DU-GAN: Generative Adversarial Networks with Dual-Domain U-Net Based Discriminators for Low-Dose CT Denoising

Zhizhong Huang,Junping Zhang,Yi Zhang,Hongming Shan

DOI: https://doi.org/10.1109/TIM.2021.3128703

2021-11-07

Abstract:LDCT has drawn major attention in the medical imaging field due to the potential health risks of CT-associated X-ray radiation to patients. Reducing the radiation dose, however, decreases the quality of the reconstructed images, which consequently compromises the diagnostic performance. Various deep learning techniques have been introduced to improve the image quality of LDCT images through denoising. GANs-based denoising methods usually leverage an additional classification network, i.e. discriminator, to learn the most discriminate difference between the denoised and normal-dose images and, hence, regularize the denoising model accordingly; it often focuses either on the global structure or local details. To better regularize the LDCT denoising model, this paper proposes a novel method, termed DU-GAN, which leverages U-Net based discriminators in the GANs framework to learn both global and local difference between the denoised and normal-dose images in both image and gradient domains. The merit of such a U-Net based discriminator is that it can not only provide the per-pixel feedback to the denoising network through the outputs of the U-Net but also focus on the global structure in a semantic level through the middle layer of the U-Net. In addition to the adversarial training in the image domain, we also apply another U-Net based discriminator in the image gradient domain to alleviate the artifacts caused by photon starvation and enhance the edge of the denoised CT images. Furthermore, the CutMix technique enables the per-pixel outputs of the U-Net based discriminator to provide radiologists with a confidence map to visualize the uncertainty of the denoised results, facilitating the LDCT-based screening and diagnosis. Extensive experiments on the simulated and real-world datasets demonstrate superior performance over recently published methods both qualitatively and quantitatively.

Image and Video Processing,Computer Vision and Pattern Recognition,Medical Physics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to improve image quality while reducing radiation dose in low - dose CT (LDCT) imaging, especially by using denoising techniques to reduce image noise and artifacts caused by the reduction of radiation dose. Low - dose CT imaging has become a focus of attention in the field of medical imaging due to its potential health risks, such as the X - ray radiation associated with CT that may harm patients. However, reducing the radiation dose will degrade the quality of the reconstructed image, thereby affecting the diagnostic performance. Therefore, the paper proposes a new method based on generative adversarial networks (GANs) - DU - GAN, which aims to improve the quality of low - dose CT images by introducing a discriminator with a dual - domain U - Net structure to simultaneously learn the global and local differences between the denoised image and the normal - dose CT image in the image domain and the gradient domain. This method can not only provide pixel - level feedback, but also pay attention to the global structure of the image, so as to better regulate the denoising model, reduce the streak artifacts caused by photon starvation, and enhance the edges of the denoised CT image. In addition, by introducing the CutMix technique, DU - GAN can also provide radiologists with a confidence map to visualize the uncertainty of the denoising results, which is helpful for screening and diagnosis using the denoised low - dose CT images.