Unconstrained Simultaneous Scheme to Fully Couple Reconstruction and Registration for Digital Breast Tomosynthesis: A Feasible Study

Digital breast tomosynthesis (DBT) provides a pseudo-3D reconstruction which addresses the limitation of superimposition of dense fibro-glandular tissue associated with conventional mammography. Registration of temporal DBT volumes searches for the optimum deformation to transform two observed images of the same object into a common reference frame. This aligns the two images via minimising an objective function that calculates the similarity between the two datasets. In this paper, we present a novel algorithm which combines reconstruction of a pair of temporal DBT acquisitions with their simultaneous registration. We approach this nonlinear inverse problem using a generic unconstrained optimisation scheme. To evaluate the performance of our method we use 2D and 3D software phantoms and demonstrate that this simultaneous approach has comparable results to performing these tasks sequentially or iteratively w.r.t both the reconstruction fidelity and the registration accuracy.