Association between quantitative and qualitative image features of contrast-enhanced mammography and molecular subtypes of breast cancer

Background: The molecular subtype of breast cancer is one of the most important factors affecting patient prognosis. The study aimed to analyze the association between quantitative and qualitative features of contrast-enhanced mammography (CEM) images and breast cancer molecular subtypes. Methods: This retrospective double-center study included women who underwent CEM between November 2017 and April 2020. Each patient had at least 1 malignant lesion confirmed by pathology. The CEM images were evaluated by 2 radiologists to obtain quantitative and qualitative image features. The molecular subtypes were studied as dichotomous outcomes, including luminal versus non-luminal, human epidermal growth factor receptor (HER2)-enriched versus non-HER2-enriched, and triple-negative breast cancer (TNBC) versus non-TNBC subtypes. The association between the image features and molecular subtypes was analyzed by multivariate logistic regression, with odds ratios (ORs) and 95% confidence intervals (CIs) provided. Results: A total of 151 patients with 160 malignant lesions were included in the study. For quantitative features, a higher standard deviation of lesion density was associated with non-luminal (OR =0.88, 95% CI: 0.81 to 0.96, P=0.004) and HER2-enriched breast cancers (OR=1.16, 95% CI: 1.04 to 1.28, P=0.006). The relative degree of enhancement (RDE) and contrast-to-noise ratio (CNR) were not associated with molecular subtypes. However, a higher CNR/lesion size (OR=1.06, 95% CI: 1.01 to 1.12, P=0.012) was associated with luminal subtype cancers, and a higher RDE/lesion size (OR=0.94, 95% CI: 0.88 to 1.00, P=0.035) or a higher CNR/lesion size (OR=0.94, 95% CI: 0.88-1.00, P=0.038) was associated with non-TNBCs. For qualitative features, the presence of calcification was associated with HER2-enriched breast cancers (OR=2.91, 95% CI: 1.10 to 7.67, P=0.031). The presence of architectural distortion was associated with luminal cancer (OR=14.50, 95% CI: 1.91 to 110.14, P=0.010) and non-TNBC (OR=0.05, 95% CI: 0.00 to 0.43, P=0.022). Non-mass enhancement (OR=2.78, 95% CI: 1.08 to 7.14, P=0.033) was associated with HER2-enriched breast cancers. An association remained after adjustments for age, breast thickness, and breast density (all adjusted P<0.050). Conclusions: The quantitative and qualitative imaging features of CEM could contribute to distinguishing breast cancer molecular subtypes.