Artifact reduction methods for truncated projections in iterative breast tomosynthesis reconstruction

Digital breast tomosynthesis (DBT) mammography is a promising imaging technique that has the potential to improve detection of early-stage breast cancers. Digital breast tomosynthesis mammography can provide quasi 3-dimensional information by reconstructing the breast volume from a number of low-dose mammograms acquired over a limited angular range. Previous studies have shown that iterative reconstruction methods such as simultaneous algebraic reconstruction technique (SART) can give satisfactory image quality in DBT. However, because of the finite size of the detector and the limited field of view, DBT reconstruction contains artifacts caused by the truncated projection-view images. We developed methods that use a local intensity equalization strategy and a geometrical tissue-compensation method to remove two types of truncation artifacts: detector boundary discontinuity and underestimation of the attenuation path length. A custom-built breast phantom and a selected DBT case were used to evaluate the improvements. A GE (GE Global Research, Niskayuna, NY, USA) prototype DBT system was used to acquire 21 projection views in 3-degree increments over a +/-30-degree angular range. Experimental results demonstrated that the artifact reduction methods can improve the image quality at the boundaries with enhanced contrast-to-noise ratio and increased background uniformity, recover the obscured breast structural information, and achieve an overall reconstruction quality comparable with the quality of those without truncation.