Dose Reduction for Digital Breast Tomosynthesis by Patch-Based Denoising in Reconstruction

Abstract

In digital breast tomosynthesis (DBT), it is desirable to achieve an appropriate level of image quality while keeping the radiation dose as low as reasonably achievable. The purpose of this study is to examine the effectiveness of a patch-based denoising algorithm in reducing noise while preserving details in DBT reconstruction. Low-dose DBT projection images were simulated with various levels of entrance exposure, based on the stochastic property of incident photons from the x-ray source. The patch-based algorithm estimates the true value of a pixel as a weighted average of all pixels in the projection image, where the weights depend on the similarity between the patches. Compared with local smoothing or filtering methods, patch-based techniques can reduce noise while preserving details. The preliminary results have demonstrated that the image quality of DBT can be potentially improved by the proposed technique by incorporating appropriate denoising into the iterative reconstruction algorithm. The suppressed noise was found to resemble the desired white noise except at sharp edges. The contrast is enhanced by more than 10% and the mean lesion signal-difference-to-noise ratio (SDNR) in homogeneous regions was increased by 131.8% and 76.4% for the entrance exposure of 0.1 R and 1 R per projection respectively. The proposed algorithm can further reduce the total imaging dose in DBT by allowing a reduced exposure for each projection view.