Aiming at the complex noise and artifacts in low-dose CT images, an improved residual encoder-decoder network for low-dose CT image denoising is proposed. The original residual encoder-decoder network consists of a series of convolution and deconvolution layers, and learns the residuals through shortcut connections. The improvement method mainly includes three aspects: firstly, batch normalization is introduced to improve the denoising performance of the network; secondly, dilated convolution is used to replace normal convolution so as to reduce the parameters in the network without degrading the denoising performance; besides, the number of feature maps in the hidden layer of the network has been adjusted to further optimize the network performance and complexity. Experimental results show that the improved network significantly reduces the complexity and also improves the denoising performance.
ZHANG Yungang,YANG Jianfeng,YI Benshun
. Improved Residual Encoder-Decoder Network for
Low-Dose CT Image Denoising[J]. Journal of Shanghai Jiaotong University, 2019
, 53(8)
: 983
-989
.
DOI: 10.16183/j.cnki.jsjtu.2019.08.014
［1］李琳, 罗德红. 低剂量CT扫描技术的临床应用 ［J］. 当代医学, 2009, 15(20): 146-149.
LI Lin, LUO Dehong. The clinical application of low-dose CT scanning ［J］. Contemporary Medicine, 2009, 15(20): 146-149.
［2］MORI I, MACHIDA Y, OSANAI M, et al. Photon starvation artifacts of X-ray CT: Their true cause and a solution ［J］. Radiological Physics & Technology, 2013, 6(1): 130-141.
［3］MANDUCA A, YU L F, TRZASKO J D, et al. Projection space denoising with bilateral filtering and CT noise modeling for dose reduction in CT ［J］. Medical Physics, 2009, 36(11): 4911-4919.
［4］KACHELRIES M, WATZKE O, KALENDER W A. Generalized multi-dimensional adaptive filtering for conventional and spiral single-slice, multi-slice, and cone-beam CT ［J］. Medical Physics, 2001, 28(4): 475-490.
［5］HSIEH J. Adaptive streak artifact reduction in computed tomography resulting from excessive x-ray photon noise ［J］. Medical Physics, 1998, 25(11): 2139-2147.
［6］WANG J, LI T F, LU H B, et al. Penalized weighted least-squares approach to sinogram noise reduction and image reconstruction for low-dose X-Ray computed tomography ［J］. IEEE Transactions on Medical Imaging, 2006, 25(10): 1272-1283.
［7］SMITH P R, PETERS T M, BATES R H T. Image reconstruction from finite numbers of projections ［J］. Journal of Physics A: Mathematical, Nuclear and General, 1973, 6(3): 361-382.
［8］董继伟. CT迭代重建技术原理及其研究进展 ［J］. 中国医学装备, 2016, 13(10): 128-133.
DONG Jiwei. The progress on research and principles of computed tomography iterative reconstruction ［J］. China Medical Equipment, 2016, 13(10): 128-133.
［9］ZENG D, HUANG J, BIAN Z Y, et al. A simple low-dose X-ray CT simulation from high-dose scan ［J］. IEEE Transactions on Nuclear Science, 2015, 62(5): 2226-2233.
［10］牛彦敏, 马燕, 王旭初. 非下采样Contourlet域中基于改进隐马尔可夫树的低剂量CT图像去噪 ［J］. 激光与光电子学进展, 2009, 46(12): 115-119.
NIU Yanmin, MA Yan, WANG Xuchu. Low-dose CT image denoising via improved hidden Markov tree model in nonsubsampled contourlet domain ［J］. Laser & Optoelectronics Progress, 2009, 46(12): 115-119.
［11］KANG D, SLOMKA P J, NAKAZATO R, et al. Image denoising of low-radiation dose coronary CT angiography by an adaptive block-matching 3D algorithm ［J］. Proceedings of SPIE. 2013, 8669: 86692G.
［12］GREEN M, MAROM E M, KIRYATI N, et al. Efficient low-dose CT denoising by locally-consistent non-local means (LC-NLM) ［C］//International Conference on Medical Image Computing and Computer-Assisted Intervention. Athens: Springer, 2016: 423-431.
［13］朱永成, 陈阳, 罗立民, 等. 基于字典学习的低剂量X-ray CT图像去噪 ［J］. 东南大学学报(自然科学版), 2012, 42(5): 864-868.
ZHU Yongcheng, CHEN Yang, LUO Limin, et al. Dictionary learning based denoising of low-dose X-ray CT image ［J］. Journal of Southeast University (Natural Science Edition), 2012, 42(5): 864-868.
［14］CHEN H, ZHANG Y, ZHANG W H, et al. Low-dose CT denoising with convolutional neural network ［C］//International Symposium on Biomedical Imaging. Melbourne: IEEE, 2017: 143-146.
［15］DONG C, LOY C C, HE K M, et al. Image super-resolution using deep convolutional networks ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(2): 295-307.
［16］CHEN H, ZHANG Y, ZHANG W H, et al. Low-dose CT via convolutional neural network ［J］. Biomedical Optics Express, 2017, 8(2): 679-694.
［17］CHEN H, ZHANG Y, KALRA M K, et al. Low-dose CT with a residual encoder-decoder convolutional neural network ［J］. IEEE Transactions on Medical Imaging, 2017, 36(12): 2524-2535.
［18］NAIR V, HINTON G E. Rectified linear units improve restricted boltzmann machines ［C］//International Conference on Machine Learning. Haifa: ACM, 2010: 807-814.
［19］HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition ［C］//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
［20］IOFFE S, SZEGEDY C. Batch normalization: Accelerating deep network training by reducing internal covariate shift ［C］//International Conference on Machine Learning. Lille: JMLR, 2015: 448-456.
［21］JIA Y Q, SHELHAMER E, DONAHUE J, et al. Caffe: convolutional architecture for fast feature embedding ［C］//ACM International Conference on Multimedia. Orlando: ACM, 2014: 675-678.
［22］KINGMA D P, BA J. Adam: A method for stochastic optimization ［C］//International Conference on Learning Representations. San Diego: Springer, 2015: 1-15.
