Dynamic Contrast-Enhanced Mammography for Differentiating Benign and Malignant Breast Lesions
2Gray Faculty of Medicine, Tel Aviv University, Israel
Background
Dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) evaluates lesion enhancement over time to differentiate benign from malignant breast lesions. However, the role of DCE in contrast-enhanced mammography (CEM) remains unclear.
Methods
This retrospective study included 41 benign and 24 malignant lesions with three projections available. Two independent readers assessed enhancement dynamics on recombined images, with persistent enhancement considered suggestive of benignity and plateau or washout as suspicious patterns. Diagnostic performance and inter-reader agreement (Cohen’s kappa) were calculated.
Results
Diagnostic performance varied by reader and projection. For Reader 1, craniocaudal (CC) combined with mediolateral oblique (MLO) versus lateral projections (CC+MLO vs. CC+Lateral) showed sensitivity 87.5% vs. 79.2%, specificity 43.9% vs. 31.7%, positive predictive value (PPV) 47.7% vs. 40.4%, negative predictive value (NPV) 85.7% vs. 72.2%, and area under the curve (AUC) 0.657 vs. 0.554; corresponding values for Reader 2 were 75.0% vs. 66.7%, 53.7% vs. 73.2%, 48.6% vs. 59.3%, 78.6% vs. 78.9%, and 0.643 vs. 0.699. Sensitivity and specificity did not differ significantly between MLO and lateral projections (all p>0.05). Inter-reader agreement was fair for CC+MLO (κ=0.328) and CC+Lateral (κ=0.360), increasing with binary grouping to κ=0.579 (80.0% agreement) and κ=0.371 (66.2%), respectively.
Conclusions
DCE CEM may aid differentiation of benign from malignant breast lesions. However, diagnostic overlap and limited inter-reader agreement may reduce its clinical utility, while the added value of the delayed lateral phase remains uncertain. Further quantitative analysis may provide a more objective assessment and improved performance.