An aspect of the present invention relates to a method for transforming generative medical images. More particularly, the invention relates to a medical image transformation method and an apparatus therefor, in which a learning model trained using a Generative Adversarial Network (GAN) performs learning at the level of feature maps generated during the encoding process of medical images, thereby outputting medical images with improved sharpness.
According to one embodiment of the present invention, the feature map encoded from the input image is compared with the feature map encoded from the target image, and the model is trained in a direction that reduces the difference between them. As a result, the difference between images is reduced at the feature map level, thereby providing a medical image transformation apparatus capable of producing output results with further improved sharpness.
Application / Registration Date
2023.08.31
Application / Registration Number
10-2574761
Country
Republic of Korea/PCT
Status
Granted
Summary
An aspect of the present invention relates to a method for transforming generative medical images. More particularly, the invention relates to a medical image transformation method and an apparatus therefor, in which a learning model trained using a Generative Adversarial Network (GAN) performs learning at the level of feature maps generated during the encoding process of medical images, thereby outputting medical images with improved sharpness.
According to one embodiment of the present invention, the feature map encoded from the input image is compared with the feature map encoded from the target image, and the model is trained in a direction that reduces the difference between them. As a result, the difference between images is reduced at the feature map level, thereby providing a medical image transformation apparatus capable of producing output results with further improved sharpness.
Link
https://doi.org/10.8080/1020220158817