Matrix inversion tomosynthesis improvements in longitudinal x-ray slice

Abstract

The present invention describes improvements to digital tomosynethesis, whereby a plurality of x-ray projection images are shifted and summed to form a matrix of tomographs at variable plane heights. Prior to tomosynthesis, the projection images are weighted by a weighting function which falls off towards shallower angles of the x-ray tube during acquisition, in order to improve the reconstructability of between-plane structures. A matrix of blurring functions is then determined, describing the blurring of structures in all planes for every tomosynthesis reconstruction. Actual patient structures are then reconstructed by solving a matrix equation containing the matrix of tomosynthesized images and the matrix of blurring functions. The patient structure images are then improved by a hybrid correction routine which combines low spatial frequencies from a wide swing of the x-ray source with high and mid spatial frequencies from a narrow swing of the x-ray source. The data, once hybrid corrected, is further corrected using an interative technique to improve the reconstruction of low spatial frequencies. This iterative technique corrects each column of each image by sequentially determining if errant frequency components are present at the edge of the images where the patient is not present. If errant frequency components are found, they are subtracted from the entire image column. Lastly, a DC level correction is performed which corrects for errant zero frequency components by adding a constant value to each image column to force the edge of each image to zero. The data is finally displayed on a display monitor. An interleaving technique is used to vary reconstructed plane heights in order to permit improved interplane resolution.