一種基于深度學習的自適應醫(yī)學超聲圖像去斑方法

付曉薇; 楊雪飛; 陳芳; 李曦

doi:10.11999/JEIT190580

一種基于深度學習的自適應醫(yī)學超聲圖像去斑方法

doi: 10.11999/JEIT190580 cstr: 32379.14.JEIT190580

付曉薇^{1, 2, 3, ,},
楊雪飛^{1, 2},
陳芳⁴,
李曦⁵

1.
武漢科技大學計算機科學與技術學院武漢 430065
2.
智能信息處理與實時工業(yè)系統(tǒng)湖北省重點實驗室武漢 430065
3.
華中科技大學材料成形與模具技術國家重點實驗室武漢 430074
4.
武漢科技大學校醫(yī)院超聲影像科武漢 430065
5.
華中科技大學人工智能與自動化學院武漢 430074

基金項目: 國家自然科學基金(61873323)，材料成形與模具技術國家重點實驗室開放課題研究基金(P2018-016)，湖北省自然科學基金(2017CFB506)，智能信息處理與實時工業(yè)系統(tǒng)湖北省重點實驗室開放課題項目(2016znss02A, znxx2018ZD01)，大學生科技創(chuàng)新基金項目(18ZRA076)

詳細信息

作者簡介:
付曉薇：女，1977年生，教授，研究方向為圖像處理、計算機視覺、信號處理與分析

楊雪飛：女，1994年生，碩士生，研究方向為圖像處理、深度學習

陳芳：女，1972年生，研究方向為肌骨超聲圖像的調節(jié)

李曦：男，1977年生，教授，研究方向為計算機應用，復雜非線性系統(tǒng)的建模和控制

通訊作者:
付曉薇　fxw_wh0409@wust.edu.cn

中圖分類號: TN911.73
計量
- 文章訪問數(shù): 3777
- HTML全文瀏覽量: 1586
- PDF下載量: 169
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2019-07-31
- 修回日期: 2020-03-18
- 網絡出版日期: 2020-04-11
- 刊出日期: 2020-07-23

An Adaptive Medical Ultrasound Images Despeckling Method Based on Deep Learning

Xiaowei FU^{1, 2, 3
, ,},
Xuefei YANG^{1, 2},
Fang CHEN⁴,
Xi LI⁵

1.
College of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan 430065, China
2.
Key Laboratory of Intelligent Information Processing and Real-time Industrial System, Wuhan 430065, China
3.
State Key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and Technology, Wuhan 430074, China
4.
Department of Ultrasound and Imaging, Wuhan University of Science and Technology Hospital, Wuhan 430065, China
5.
School of Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan 430074, China

Funds: The National Natural Science Foundation of China (61873323), The Open Fund Project of State Key Laboratory of Material Processing and Die & Mould Technology (P2018-016), The Natural Science Foundation of Hubei Provincial (2017CFB506), The Open Fund Project of Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System (2016znss02A, znxx2018ZD01), The University Student Science and Technology Innovation Fund Project (18ZRA076)

摘要

摘要:
針對傳統(tǒng)醫(yī)學超聲圖像去斑方法的不足，該文提出一種自適應多曝光融合框架和前饋卷積神經網絡模型圖像去斑方法。首先，制作超聲圖像訓練數(shù)據集；然后，提出一種自適應增強因子的多曝光融合框架，增強圖像進行有效特征提??；最后，通過網絡訓練去斑模型并獲得去斑后的圖像。實驗結果表明，該文較已有的方法，能更有效地濾除醫(yī)學超聲圖像中的斑點噪聲并更多的保留圖像細節(jié)。
- 超聲圖像 /
- 深度學習 /
- 多曝光融合框架 /
- 乘性斑點噪聲
Abstract:
Considering the shortage of traditional medical ultrasound image despeckle methods, an adaptive multi-exposure fusion framework and feedforward convolutional neural network model image despeckle method is proposed. Firstly, an ultrasound image training data set is produced. Then, a multi-exposure fusion framework with adaptive enhancement factors is proposed to enhance the image for effective feature extraction.Finally, a speckle model is trained through the network and a speckle image is obtained. Experimental results show that, compared with the existing methods, this paper can more effectively remove speckle noise in medical ultrasound images and retain more image details.
- Ultrasound image /
- Deep learning /
- Multi-exposure fusion framework /
- Multiplicative speckle noise

