Purpose: Contrast-enhanced digital breast tomosynthesis (CEDBT) is being investigated for breast cancer detection. The image quality of CEDBT is degraded by the cupping artifacts due to scattered radiations, which hinders the characterization of lesion enhancement. We aim to develop an iterative kernel-based method for scatter correction using scatter point-spread-function (SPSF) from Monte-Carlo (MC) simulation.
Methods: The SPSF were simulated using GATE for breasts with 50% glandularity and different thicknesses for various DBT projection angles, and used to compute the scatter-to-primary-ratio (SPR) kernels. The SPR kernels were convolved with the projection images to generate the initial estimate of scatter. The errors in scatter estimate were further reduced by iterative image convolution. The accuracy of the scatter estimate was tested on phantom images. Clinical images of CEDBT from an IRB-approved study were used to evaluate the reduction of cupping artifacts for filtered-back-projection (FBP) reconstruction and simultaneous-algebraic-reconstruction-technique (SART).
Results: The phantom study shows that the proposed method provides scatter correction close to that when anti-scatter grid is used. The cupping artifacts on the clinical images of CEDBT from FBP and SART reconstruction are reduced after scatter correction.
Conclusion: The proposed method provides more accurate scatter estimate for projection images in DBT and reduces cupping artifacts in CEDBT. It may improve the assessment of contrast-enhanced lesions in CEDBT.