Contrast-enhanced mammography (CEM) combines low-energy images with recombined images that show contrast enhancement. In a 2026 European Radiology analysis, lesion conspicuity on recombined CEM images was assessed as a potential predictor of breast malignancy. Conspicuity was graded as low, moderate or high relative to background parenchymal enhancement. The findings show that enhancement strength alone does not distinguish benign from malignant lesions.
Conspicuity Assessment on Recombined Images
The final sample comprised women with a median age of 49 years. The CEM examinations all had initial Breast Imaging Reporting and Data System assessments of 0, 3, 4 or 5. Examinations performed for biopsy-proven malignancy, evaluation of disease extent, short-term follow-up of a previous finding or without two-year imaging follow-up or biopsy pathology results did not form part of the final sample.
Must Read: Mammography-Linked Lung Screening Increases Uptake
Two breast radiologists reviewed the examinations while blinded to follow-up imaging and pathology outcomes. Each enhancing lesion previously identified on recombined images received a conspicuity grade of low, moderate or high. Low conspicuity meant enhancement equal to or slightly greater than background parenchymal enhancement. High conspicuity meant enhancement much greater than the background. Moderate conspicuity fell between these categories.
The radiologists also recorded enhancement pattern, including asymmetry, mass and non-mass enhancement with or without calcifications. They assessed whether each lesion appeared only on recombined images or also had a low-energy correlate. Patient age, breast cancer risk factors, breast density, background parenchymal enhancement and CEM vendor were also collected. One hundred randomly selected lesions received independent assessment by both radiologists to measure agreement.
Conspicuity Did Not Separate Benign and Malignant Lesions
Among 563 enhancing lesions, 55 were malignant and 508 were benign. Most lesions had low or moderate conspicuity. Low conspicuity accounted for 292 lesions, moderate conspicuity for 213 and high conspicuity for 58. The positive predictive value for malignancy remained low across all three categories, at 9% for low conspicuity, 11% for moderate conspicuity and 10% for high conspicuity.
Differences in positive predictive value between low, moderate and high conspicuity were not significant. The same pattern held across lesion subgroups, including lesions appearing only on recombined images, lesions appearing on both recombined and low-energy images, enhancing asymmetries, enhancing masses and non-mass enhancement. A multivariable model adjusting for risk factors, breast density and background parenchymal enhancement was also not significant.
Within the malignant lesions, 14 were ductal carcinomas in situ and 41 were invasive cancers, including four microinvasive carcinomas. Conspicuity did not differ significantly between ductal carcinomas in situ and invasive carcinomas. It also did not differ significantly by cancer enhancement pattern. Inter-reader agreement was high: the two radiologists agreed on conspicuity in 93 of 100 jointly assessed lesions, with almost-perfect agreement by kappa analysis.
Low-Conspicuity Cancers Often Lacked Low-Energy Correlates
The strongest difference among malignant lesions concerned whether cancers appeared on recombined images alone or also on low-energy images. Most high-conspicuity cancers appeared on both recombined and low-energy images. By contrast, most low-conspicuity and moderate-conspicuity cancers appeared only on recombined images. Among low-conspicuity cancers, 21 of 25 lacked a low-energy correlate. Among moderate-conspicuity cancers, 17 of 24 lacked a low-energy correlate.
Examples included low-conspicuity invasive lobular carcinoma, low-conspicuity invasive carcinoma and moderate-conspicuity invasive carcinoma seen on recombined images without a low-energy correlate. A high-conspicuity invasive carcinoma showed a low-energy correlate, while a high-conspicuity benign lesion was consistent with fat necrosis on subsequent imaging.
Technical features of CEM may affect the visual strength of enhancement. Unlike MRI, CEM does not image both breasts simultaneously in the axial plane. Image acquisition order varies between sides and planes, which makes conspicuity dependent on enhancement kinetics at the moment of acquisition. CEM enhancement is not an absolute iodine quantification. Recombined images arise from high-energy and low-energy acquisitions after iodinated contrast administration, followed by logarithmic transformation and weighted subtraction. Breast thickness, composition, density, scatter, misregistration, high attenuation artefacts, vendor systems and software processing can all affect image appearance. Lesion conspicuity is therefore unlikely to reflect lesion vascularity or tumour angiogenesis alone.
Lesion conspicuity on recombined CEM images is not independently associated with benignity or malignancy. Most enhancing lesions, including most cancers, show low or moderate conspicuity. Low conspicuity alone should not support dismissal of an enhancing CEM finding as benign, even when no low-energy correlate is visible. The evidence places emphasis on careful assessment of all enhancing lesions on recombined images, rather than reliance on enhancement strength as an independent malignancy predictor.
Source: European Radiology
Image Credit: iStock
References:
Amir T, Lee CH, Hogan MP et al. (2026) Clinical significance of lesion conspicuity on contrast-enhanced mammography. Eur Radiol. https://doi.org/10.1007/s00330-026-12660-y
