Learning device, image generation device, learning method, image generation method, learning program, and image generation program

Abstract

A learning device detects a first region of interest including a calcification and a second region of interest including an other lesion on the basis of any one of a composite two-dimensional image obtained from a plurality of projection images captured by tomosynthesis imaging or a plurality of tomographic images, the tomographic image, or a normal two-dimensional image, and trains an image generation model, in which a weight for the first region of interest is largest and a weight for the second region of interest is set to be equal to or larger than a weight for a region other than the first region of interest and the second region of interest, by updating a weight for a network of the image generation model on the basis of a loss between a pseudo two-dimensional image output by the image generation model and the normal two-dimensional image and/or the composite two-dimensional image to reduce the loss.