基于改进U-Net网络的甲状腺结节超声图像分割方法

王波; 李梦翔; 刘侠

doi:10.11999/JEIT210015

基于改进U-Net网络的甲状腺结节超声图像分割方法

doi: 10.11999/JEIT210015

王波^{1, 2,},
李梦翔¹,
刘侠^{1, 2, ,}

1.
哈尔滨理工大学自动化学院哈尔滨 150080
2.
黑龙江省复杂智能系统与集成重点实验室哈尔滨 150080

基金项目: 国家自然科学基金(61172167)，哈尔滨理工大学“理工英才”计划科学研究项目(LGYC2018JC013)，黑龙江省青年科学基金项目(QC2017076)

详细信息

作者简介:
王波：男，1982年生，博士，副教授，研究方向为医学图像处理、模式识别、机器学习

李梦翔：男，1996年生，硕士生，研究方向为医学图像分割、机器学习

刘侠：男，1975年生，博士，教授，研究方向为医学图像处理、模式识别、机器学习

通讯作者:
刘侠　liuxia@hrbust.edu.cn

中图分类号: TN911.73; TP391
计量
- 文章访问数: 3360
- HTML全文浏览量: 1483
- PDF下载量: 402
- 被引次数: 0
出版历程
- 收稿日期: 2021-01-05
- 修回日期: 2021-03-31
- 网络出版日期: 2021-04-16
- 刊出日期: 2022-02-25

Ultrasound Image Segmentation Method of Thyroid Nodules Based on the Improved U-Net Network

WANG Bo^{1, 2
,},
LI Mengxiang¹,
LIU Xia^{1, 2
, ,}

1.
School of Automation, Harbin University of Science and Technology, Harbin 150080, China
2.
Heilongjiang Provincial Key Laboratory of Complex Intelligent System and Integration, Harbin 150080, China

Funds: The National Natural Science Foundation of China (61172167), The Scientific Research Project of Talent Plan of Harbin University of Science and Technology (LGYC2018JC013), The Youth Science Foundation of Heilongjiang Province (QC2017076)

摘要

摘要: 针对甲状腺结节尺寸多变、超声图像中甲状腺结节边缘模糊导致难以分割的问题，该文提出一种基于改进U-net网络的甲状腺结节超声图像分割方法。该方法首先将图片经过有残差结构和多尺度卷积结构的编码器路径进行降尺度特征提取；然后，利用带有注意力模块的跳跃长连接部分对特征张量进行边缘轮廓保持操作；最后，使用带有残差结构和多尺度卷积结构的解码器路径得到分割结果。实验结果表明，该文所提方法的平均分割Dice值达到0.7822，较传统U-Net方法具有更优的分割性能。
- 图像分割 /
- 甲状腺结节超声图像 /
- 注意力机制 /
- U-Net
Abstract: An ultrasound image segmentation method of thyroid nodules based on the improved u-net network is proposed in this paper, in order to solve the problem of changeable size of thyroid nodules and difficulty in segmentation due to edge blur of thyroid nodules in the ultrasound image. Firstly, the image is downscaled to extract the features through an encoder path with a residual structure and a multi-scale convolution structure. Secondly, the long skip connection with an attention module is used to maintain the edge contour of characteristic tensor. Finally, the segmentation result is obtained by a decoder path with a residual structure and a multi-scale convolution structure. The experimental results show that with the method proposed in this paper, the average segmentation Dice value reaches 0.7822. It indicates that this method has better segmentation performance than the traditional U-Net method.
- Image segmentation /
- Ultrasound image of thyroid nodule /
- Attention mechanism /
- U-Net

HTML全文

图 1 基于残差多尺度卷积和注意力机制的深度卷积神经网络模型

下载: 全尺寸图片幻灯片

图 2 本文中所使用的残差块

下载: 全尺寸图片幻灯片

图 3 多尺度卷积结构

下载: 全尺寸图片幻灯片

图 4 注意力模块网络结构

下载: 全尺寸图片幻灯片

图 5 不同网络分割结果对比

下载: 全尺寸图片幻灯片

图 6 空间注意力权重图可视化后的结果

下载: 全尺寸图片幻灯片

图 7 通道注意力模块输出特征张量的不同通道可视化后的结果

下载: 全尺寸图片幻灯片

表 1 不同模型的量化分割结果

方法	${\rm{DSC}}$	${\rm{IoU}}$	${\rm{Hausdorff}}$	${\rm{FPR}}$	${\rm{FNR}}$
U-Net	0.6254±0.0719	0.4550±0.0709	35.1372±4.2380	0.0703±0.0216	0.1857±0.0532
A-Deeplabv3+	0.6874±0.0241	0.5258±0.0280	26.7047±2.1334	0.0553±0.0123	0.2364±0.0244
BCDU-Net	0.7132±0.0086	0.5554±0.0104	22.8483±2.9490	0.0540±0.0245	0.2052±0.0416
U-Net++	0.7177±0.0307	0.5597±0.0214	27.2110±2.3588	0.0407±0.0037	0.2556±0.0785
AU-Net	0.7413±0.0451	0.5889±0.0581	25.2377±1.1363	0.0426±0.0023	0.3024±0.0785
AU-Net+SE	0.7432±0.0231	0.5934±0.0289	24.3763±2.0682	0.0366±0.0022	0.2720±0.0301
本文方法	0.7822±0.0113	0.6423±0.0151	19.2769±1.2694	0.0306±0.0064	0.1991±0.0197

