A mammography image set for research purposes using BI-RADS density classification.

Area of Science:

• Radiology
• Medical Imaging Analysis

Background:

• Consistent breast density categorization is crucial for research.
• Minimizing inter- and intra-observer variability is essential for reliable mammography research.

Purpose of the Study:

• To validate a set of mammography images for visual breast density estimation.
• To ensure consistency in future research projects.
• To determine observer performance in breast density assessment.

Main Methods:

• Utilized the Breast Imaging Reporting and Data System (BI-RADS) as the visual grading scale.
• Assigned 50 mammography images for density grade scoring by 8 observers.
• Evaluated observer agreement using kappa statistics and delta variance.

Main Results:

• Six out of 8 observers achieved near-complete intraobserver agreement (kappa > 0.81).
• Strong interobserver agreement (kappa = 0.61-0.8) was observed in 10-12 out of 28 paired observation episodes across two iterations.
• No observers exceeded a delta variance of 1, indicating limited variability in density estimation.

Conclusions:

• The validated mammography images are suitable for research purposes.
• While some observer variability exists, overall breast density classification agreement was strong.
• Future studies should consider digitally acquired images and repeat this validation process.