Image simulation and realism evaluation for mammography and tomosynthesis based on the detailed breast phantom

Abstract

The virtual image system is a key component of virtual clinical trials (VCTs) which can be utilized to evaluate and improve medical imaging devices. The main limitation of VCTs in the Chinese female breast is the lack of detailed structure in the breast model employed in the Chinese specification for testing of quality control in mammography. In this paper, based on Chinese female breast parameters, detailed breast phantoms with different glandularity (fatty, glandular, dense) and compressed breast thicknesses (CBT) (2–7 cm) were generated. Digital mammography (DM) and digital breast tomosynthesis (DBT) projections for these phantoms were simulated with clinical system configuration. Compressed breast volumes were reconstructed through DBT projections with different angles. Power spectrum analysis and fractal dimension measurement were applied for simulated and clinical images. The results show that the average power law exponents and standard deviation (DM, DBT) for clinical and simulated images were (3.26 ± 0.29, 3.33 ± 0.42) and (3.63 ± 0.25, 3.18 ±0.36) respectively. The average fractal dimensions and standard deviation (DM, DBT) for clinical and simulated images were (2.36 ± 0.03, 2.41 ± 0.05) and (2.10 ± 0.01, 2.26 ± 0.06) respectively. We constructed two virtual imaging systems for DM and DBT and obtained the imaging data based on the three-dimensional (3D) detailed breast phantoms. These results indicate that the texture indexes of simulated images are similar to that of the clinical images, and validate that these breast phantoms are suitable for imaging in VCTs requiring realistic anatomy with the Chinese female population. Results also show that the values of power law exponents for clinical and simulated images in DM and DBT are positively correlated with the value of breast glandularity, but the values of fractal dimensions keep steady for different glandularity.