast imaging accuracy and efficiency, particularly in mammographic screening, but widespread clinical adoption must be guided by robust evidence and continuous monitoring. Future research should focus on long-term outcomes, cost-effectiveness, and large-scale validation.
References
European Society of Breast Imaging Guidelines 2024 -- Key Insights on AI Utilization in Breast Imaging
Radiologists assigned to receive step-by-step explanations from a large language model achieved higher diagnostic accuracy in a randomized vignette study, while differential-diagnosis outputs may have increased inappropriate reliance on incorrect model suggestions.
