Clinical Implementation of MITS Reconstruction for Linear Tomosynthesis

Tomosynthesis is an X-ray based tomographic modality, which calculates slice images of the examined volume from a set of low dose projection images acquired over a limited scan angle. Due to the limited scan angle, the reconstruction is an extremely ill-posed inverse problem. Therefore, many artefacts degrade the calculated slice images in the case of any reconstruction method. If linear, shift invariant system based algorithm reconstructs these slices, they can be interpreted by utilizing linear systems theory. MITS [1] is one of the most adequate of these methods; although its clinical implementation is not trivial. This paper examines the main difficulties of its clinical implementation from both theoretical and practical viewpoints, proposes modifications to treat these and compensate the possible artifacts (i.e. the intensity oscillation artifact, the partial volume artifact and the effects caused by the low condition of the inverse problem). The proposed modifications preserve the linear, shift invariant nature of the reconstruction, therefore, the reconstruction after applying these modifications, can also be interpreted by analyzing its Transfer Function. The paper also describes the method of this interpretation. A clinical implementation of a linear tomosynthesis system is used for validating the proposed, modified MITS algorithm. The paper concludes that, based on this validation, the proposed modifications significantly decrease the artifacts of the MITS algorithm in practical utilization and improve the quality of the reconstructed slice images, while the whole reconstruction remains interpretable (which mean, that we exactly know how any voxel of a reconstruction depend on the internal structure of the examined volume).