Clinical Report: AI Integration Reduces Radiologist Time in Mammogram Screening
Overview
This study evaluates the potential time savings for radiologists by integrating AI into mammogram screen-reading within BreastScreen Norway. Findings suggest that substituting one of two radiologists with AI could significantly reduce workload without compromising diagnostic accuracy.
Background
Mammographic screening is essential for early breast cancer detection and typically involves double reading by two radiologists to ensure accuracy. This process, while effective, is resource-intensive and challenged by a global shortage of radiologists. AI systems have demonstrated comparable diagnostic performance and may reduce radiologist workload by replacing or assisting one reader. However, the impact of AI on time savings and cost-effectiveness in real-world screening programs remains underexplored.
Data Highlights
Parameter
Value
Target population
~680,000 women aged 50–69
Attendance rate (2024)
77%
Consensus rate
7.2%
Recall rate
3.2%
Screen-detected cancer rate
0.63%
Mean reading time per mammogram
41 seconds (after exclusions)
Median reading time per mammogram
25 seconds
Consensus/arbitration time
5 min × 2.5 radiologists = 12.5 min per case
Recall assessment time
60 min per case by one radiologist
Working hours
7 hours/day, 5 days/week, 180 days/year
Key Findings
Double reading by two radiologists is standard in BreastScreen Norway, with consensus discussions for cases scored ≥2.
Mean reading time per mammogram is approximately 41 seconds, with a median of 25 seconds after excluding outliers.
Consensus/arbitration discussions involve about 12.5 minutes of radiologist time per case (5 minutes × 2.5 radiologists).
Recall assessments require approximately 60 minutes of radiologist time per positive screening case.
Replacing one radiologist with an AI system could substantially reduce the initial reading workload and overall radiologist time commitment.
Time estimates were conservatively set higher than measured values to avoid underestimation of radiologist workload.
Clinical Implications
Integrating AI as a substitute for one radiologist in mammographic screening can alleviate the workload pressures on radiology services, especially amid personnel shortages. This approach may maintain diagnostic accuracy while improving efficiency and resource allocation in population-based screening programs. Careful selection of AI risk thresholds and workflow integration is essential to optimize benefits.
Conclusion
AI integration in mammogram screen-reading offers a promising strategy to reduce radiologist workload and save time without compromising screening quality. Adoption in organized programs like BreastScreen Norway could enhance sustainability and efficiency of breast cancer screening services.
References
BreastScreen Norway Program Overview and Protocols
Cancer Registry Regulations and Personal Health Data Filing System Act
Studies on AI Diagnostic Performance in Mammography
Workload and Time Estimates in Breast Radiology Departments
