Observer variation in the classification of mammographic parenchymal patterns

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Wolfe's classification identifies breast parenchyma patterns linked to cancer risk.
• Limited data exists on radiologist consistency in applying these mammographic patterns.

Purpose of the Study:

• To evaluate inter-rater reliability for Wolfe's four mammographic patterns.
• To compare agreement on pattern classification versus diagnosis and recommendations.

Main Methods:

• Ten expert radiologists classified 100 xeromammograms.
• Classification included mammographic pattern, radiological signs, diagnosis, and recommendations.
• Inter-rater agreement was measured using the Kappa statistic, correcting for chance agreement.

Main Results:

• High levels of agreement were observed for mammographic pattern classification.
• Agreement on mammographic pattern was significantly higher than for diagnosis or recommendations.

Conclusions:

• Radiologists demonstrate substantial agreement in classifying Wolfe's mammographic patterns.
• Mammographic pattern classification shows better reliability than other interpretation elements.