Low-Dose CT Image Denoising Using a Generative Adversarial Network With a Hybrid Loss Function for Noise Learning

Yinjin Ma,Xiaodong Guo,Biao Wei,Ge Wang,Peng He,Peng Feng

doi:10.1109/access.2020.2986388

Abstract

Potential risk of X-ray radiation from computed tomography (CT) has been a concern of the public. However, simply decreasing the dose will degrade quality of the CT images and compromise diagnostic performance. In this paper, we propose a noise learning generative adversarial network coupling with least squares, structural similarity and L1 losses for low-dose CT denoising. In our method, noise distributed in the input low-dose CT image is learned by the generator network and then subtracted from the input to generate the final denoised version. The denoised CT images are penalized by the least squares loss function, and they are pulled toward boundary of the decision even though they are classified as normal-dose CT. Least squares stabilize the training process without regularization. Structural similarity and L1 losses are utilized to keep textural details and sharpness of the denoised CT images respectively. Experiments and results show that our method can effectively suppress noise and remove artifacts compared with the state-of-the-art methods. The texture statistical properties, which include mean, standard deviation, uniformity, and entropy, further confirm that the generated noise-reduced CT image is as close as to that of the normal-dose counterpart.

Highlights

For recent decades, X-ray computed tomography (CT) is one of the most practical imaging modalities, which is extensively utilized in medical imaging, industrial evaluation, and other applications [1], [2]
DATASETS FOR EXPERIMENTS To exhibit the capacity of our adapted generative adversarial network (GAN) for low-dose CT image denoising, a real clinical CT image dataset was adopted in our study, which was authorized by Mayo Clinic for ‘‘2016 NIH-Association of Physicists in Medicine (AAPM)-Mayo Clinic Low Dose CT Grand Challenge’’
The deep learning approach needs a large number of training samples, but collecting medical images is constrained by complex formalities and extensive efforts

Summary

Introduction

X-ray computed tomography (CT) is one of the most practical imaging modalities, which is extensively utilized in medical imaging, industrial evaluation, and other applications [1], [2]. With the widespread use of medical CT, the potential risk of ionizing X-ray radiation to patients has been raised concerns of the public [3], [4]. The lower the X-ray, the noisier a CT image, which results in the degradation of the signalto-noise ratio and compromise of diagnostic performance [5]. To tackle this inherent problem, many algorithms have been developed to improve low-dose CT images [6]. These algorithms are classified into three categories: (a) sinogram filtration, (b) iterative reconstruction, and (c) post-processing

Objectives

Methods

Results

Conclusion