3D medical image reconstruction on Digital Breast Tomosynthesis

Abstract

The present work aim to evaluate image reconstruction algorithms for Digital Breast Tomosynthesis. Simulated data was used to perform and assess the 3D reconstruction. The images were acquired in a Monte Carlo platform and using an analytical phantom. The reconstruction step was performed, implementing Filtered Back Projection (FBP), Maximum-Likelihood Expectation-Maximization (ML-EM) and Algebraic Reconstruction Technique (ART). Comparable results were obtained by the three algorithms. The FBP algorithm presented more blurring images than the ML-EM and ART algorithms. However, it was the one more capable to localize the structures on the 3D space, including the smallest details. The results of the 3D reconstruction allow the discrimination even of very small structures which could not be differentiated on the simple projections that result from the simulations. This indicates that the accuracy of Digital Breast Tomosynthesis can be better than the Mammography.