Related Experiment Video
Updated: Apr 14, 2026
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
1.5K
Comparison of artificial intelligence (AI) services for Breast Imaging-Reporting and Data System (BI-RADS)
Yuriy Vasilev1, Anna Mayorova1, Denis Rumyantsev1
1Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Health Care Department, Moscow, Russia.
Quantitative Imaging in Medicine and Surgery
|April 13, 2026
Summary
Artificial intelligence (AI) shows suboptimal accuracy for individual mammogram BI-RADS categories but excels in binary classification to confirm absence of pathology. Further research is needed for clinical integration.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Existing AI research for breast cancer detection primarily focuses on malignant tumors (BI-RADS categories 4 and 5).
- The diagnostic performance of AI for other BI-RADS categories remains understudied.
- This study addresses the gap in understanding AI's accuracy across all BI-RADS categories.
Purpose of the Study:
- To compare the diagnostic accuracy of three mammographic AI services in predicting individual BI-RADS categories.
- To evaluate the potential for integrating AI into routine clinical practice for mammography analysis.
Main Methods:
- Utilized 81,895 anonymized screening mammograms from patients aged 40-75.
- Excluded mammograms without BI-RADS categories or with categories 0 and 6.
- Assessed AI performance against radiologists' opinions as ground truth, including calibration tests.
Main Results:
- Median accuracy across AI services was 76.9% with a positive predictive value (PPV) of 11.8%.
- Highest negative predictive values (NPV) were for BI-RADS 2 (78.5-83.4%) and BI-RADS 1, 3, 4, 5 (over 84.7%).
- Binary classification showed improved performance (median accuracy 80.5%, PPV 98.6%) compared to individual category prediction.
Conclusions:
- Most AI services demonstrated suboptimal metrics for individual BI-RADS prediction, possibly due to reliance on variable radiologist input and lack of histological calibration.
- Binary classification by AI showed higher performance and can be recommended to confirm the absence of pathology.
- Successful clinical integration of AI requires diverse diagnostic accuracy assessment methods tailored to specific use cases.
