Related Experiment Video
Updated: May 2, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
42.6K
Predicting error in detecting mammographic masses among radiology trainees using statistical models based on BI-RADS
Lars J Grimm1, Sujata V Ghate1, Sora C Yoon1
1Department of Radiology, Duke University Medical Center, Box 3808, Durham, North Carolina 27710.
Medical Physics
|March 6, 2014
Summary
Breast Imaging-Reporting and Data System (BI-RADS) features can predict radiology trainee errors in mammogram mass detection. This may help create personalized educational tools for improving diagnostic accuracy.
Area of Science:
- Radiology
- Medical Imaging
- Diagnostic Accuracy
Background:
- Mammography is crucial for breast cancer detection.
- Radiology trainees face challenges in accurately detecting masses.
- Standardized reporting systems like BI-RADS aim to improve consistency.
Purpose of the Study:
- To investigate if Breast Imaging-Reporting and Data System (BI-RADS) features predict individual errors made by radiology trainees in mammogram mass detection.
- To assess the utility of BI-RADS lexicon in understanding trainee performance variations.
Main Methods:
- Ten radiology trainees and three experts reviewed 100 mammograms.
- BI-RADS features (breast density, mass shape, margin, density) and location were recorded for abnormalities.
- User-specific models predicted trainee error likelihood based on BI-RADS features, assessed by AUC.
Main Results:
- Individual models predicted trainee error likelihood with a mean AUC of 0.611 (p < 0.002).
- Despite inter-trainee error variability, BI-RADS features showed predictive power.
- The models demonstrated statistically significant ability to forecast errors.
Conclusions:
- Detection error patterns in mammographic masses by radiology trainees can be modeled using BI-RADS features.
- These findings suggest potential for developing personalized educational materials tailored to individual trainee needs.
- BI-RADS features offer a framework for analyzing and improving trainee performance in mammography.
