Mammography test sets: reading location and prior images do not affect group performance

Area of Science:

• Radiology
• Medical Imaging Analysis

Background:

• Mammography interpretation is crucial for breast cancer detection.
• Optimizing reading conditions can potentially improve diagnostic accuracy.

Purpose of the Study:

• To evaluate the influence of reading location (clinical vs. laboratory) and prior image availability on mammographic interpretation performance.
• To assess the impact on group and individual reader performance metrics.

Main Methods:

• 10 expert readers interpreted 200 mammographic cases under varied conditions (clinical/laboratory, with/without prior images).
• Region-of-interest (ROI) figure-of-merit (FOM), sensitivity, and specificity were analyzed using JAFROC and Wilcoxon tests.
• Reader agreement was assessed using Kendall's coefficient of concordance.

Main Results:

• Group performance metrics (ROI FOM, sensitivity, specificity) and reader agreement were comparable across all tested reading conditions.
• Individual analysis revealed decreased ROI FOMs in two readers when prior images were unavailable.
• The impact of reading location on individual performance was inconsistent.

Conclusions:

• Neither reading location nor the availability of prior images significantly affected overall group performance in breast test-set reading.
• While group performance remained stable, individual reader performance can be affected by the absence of prior images.