HTML全文

圖 1 殘差網絡的基本結構

下載: 全尺寸圖片幻燈片

圖 2 本文流程圖

下載: 全尺寸圖片幻燈片

圖 3 網絡結構圖

下載: 全尺寸圖片幻燈片

圖 4 模擬斑點肝臟超聲圖像1實驗比較

下載: 全尺寸圖片幻燈片

圖 5 模擬斑點肝臟超聲圖像2實驗比較

下載: 全尺寸圖片幻燈片

圖 6 真實醫(yī)學超聲圖像去斑后結果比較

下載: 全尺寸圖片幻燈片

表 1 模擬斑點肝臟超聲圖像1不同方法PSNR結果(dB)

方法	斑點噪聲的標準差σ
方法	0.5	0.6	0.7	0.8	0.9
BI-DTCWT	35.4225	33.8461	32.5368	31.4265	30.3688
NPSM	34.5827	32.9573	31.5366	30.3250	29.2056
NL-means	34.9289	34.1934	33.3554	32.5658	31.6134
BM3D	36.1701	35.7552	35.2485	34.8951	34.2035
Local_entropy_qsp	36.7812	36.1083	35.4363	35.0726	34.5014
DnCNN	35.7701	35.8394	35.8180	35.6769	35.3885
本文方法	36.7203	36.7139	36.6025	36.3568	35.9492

下載: 導出CSV

表 4 模擬斑點肝臟超聲圖像2不同方法$\beta $結果

方法	斑點噪聲的標準差σ
方法	0.5	0.6	0.7	0.8	0.9
BI-DTCWT	0.7078	0.6359	0.5612	0.5099	0.4661
NPSM	0.6830	0.6197	0.5479	0.4990	0.4517
NL-means	0.7191	0.6761	0.6037	0.5449	0.4899
BM3D	0.8030	0.7950	0.7826	0.7683	0.7355
Local_entropy_qsp	0.8263	0.8090	0.7787	0.7567	0.7384
DnCNN	0.9286	0.9238	0.9156	0.9029	0.8812
本文方法	0.9394	0.9325	0.9217	0.9653	0.8836

下載: 導出CSV

表 2 模擬斑點肝臟超聲圖像2不同方法PSNR結果(dB)

方法	斑點噪聲的標準差σ
方法	0.5	0.6	0.7	0.8	0.9
BI-DTCWT	31.0477	29.5409	28.0856	27.4342	26.2056
NPSM	31.5374	30.0985	28.6745	27.6699	26.6843
NL-means	32.7360	31.7539	30.4860	29.5105	28.4174
BM3D	33.8786	33.3096	32.5436	32.0199	31.2079
Local_entropy_qsp	34.3157	33.2426	32.1706	31.5329	30.8599
DnCNN	34.9760	35.0382	34.8497	34.3851	33.6562
本文方法	35.9280	35.9170	35.6289	35.0301	34.1677

下載: 導出CSV

表 3 模擬斑點肝臟超聲圖像1不同方法$\beta $結果

方法	斑點噪聲的標準差σ
方法	0.5	0.6	0.7	0.8	0.9
BI-DTCWT	0.6416	0.5611	0.4823	0.4291	0.3846
NPSM	0.5972	0.5154	0.4352	0.3817	0.3393
NL-means	0.4522	0.4102	0.3564	0.3262	0.2949
BM3D	0.5969	0.5820	0.5685	0.5477	0.5016
Local_entropy_qsp	0.6540	0.6287	0.5991	0.5842	0.5621
DnCNN	0.7803	0.7726	0.7595	0.7393	0.7106
本文方法	0.8128	0.8011	0.7831	0.7564	0.7208

