Computer-implemented method and data processing facility for scattered radiation correction and for parameterization of a determination algorithm for determining a measure of quality, computer program and data medium

A method for parameterization of a determination algorithm that is configured to determine a measure of quality, the method comprising:

• predetermining a number of training datasets, that each comprise a number of pairs consisting of a reference projection image and a scattered radiation distribution assigned to the respective reference projection image;
• generating a respective intermediate image by the assigned value of the assigned scattered radiation distribution being added in each case to the image values of the image points of the respective reference projection image;
• determining a proportion of scattered radiation for the respective training dataset depending on the image values of the reference projection images of the training dataset and the value of the scattered radiation distribution assigned in each case;
• and
• minimizing a cost function that depends on the respective deviation between a measure of quality determined by the determination algorithm and the proportion of scattered radiation for the number of training datasets, by variation of a number of parameters of the determination algorithm;
• wherein the determination algorithm processes as its input data a reconstructed three-dimensional image dataset or processing data that is chosen from the three-dimensional image dataset and/or is determined depending on the three-dimensional image dataset;
• wherein the three-dimensional image dataset is based on intermediate images.

A non-transitory computer implemented storage medium that stores machine-readable instructions for scattered radiation correction, the machine-readable instructions comprising:

• obtaining projection images;
• optimizing, by a determination algorithm, a measure of quality by variation of correction parameters depending on the projection images, wherein the determination algorithm is trained by machine learning and processes as input data a reconstructed three-dimensional image dataset or processing data that is selected from the reconstructed three-dimensional image dataset and/or is determined depending on the reconstructed three-dimensional image dataset, wherein the reconstructed three-dimensional image dataset is based on corrected projection images, wherein a respective corrected projection image of the corrected projection images is a result of application of a correction algorithm to a respective one of the projection images, wherein the correction algorithm is parameterized by the correction parameters or by a subgroup of the correction parameters assigned to the respective projection image;
• and
• providing radiation scatter-corrected projection images by application of the correction algorithm to the respective projection image.