Implementation of Majorization-Minimization (MM) Algorithm for 3D Total Variation Minimization in DBT Image Reconstruction

Abstract

Digital Breast Tomosynthesis (DBT) is a developing imaging modality which produces 3D images of a breast. Iterative image reconstruction techniques, such as Algebraic reconstruction technique (ART), have been proposed to help increasing success in detecting masses and micro-calcifications. To enhance the quality of reconstructed image, total variation (TV) minimization was applied to the images reconstructed by ART. Nowadays, the number of published papers dealing with 3D TV minimization on ART (ART+TV3D) tends to increase. On the other hand, in the signal processing literature, a new majorization-minimization (MM) algorithm on TV denoising is described for an N-point x(n) as 1D signal. According to our literature review, this 1D MM algorithm has not been applied to DBT studies yet. In this paper, we propose a method to combine MM1D algorithm with ART+TV3D: "ART+TV3D+MM1D". Both quantitative and qualitative analyses of the proposed method ART+TV3D+MM1D, ART+TV3D, and ART are performed for a phantom that mimics 3D breast and a real 3D breast phantom with 301x236x8-dimensions.