下載: 導出CSV

表 5 真實斑點超聲圖像不同方法ENL結果

方法	ENL等效視數(shù)值
BI-DTCWT	61.2209
NPSM	64.6016
NL-means	109.5584
BM3D	93.4877
Local_entropy_qsp	79.1016
DnCNN	132.9184
本文方法	134.3287

下載: 導出CSV

表 6 50張真實斑點超聲圖像不同方法ENL平均值比較

方法	ENL等效視數(shù)值平均值
BI-DTCWT	75.5182
NPSM	75.5941
NL-means	110.6393
BM3D	110.9127
Local_entropy_qsp	93.7911
DnCNN	140.3622
本文方法	147.0689

下載: 導出CSV

參考文獻(28)

KANAYAMA Y and YANO M. Ultrasound diagnosis apparatus and ultrasound imaging method[P]. USA Patent, 10231710, 2019.

ZHOU Yingyue, ZANG Hongbin, XU Su, et al. An iterative speckle filtering algorithm for ultrasound images based on bayesian nonlocal means filter model[J]. Biomedical Signal Processing and Control, 2019, 48: 104–117. doi: 10.1016/j.bspc.2018.09.011

JOEL T and SIVAKUMAR R. An extensive review on Despeckling of medical ultrasound images using various transformation techniques[J]. Applied Acoustics, 2018, 138: 18–27. doi: 10.1016/j.apacoust.2018.03.023

沈民奮, 陳婷婷, 張瓊, 等. 醫(yī)用超聲圖像散斑去噪方法綜述[J]. 中國醫(yī)療器械信息, 2013, 19(3): 17–22. doi: 10.3969/j.issn.1006-6586.2013.03.003

SHEN Minfen, CHEN Tingting, ZHANG Qiong, et al. The review of speckle denoising in medical ultrasound imaging[J]. China Medical Device Information, 2013, 19(3): 17–22. doi: 10.3969/j.issn.1006-6586.2013.03.003

KUAN D, SAWCHUK A, STRAND T, et al. Adaptive restoration of images with speckle[J]. IEEE Transactions on Acoustics, Speech, and Signal Processing, 1987, 35(3): 373–383. doi: 10.1109/TASSP.1987.1165131

DABOV K, FOI A, KATKOVNIK V, et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. IEEE Transactions on Image Processing, 2007, 16(8): 2080–2095. doi: 10.1109/TIP.2007.901238

肖佳, 張俊華, 梅禮曄. 改進的三維塊匹配去噪算法[J]. 計算機科學, 2019, 46(6): 288–294. doi: 10.11896/j.issn.1002-137X.2019.06.043

XIAO Jia, ZHANG Junhua, and MEI Liye. Improved block-matching 3D denoising algorithm[J]. Computer Science, 2019, 46(6): 288–294. doi: 10.11896/j.issn.1002-137X.2019.06.043

PERONA P and MALIK J. Scale-space and edge detection using anisotropic diffusion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1990, 12(7): 629–639. doi: 10.1109/34.56205

YU Yongjian and ACTON S T. Speckle reducing anisotropic diffusion[J]. IEEE Transactions on Image Processing, 2002, 11(11): 1260–1270. doi: 10.1109/TIP.2002.804276

TIAN Jing and CHEN Li. Image despeckling using a non-parametric statistical model of wavelet coefficients[J]. Biomedical Signal Processing and Control, 2011, 6(4): 432–437. doi: 10.1016/j.bspc.2010.11.006

SENDUR L and SELESNICK I W. Bivariate shrinkage functions for wavelet-based denoising exploiting interscale dependency[J]. IEEE Transactions on Signal Processing, 2002, 50(11): 2744–2756. doi: 10.1109/TSP.2002.804091

BURGER H C, SCHULER C J, and HARMELING S. Image denoising: Can plain neural networks compete with BM3D?[C]. 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, USA, 2012: 2392–2399.

