Generative adversarial network training method and device and image enhancement method and device

Abstract

The invention discloses a generative adversarial network training method and device, an image enhancement method and device, an electronic device and a storage medium. The training method comprises the steps that a first sample set and a second sample set are acquired, the first sample set comprises first image data and corresponding enhanced image data, and the second sample set comprises secondimage data; inputting the first image data into a generative adversarial network, and calculating a first loss error according to the enhanced image data and an output result of the generative adversarial network so as to adjust network parameters of the generative adversarial network; and inputting the second image data into a generative adversarial network, and calculating a second loss error according to an output result of the generative adversarial network so as to adjust network parameters of the generative adversarial network. According to the method, the generative adversarial networkis trained based on semi-supervised deep learning, and the accuracy and robustness of the generative adversarial network are improved while the sample data collection difficulty is reduced.

Description

technical field [0001] The invention relates to the technical field of medical imaging, in particular to a training method and device of a generative confrontation network, an image enhancement method and equipment, electronic equipment, and a storage medium. Background technique [0002] Medical image enhancement (Image Enhancement) is a class of inverse problems, including image denoising (Denoising), artifact removal (Artifact Reduction), de-blur (De-blur), image restoration (Recovery), etc. the process of. [0003] At present, image enhancement algorithms based on deep learning are generally used to achieve image enhancement. Traditional deep learning-based algorithms require a large number of image pairs composed of low-quality images and high-quality images that completely match the structural information as a training set. However, such image pairs It is difficult to obtain in practical applications, especially in the field of medical image processing. For example, f...

AI Technical Summary

Problems solved by technology

[0003] At present, image enhancement algorithms based on deep learning are generally used to achieve image enhancement. Traditional deep learning-based algorithms require a large number of image pairs composed of low-quality images and high-quality images that completely match the structural information as a training set. However, such image pairs It is difficult to obtain in practical applications, especially in the field of medical image processing. For example, for the enhancement task of low-dose CT images, it is impossible to perform low-dose scanning and normal-dose scanning on the same patient.

[0004] In view of the particularity of medical images, it is difficult to obtain sample data of medical images, which cannot meet the requirements of the diversity of sample data for network training. As a result, the accuracy of the model established by the image enhancement algorithm based on deep learning is not high. The reconstruction effect of such detailed information is not ideal, which is not conducive to the doctor's diagnosis

Images