Quantitative evaluation of iodine detectability in contrast-enhanced mammography using an anthropomorphic phantom

Abstract

Abstract Repeatable and standardized quantitative assessment of diagnostic image quality in contrast-enhanced mammography (CEM) is limited by the absence of realistic pseudo-anthropomorphic phantoms. We extended the L2 phantom mimicking realistic anatomical noise by integrating a plate containing 3D-printed iodine inserts of varying areal densities and diameters, enabling systematic evaluation of iodine detectability with a structured background. A key feature of the L2 phantom is the possibility to generate different background patterns by agitating the phantom. CEM images were acquired on a Siemens Mammomat Revelation system at 28 kV in TiCEM dual-energy mode with automatic exposure control. 8 readers participated in a Four-Alternative Forced Choice (4AFC) study with 12 independent acquisitions per condition, each performed after gentle phantom agitation to generate unique background configurations. Psychometric functions were fitted for each reader to determine detection thresholds set at 62.5% correct readings as a function of both cylinder diameter and iodine concentration. Detectability improved with increasing iodine content and cylinder size. Intra- and inter-reader variability decreased with higher signal conspicuity, and geometric distortion was negligible ( $$\approx 0.4\%$$ ). As a proof-of-concept study, the thresholds determined provide quantitative benchmarks for phantom-based comparison of CEM acquisition protocols and technical performance assessment on the investigated system. Overall, the L2 phantom with the added iodine contrast detail plate, combined with the 4AFC framework, offers a promising tool for evaluating image quality and assessing iodine detectability in CEM under controlled phantom conditions.