SCHMIDT U and ROTH S. Shrinkage fields for effective image restoration[C]. The IEEE Conference on Computer Vision and Pattern Recognition, Columbus, USA, 2014: 2774–2781.

CHEN Yunjin and POCK T. Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restoration[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1256–1272. doi: 10.1109/TPAMI.2016.2596743

MAO Xiaojiao, SHEN Chunhua, and YANG Yubin. Image restoration using very deep convolutional encoder-decoder networks with symmetric skip connections[C]. Advances in Neural Information Processing Systems, Red Hook, USA, 2016: 2802–2810.

ZHANG Kai, ZUO Wangmeng, CHEN Yunjin, et al. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing, 2017, 26(7): 3142–3155. doi: 10.1109/TIP.2017.2662206

NADEEM M, HUSSAIN A, and MUNIR A. Fuzzy logic based computational model for speckle noise removal in ultrasound images[J]. Multimedia Tools and Applications, 2019, 78(9): 18531–18548.

呂曉琪, 吳涼, 谷宇, 等. 基于深度卷積神經網絡的低劑量CT肺部去噪[J]. 電子與信息學報, 2018, 40(6): 1353–1359. doi: 10.11999/JEIT170769

Lü Xiaoqi, WU Liang, GU Yu, et al. Low dose CT lung denoising model based on deep convolution neural network[J]. Journal of Electronics &Information Technology, 2018, 40(6): 1353–1359. doi: 10.11999/JEIT170769

YING Zhenqiang, LI Ge, REN Yurui, et al. A new image contrast enhancement algorithm using exposure fusion framework[C]. The 17th International Conference on Computer Analysis of Images and Patterns, Ystad, Sweden, 2017: 36–46.

ZHENG D, WANG J, and XIAO Z. Image enhancement based on local standard deviation[J]. Journal of Information and Computational Science, 2005, 2(2): 429–437.

JIFARA W, JIANG Feng, RHO S, et al. Medical image denoising using convolutional neural network: A residual learning approach[J]. The Journal of Supercomputing, 2019, 75(2): 704–718. doi: 10.1007/s11227-017-2080-0

SIMONYAN K and ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. arXiv, 2014, 1409.1556.

VEDALDI A and LENC K. MatconvNet: Convolutional neural networks for MATLAB[C]. The 23rd ACM International Conference on Multimedia, Brisbane, Australia, 2015: 689–692.

FU Xiaowei, WANG Yi, CHEN Li, et al. An image despeckling approach using quantum-inspired statistics in dual-tree complex wavelet domain[J]. Biomedical Signal Processing and Control, 2015, 18: 30–35. doi: 10.1016/j.bspc.2014.11.005

BUADES A, COLL B, and MOREL J M. A non-local algorithm for image denoising[C]. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, USA, 2005: 60–65.

付曉薇, 代蕓, 陳黎, 等. 基于局部熵的量子衍生醫(yī)學超聲圖像去斑[J]. 電子與信息學報, 2015, 37(3): 560–566. doi: 10.11999/JEIT140587

FU Xiaowei, DAI Wei, CHEN Li, et al. Quantum-inspired despeckling of medical ultrasound images based on local entropy[J]. Journal of Electronics &Information Technology, 2015, 37(3): 560–566. doi: 10.11999/JEIT140587

PIZURICA A, PHILIPS W, LEMAHIEU I, et al. A versatile wavelet domain noise filtration technique for medical imaging[J]. IEEE Transactions on Medical Imaging, 2003, 22(3): 323–331. doi: 10.1109/TMI.2003.809588

RODTOOK A and MAKHANOV S S. Multi-feature gradient vector flow snakes for adaptive segmentation of the ultrasound images of breast cancer[J]. Journal of Visual Communication and Image Representation, 2013, 24(8): 1414–1430. doi: 10.1016/j.jvcir.2013.09.009

施引文獻

資源附件(0)

訪問統(tǒng)計