Methods for training a tomosynthesis reconstruction model, or for generating at least one contrast tomogram depicting a target body part during an injection of contrast agent

A method, comprising implementation, by a data processor of a server, of steps of:

• (a) obtaining a pre-contrast tomogram depicting a target body part prior to an injection of contrast agent, and raw projection data acquired over a limited rotation angle span during the injection of contrast agent;
• (b) segmenting said target body part on the pre-contrast tomogram;
• (c) generating a contrast tomogram depicting the target body part during the injection of contrast agent, by application of a tomosynthesis reconstruction model to said segmented target body part and the obtained raw projection data;
• wherein said contrast tomogram and said raw projection data are respectively a i-th contrast tomogram and i-th raw projection data, with i>0, the method further comprising steps of:
• (d) obtaining i+1-th raw projection data acquired over said limited rotation angle span during the injection of contrast agent;
• and
• (e) combining the segmented target body part with at least one j-th contrast tomogram, 0<j≤i, into a combined segmented tomogram and generating a i+1-th contrast tomogram by application of the tomosynthesis reconstruction model to the combined segmented tomogram and said i+1-th raw projection data.