The Gated Recurrent Conditional Generative Adversarial Network (GRC-GAN): application to denoising of low-dose CT images

Abstract

The ionizing radiation that propagates through the human body at Computed Tomography (CT) exams is known to be carcinogenic. For this reason, the development of methods for image reconstruction that operate with reduced radiation doses is essential. If we reduce the electrical current in the electrically powered X-ray tubes of CT scanners, the amount of radiation that passes through the human body during a CT exam is reduced. However, significant image noise emerges in the reconstructed CT slices if standard reconstruction methods are applied. To estimate routine-dose CT images from low-dose CT images and thus reduce noise, the Conditional Generative Adversarial Network (cGAN) was recently proposed in the literature. In this work, we introduce the Gated Recurrent Conditional Generative Adversarial Network (GRC-GAN) that is based on the usage of network gates to learn the specific regions of the input image to be updated using the cGAN denoising operation. Moreover, the GRC-GAN is executed recurrently in multiple time steps. At each time step, different parts of the input image are denoised. As a result, our GRC-GAN better focus on the denoise criterium than the regular cGAN in the LoDoPaB-CT benchmark.

The Medical Image Denoising Method Based on the CycleGAN and the Complex Shearlet Transform

“…According to [32,33], the deep learning-based methods outperform the traditional methods, such as the Wiener filtering, polynomial regression, or the wavelet denoising, so the proposed method is compared with five state-of-the-art denoising methods, i.e. the NSST-BM3D model [18], the FFDNet model [30], the FFCNN model [31], the GAN model [35] and the NSTBNet model [36]. The parameters, such as the size of the convolutional layer, the network depth, are set to be the same as they are reported in the corresponding literature.…”

“…Recently, the good denoising results have been obtained by the generative adversarial network (GAN) model for it models the denoising task to be the game between the noisy image and denoising image, which is implemented by the learning strategy on the generative and the discriminator network [34]. Furthermore, to suppress the influence of the diversity of the noise and control the sampling variables, the conditional generative adversarial networks (CGAN) is proposed for removing the noise of the low-dose CT images [35]. The deep learning-based methods outperform the other methods for their strong representation and generalization ability of the deep level features.…”