A self-directed learning intervention for radiographers rating mammographic breast density

Area of Science:

• Radiology
• Medical Imaging
• Breast Cancer Screening

Background:

• Subjective mammographic breast density (MBD) assessment by radiologists is susceptible to inter-reader variability.
• Standardized assessment criteria, like the Breast Imaging and Data Systems (BI-RADS^®) Atlas, aim to improve consistency.
• Variability in MBD assessment can impact patient management and screening outcomes.

Purpose of the Study:

• To evaluate the effectiveness of a concise, self-directed learning intervention in enhancing radiographers' reproducibility of MBD assessments.
• To quantify the impact of the intervention on inter-reader agreement among radiographers.
• To assess changes in agreement between radiographers and expert readers (radiologists and Volpara™ software) post-intervention.

Main Methods:

• Seven radiographers assessed MBD using BI-RADS^® percentage descriptors on two image sets (pre- and post-intervention).
• Inter-reader agreement among radiographers was measured using Weighted Kappa (к_w).
• Agreement between radiographers and expert readers (radiologists and Volpara™) was also assessed using Weighted Kappa (к_w).

Main Results:

• Radiographers' inter-reader agreement significantly improved from substantial (к_w=0.79) to almost perfect (к_w=0.84) after the intervention.
• Pre-intervention agreement between radiographers and expert readers (radiologists and Volpara™) was fair (к_w=0.24 for both).
• Post-intervention, agreement with expert readers showed a slight improvement but remained fair (к_w=0.31 with radiologists, к_w=0.28 with Volpara™).

Conclusions:

• A short, self-directed learning intervention effectively reduces inter-reader variability in mammographic breast density classification among radiographers.
• The intervention demonstrated a negligible impact on improving the agreement between less experienced radiographers and expert readers.
• Further strategies may be needed to bridge the agreement gap between novice and expert MBD assessment.