Application of non-uniformity correction framework in digital breast tomosynthesis by incorporating a total variation penalty

Abstract

Digital breast tomosynthesis (DBT) is well-established in breast imaging modality to overcome the loss of depth-information in a three-dimension (3D) object in 2D mammography. However, the reconstructed DBT images suffer from shading artifacts caused by scatter radiation, beam-hardening effect, and unique geometric conditions. In this study, we investigated the application of a non-uniformity correction framework with TV-penalty in the DBT system for substantial improvement of its image visibility. We modeled a modified solver to estimate the non-uniformity field component, which is inspired by the normalized intensity-gradient joint-histogram-based energy minimization. The proposed framework was compared to the reference slice with only primary component in Monte-Carlo simulation and the uniformity was experimentally improved using the projection at low exposure condition. To quantitative evaluate image performance, percent integral uniformity (PIU) was calculated in acquired images. Our results indicated that the proposed framework well-operated to compensate for distorted flatness due to beam hardening, scatter, and unique DBT geometry. Especially, the measured PIU value of the corrected DBT slice was 14.2, it is almost same compared to that of mono-energy DBT slice (primary only) in simulation study. Moreover, the profile flatness of the corrected DBT slice was significantly improved from conventional DBT slice with much noise component. The uniformity correction of the DBT slice is necessary to reduce the misdiagnosis. Consequently, our results demonstrate that the proposed framework successfully corrects the non-uniformity at low X-ray exposure condition, and its scheme is expected to be applicable for efficient DBT imaging performance. © 2021 Elsevier GmbH