US 11,690,579 B2
Attention-driven image domain translation
Srikrishna Karanam, Cambridge, MA (US); Ziyan Wu, Cambridge, MA (US); and Terrence Chen, Cambridge, MA (US)
Assigned to Shanghai United Imaging Intelligence Co., LTD., Shanghai (CN)
Filed by Shanghai United Imaging Intelligence Co., LTD., Shanghai (CN)
Filed on Jun. 16, 2020, as Appl. No. 16/902,760.
Prior Publication US 2021/0386391 A1, Dec. 16, 2021
Int. Cl. A61B 6/00 (2006.01); A61B 6/03 (2006.01); G06N 3/084 (2023.01); G06F 18/22 (2023.01); G06F 18/241 (2023.01); G06N 3/045 (2023.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G06V 40/16 (2022.01)
CPC A61B 6/032 (2013.01) [A61B 6/5211 (2013.01); A61B 6/54 (2013.01); G06F 18/22 (2023.01); G06F 18/241 (2023.01); G06N 3/045 (2023.01); G06N 3/084 (2013.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G06V 40/171 (2022.01)] 17 Claims
OG exemplary drawing
 
1. An apparatus comprising a processor, the processor configured to:
receive input image data corresponding to output image data of a first radiology scanner device;
translate the input image data into a format corresponding to output image data of a second radiology scanner device by:
in a first training phase, training a discriminator of a Generative Adversarial Network (GAN) to classify between images of the first radiology scanner device and images of the second radiology scanner device;
generating a target attention image map using a prediction of the discriminator trained in the first phase; and
in a second training phase, training the discriminator with an input image corresponding to a real input image of the first radiology scanner device, a synthesized image generated by the generator of the GAN, and a set of images corresponding to real images of the second radiology scanner device;
determining a degree of similarity between the synthesized image and the set of real images;
determining a degree of dissimilarity between the real input image and the set of real images;
generating an attention map based on the determined similarity between the synthesized image and the set of real images and the determined dissimilarity between the real input image and the set of real images;
determining a degree of similarity between the target attention image map generated in the first training phase and the attention map generated in the second training phase; and
training the GAN using the determined degree of similarity; and
generate an output image corresponding to the translated input image data on a post processing imaging device associated with the first radiology scanner device.