ROI-Based Quantitative Analysis in Contrast-Enhanced Mammography: A Feasibility Study for Reproducible Enhancement...

Background/Objectives: Contrast-enhanced mammography (CEM) has demonstrated high diagnostic accuracy in breast lesion detection and characterization. However, interpretation of enhancement intensity remains largely qualitative and subject to interobserver variability. This study aimed to evaluate the feasibility and reproducibility of a standardized region-of-interest (ROI)-based quantitative approach for enhancement assessment in CEM. Methods: In this single-center prospective study, 22 women undergoing CEM for suspicious findings on conventional breast imaging were enrolled, yielding a total of 27 lesions. Quantitative analysis was performed on recombined mediolateral oblique images acquired in both the early and delayed post-contrast phases. Three radiologists with different levels of experience independently performed ROI measurements using two standardized ellipsoidal ROI sizes (0.1 cm² and 0.2 cm²). ROIs were systematically positioned on the lesion, fibroglandular tissue, subcutaneous fat, and pectoralis major muscle. Relative ROI (rROI) metrics were calculated to reduce variability related to acquisition and post-processing factors. The primary normalized parameter was the lesion-to-parenchyma ratio (nROI). Histopathology served as the reference standard. Results: Of the 27 lesions, eight were benign (29.6%), eight were high-risk/B3 lesions (29.6%), and 11 were malignant (40.7%). Breast density categories were B in 11 lesions (40.7%), C in 13 lesions (48.1%), and D in three lesions (11.1%). Inter-reader reproducibility was good to excellent. Absolute lesion gray-scale values showed excellent agreement, with intraclass correlation coefficient (ICC) values ranging from 0.941 to 0.974, while nROI showed good reproducibility, with ICC values ranging from 0.861 to 0.894. No significant inter-reader differences were observed for normalized ratios. Comparison between ROI sizes showed small but statistically significant early phase differences for nROI, with median values of 1.009 for 0.1 cm² ROIs and 1.005 for 0.2 cm² ROIs (median paired difference -0.002; p = 0.002). Early phase nROI measured with the 0.1 cm² ROI differed significantly across breast density categories, increasing from density B to D (0.996, 1.010, and 1.015, respectively; p = 0.026). Conclusions: Standardized ROI-based quantitative analysis in CEM is feasible, and normalized enhancement ratios showed good inter-reader agreement, suggesting their potential to reduce variability related to subjective visual assessment and technical acquisition factors. Consequently, the proposed method provides a reproducible methodological framework for future larger studies aimed at validating quantitative CEM metrics.