Generative adversarial network for synthesizing medical images

Abstract

The invention discloses a generative adversarial network for synthesizing medical images. The topological structure of the generator is that the generator has eight levels, and the levels of the levels are logarithms, with 2 as the bottom, of the resolution of images learned by the levels; the resolution of the image of the low-level hierarchical learning is doubled and smoothly transited to the resolution of the image of the high-level hierarchical learning; the feature map of the convolution block is a feature map with self-attention. The topological structure of the discriminator is opposite to that of the generator, and the resolution of the image of high-level hierarchical learning is converted into the resolution of the image of low-level hierarchical learning in a half-smooth transition manner; the lowest level of the discriminator further comprises a batch standard deviation; the generator and the discriminator use a loss function of the WGAN-GP; the weight is initialized by using standard normal distribution, and the weight is scaled in an operation stage. Compared with a large-resolution GAN network generated by a LapGAN, a StackGAN and the like, most iteration in the training process is completed at a low resolution, and the generation speed is greatly increased on the premise that the quality of the synthesized picture is guaranteed.

Description

technical field [0001] The invention relates to the field of medical images of computer vision, in particular to a generative confrontation network for synthesizing medical images. Background technique [0002] With the improvement of computing power and the rapid increase of data volume in various industries, artificial intelligence has achieved rapid development. In 2014, Hinton's doctoral student, Lan Goodfellow, proposed the GAN network (Generative Adversarial Networks Generative Adversarial Networks), and then the improvement of GAN has been studied in full swing in the academic community, and there is still a lot of room for development. [0003] With the continuous research on GAN, various fields try to combine GAN with work in the field. Since 2017, the application of GAN in the field of medical images has risen sharply, among which image synthesis, denoising, reconstruction, segmentation, detection and classification have become the main directions of GAN in the fi...

AI Technical Summary

Problems solved by technology

On the other hand, due to the protection of patient privacy, public high-quality data sets in the medical field are very scarce, which limits the development of deep learning in the medical field.

Images