Evaluation of a breast software model for 2D and 3D X-ray imaging studies of the breast

Introduction: In X-ray imaging, test objects reproducing breast anatomy characteristics are realized to optimize issues such as image processing or reconstruction, lesion detection performance, image quality and radiation induced detriment. Recently, a physical phantom with a structured background has been introduced for both 2D mammography and breast tomosynthesis. A software version of this phantom and a few related versions are now available and a comparison between these 3D software phantoms and the physical phantom will be presented. Methods: The software breast phantom simulates a semi-cylindrical container filled with spherical beads of different diameters. Four computational breast phantoms were generated with a dedicated software application and for two of these, physical phantoms are also available and they are used for the side by side comparison. Planar projections in mammography and tomosynthesis were simulated under identical incident air kerma conditions. Tomosynthesis slices were reconstructed with an in-house developed reconstruction software. In addition to a visual comparison, parameters like fractal dimension, power law exponent beta and second order statistics (skewness, kurtosis) of planar projections and tomosynthesis reconstructed images were compared. Results: Visually, an excellent agreement between simulated and real planar and tomosynthesis images is observed. The comparison shows also an overall very good agreement between parameters evaluated from simulated and experimental images. Conclusion: The computational breast phantoms showed a close match with their physical versions. The detailed mathematical analysis of the images confirms the agreement between real and simulated 2D mammography and tomosynthesis images. The software phantom is ready for optimization purpose and extrapolation of the phantom to other breast imaging techniques. (C) 2017 Associazione Italiana di Fisica Medica. Published by Elsevier Ltd.