Optimization of back projection with different geometry configurations of X-ray sources in digital breast tomosynthesis: A Monte Carlo simulation study

Breast cancer is the second leading cause of cancer-related mortality among women all over the world. The earlier the breast cancer is detected, the higher the possibility of cure. Digital breast tomosynthesis (DBT) is an effective technique for early breast cancer diagnosis. The DBT system obtains several projection images at a limited range of view angle and thus reduces the radiation dose delivered to patient effectively. Back projection (BP) is a classic algorithm for image reconstruction. In this work, 25 projection images were acquired by simulating seven different geometry configurations of X-ray sources on GATE (Geant4 Application for Tomographic Emission) platform. BP algorithm and its two variants (i.e., alpha-trimmed BP and PCA-based BP) were used for image reconstruction. Our experimental results demonstrated that among the seven geometry configurations of X-ray sources, the rectangular non-uniform configuration performed the best; it showed a better capability in eliminating out-of-plane artifacts and resolving spatial details than others. In addition, the a-trimmed BP had a better image quality among all of the three algorithms in the modulation transfer function analysis.