Investigating poisson noise filtering in Digital Breast Tomosynthesis

Abstract

Digital Breast Tomosynthesis (DBT) is a potential candidate to substitute digital mammography in breast cancer screening. In DBT, projection images are acquired with low levels of radiation, which significantly increases image noise. In this work, we evaluate the effect of a denoising filter, designed for digital mammography, on the reduction of quantum noise in DBT images. This filter is based on an adaptive Wiener filter and the Anscombe transformation, to reduce Poisson noise without significantly affecting image sharpness. Denoising was applied to a set of synthetic DBT images generated using a 3D anthropomorphic software breast phantom. Images without noise was also created to provide ground-truth information. In order to evaluate the denoising performance in different steps of the DBT imaging, filtering was applied separately to the projections (before reconstruction) and to the tomographic slices (after reconstruction). The performance of the filter was evaluated considering qualitative and quantitative analysis of the images before and after denoising.