SACNN: Self-Attention Convolutional Neural Network for Low-Dose CT Denoising With Self-Supervised Perceptual Loss Network

被引:238
作者
Li, Meng [1 ]
Hsu, William [2 ]
Xie, Xiaodong [1 ]
Cong, Jason [3 ]
Gao, Wen [1 ]
机构
[1] Peking Univ, Dept Elect Engn & Comp Sci, Beijing 100871, Peoples R China
[2] Univ Calif Los Angeles, David Geffen Sch Med, Dept Radiol Sci, Los Angeles, CA 90024 USA
[3] Univ Calif Los Angeles, Dept Comp Sci, Los Angeles, CA 90095 USA
基金
中国国家自然科学基金;
关键词
Low-dose CT; denoising; self-attention; autoencoder; perceptual loss; COMPUTED-TOMOGRAPHY; IMAGE-RECONSTRUCTION; NOISE-REDUCTION; ALGORITHM;
D O I
10.1109/TMI.2020.2968472
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
Computed tomography (CT) is a widely used screening and diagnostic tool that allows clinicians to obtain a high-resolution, volumetric image of internal structures in a non-invasive manner. Increasingly, efforts have been made to improve the image quality of low-dose CT (LDCT) to reduce the cumulative radiation exposure of patients undergoing routine screening exams. The resurgence of deep learning has yielded a new approach for noise reduction by training a deep multi-layer convolutional neural networks (CNN) to map the low-dose to normal-dose CT images. However, CNN-based methods heavily rely on convolutional kernels, which use fixed-sizefilters to process one local neighborhood within the receptive field at a time. As a result, they are not efficient at retrieving structural information across large regions. In this paper, we propose a novel 3D self-attention convolutional neural network for the LDCT denoising problem. Our 3D self-attention module leverages the 3D volume of CT images to capture a wide range of spatial information both within CT slices and between CT slices. With the help of the 3D self-attention module, CNNs are able to leverage pixels with stronger relationships regardless of their distance and achieve better denoising results. In addition, we propose a self-supervised learning scheme to train a domain-specific autoencoder as the perceptual loss function. We combine these two methods and demonstrate their effectiveness on both CNN-based neural networks and WGAN-based neural networks with comprehensive experiments. Tested on the AAPM-Mayo Clinic Low Dose CT Grand Challenge data set, our experiments demonstrate that self-attention (SA) module and autoencoder (AE) perceptual loss function can efficiently enhance traditional CNNs and can achieve comparable or better results than the state-of-the-art methods.
引用
收藏
页码:2289 / 2301
页数:13
相关论文
共 61 条
[1]   Learned Primal-Dual Reconstruction [J].
Adler, Jonas ;
Oktem, Ozan .
IEEE TRANSACTIONS ON MEDICAL IMAGING, 2018, 37 (06) :1322-1332
[2]   K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation [J].
Aharon, Michal ;
Elad, Michael ;
Bruckstein, Alfred .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2006, 54 (11) :4311-4322
[3]  
[Anonymous], 2016, P 21 C EMPIRICAL MET
[4]  
[Anonymous], 2017, arXiv:1705.04267
[5]  
Arjovsky M., 2017, ARXIV170107875
[6]  
Arjovsky M, 2017, PR MACH LEARN RES, V70
[7]   Ray Contribution Masks for Structure Adaptive Sinogram Filtering [J].
Balda, Michael ;
Hornegger, Joachim ;
Heismann, Bjoern .
IEEE TRANSACTIONS ON MEDICAL IMAGING, 2012, 31 (06) :1228-1239
[8]   PatchMatch: A Randomized Correspondence Algorithm for Structural Image Editing [J].
Barnes, Connelly ;
Shechtman, Eli ;
Finkelstein, Adam ;
Goldman, Dan B. .
ACM TRANSACTIONS ON GRAPHICS, 2009, 28 (03)
[9]   Tracking Dosimetric Changes Due to Lung Patient Physical Changes During Proton Therapy Treatment [J].
Bennett, M. ;
Hoppe, B. ;
Li, Z. ;
Flampouri, S. .
MEDICAL PHYSICS, 2013, 40 (06)
[10]   A non-local algorithm for image denoising [J].
Buades, A ;
Coll, B ;
Morel, JM .
2005 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, VOL 2, PROCEEDINGS, 2005, :60-65