Attenuation coefficient image generation method, nuclear medicine diagnostic apparatus, and trained model generation method

A nuclear medicine diagnostic apparatus comprising:

• a detector configured to detect radiation generated from a radiopharmaceutical agent in a subject;
• and
• a processor configured to generate a radioactivity distribution image of the subject based on detection of the radiation by the detector,
• wherein the processor is configured to
• generate an input image by performing imaging processing on measurement data acquired based on detection of the radiation emitted from the subject,
• generate a plurality of tissue composition ratio images indicating ratios of a plurality of tissues included in each pixel or a plurality of certainty degree images indicating probability of whether each pixel belongs to one of a plurality of tissues, based on the input image, and
• generate an attenuation coefficient image for generating the radioactivity distribution image, based on weighting addition by multiplying each of the plurality of tissue composition ratio images or the certainty degree images by known attenuation coefficients of the plurality of tissues and then adding up each of the images multiplied by the known attenuation coefficients.

A trained model generation method for a nuclear medicine diagnostic apparatus, the method comprising the steps of:

• preparing tissue label images indicating a tissue to which each pixel belongs;
• generating pseudo-radioactivity distribution images generated by assigning radioactive concentration to each tissue of the tissue label images, corresponding to a radioactivity distribution image generated by the nuclear medicine diagnostic apparatus and pseudo-attenuation coefficient images generated by assigning an attenuation coefficient to each integrated tissue of the tissue label images, based on the tissue label images;
• generating pseudo-measurement data corresponding to measurement data generated by the nuclear medicine diagnostic apparatus by performing simulation calculations, based on the pseudo-radioactivity distribution images and the pseudo-attenuation coefficient images;
• generating pseudo images by performing imaging processing on the pseudo-measurement data, the pseudo images corresponding to an input image generated by the nuclear medicine diagnostic apparatus by performing imaging processing on the measurement data;
• and
• generating a trained model using the pseudo images based on the tissue label images as training input images and a plurality of tissue composition ratio images indicating ratios of a plurality of tissues included in each pixel based on the tissue label image as training output images.