AI Decision-Support Model Improves Consistency of Breast Density and BPE Assessment on Contrast-Enhanced Mammography

An AI-assisted decision-support model using patient-level variables reduced categorization disagreement by 26% for breast density and background parenchymal enhancement on contrast-enhanced mammography, achieving AUC 0.75, precision 0.72, and recall 0.69, with greatest benefit i…

• Retrospective analysis of 213 CEM exams with BI-RADS 4–5 malignancy showed AI decision-support associated with 26% fewer categorization disagreements and higher inter-reader agreement (Fleiss' κ) vs. conventional assessment.
• Model used expert-derived variables (BPE grade, breast density, age) to generate continuous estimates, not autonomous image grading; performance AUC 0.75, precision 0.72, recall 0.69, particularly aiding dense breasts (BI-RADS C/D).
• Single-center, malignancy-enriched design limits generalizability to screening; external testing on VinDr-Mammo evaluated only density prediction robustness, not the full framework.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.