Tomosynthesis reconstruction using an accelerated expectation maximisation algorithm with novel data structure based on sparse matrix ray-tracing method

Abstract

Digital Breast Tomosynthesis (DBT) is a novel imaging technology to improve early breast cancer detection. It provides three-dimensional information of the breast to overcome the critical issues of overlapping anatomical structures of the breast. Among current available DBT reconstruction algorithms, Maximum Likelihood Expectation-Maximisation (MLEM) is a time-consuming iterative method to reconstruct three-dimensional image of the breast. In this paper, we proposed an accelerated MLEM algorithm with novel data structure based on sparse matrix ray-tracing method for DBT reconstruction. Compared with the standard MLEM, the proposed algorithm is effective to generate relative fast-speed tomosynthesis reconstruction and maintain the same image quality.