下载: 导出CSV

参考文献(24)

[1]	HAUGEN B R, ALEXANDER E K, BIBLE K C, et al. 2015 American Thyroid Association management guidelines for adult patients with thyroid nodules and differentiated thyroid cancer: The American Thyroid Association guidelines task force on thyroid nodules and differentiated thyroid cancer[J]. Thyroid, 2016, 26(1): 1–133. doi: 10.1089/thy.2015.0020
[2]	MAROULIS D E, SAVELONAS M A, IAKOVIDIS D K, et al. Variable background active contour model for computer-aided delineation of nodules in thyroid ultrasound images[J]. IEEE Transactions on Information Technology in Biomedicine, 2007, 11(5): 537–543. doi: 10.1109/TITB.2006.890018
[3]	SAVELONAS M A, IAKOVIDIS D K, LEGAKIS I, et al. Active contours guided by echogenicity and texture for delineation of thyroid nodules in ultrasound images[J]. IEEE Transactions on Information Technology in Biomedicine, 2009, 13(4): 519–527. doi: 10.1109/TITB.2008.2007192
[4]	邵蒙恩, 严加勇, 崔崤峣, 等. 基于CV-RSF模型的甲状腺结节超声图像分割算法[J]. 生物医学工程研究, 2019, 38(3): 336–340. doi: 10.19529/j.cnki.1672-6278.2019.03.15 SHAO Meng’en, YAN Jiayong, CUI Xiaoyao, et al. Ultrasound image segmentation of thyroid nodule based on CV-RSF algorithm[J]. Journal of Biomedical Engineering Research, 2019, 38(3): 336–340. doi: 10.19529/j.cnki.1672-6278.2019.03.15
[5]	ZHAO Jie, ZHENG Wei, ZHANG Li, et al. Segmentation of ultrasound images of thyroid nodule for assisting fine needle aspiration cytology[J]. Health Information Science and Systems, 2013, 1: 5. doi: 10.1186/2047-2501-1-5
[6]	ALRUBAIDI W M H, PENG Bo, YANG Yan, et al. An interactive segmentation algorithm for thyroid nodules in ultrasound images[C]. The 12th International Conference on Intelligent Computing, Lanzhou, China, 2016: 107–115. doi: 10.1007/978-3-319-42297-8_11.
[7]	RONNEBERGER O, FISCHER P, and BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]. The 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Germany, 2015: 234–241. doi: 10.1007/978-3-319-24574-4_28.
[8]	CHAN T F and VESE L A. Active contours without edges[J]. IEEE Transactions on Image Processing, 2001, 10(2): 266–277. doi: 10.1109/83.902291
[9]	LI Chunming, KAO C Y, GORE J C, et al. Minimization of region-scalable fitting energy for image segmentation[J]. IEEE Transactions on Image Processing, 2008, 17(10): 1940–1949. doi: 10.1109/TIP.2008.2002304
[10]	DING Jianrui, HUANG Zichen, SHI Mengdie, et al. Automatic thyroid ultrasound image segmentation based on u-shaped network[C]. The 12th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, Suzhou, China, 2019: 1–5. doi: 10.1109/CISP-BMEI48845.2019.8966062.
[11]	WANG Jianrong, ZHANG Ruixuan, WEI Xi, et al. An attention-based semi-supervised neural network for thyroid nodules segmentation[C]. 2019 IEEE International Conference on Bioinformatics and Biomedicine, San Diego, USA, 2019: 871–876. doi: 10.1109/BIBM47256.2019.8983288.
[12]	WU Yating, SHEN Xueliang, BU Feng, et al. Ultrasound image segmentation method for thyroid nodules using ASPP fusion features[J]. IEEE Access, 2020, 8: 172457–172466. doi: 10.1109/ACCESS.2020.3022249
[13]	ABDOLALI F, KAPUR J, JAREMKO J L, et al. Automated thyroid nodule detection from ultrasound imaging using deep convolutional neural networks[J]. Computers in Biology and Medicine, 2020, 122: 103871. doi: 10.1016/j.compbiomed.2020.103871
[14]	OKTAY O, SCHLEMPER J, FOLGOC L L, et al. Attention U-Net: Learning where to look for the pancreas[J]. arXiv preprint arXiv: 1804.03999, 2018.
[15]	WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]. The 15th European Conference on Computer Vision, Munich, Germany, 2018: 3–19. doi: 10.1007/978-3-030-01234-2_1.
[16]	FU Jun, LIU Jing, TIAN Haijie, et al. Dual attention network for scene segmentation[C]. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 3146–3154. doi: 10.1109/CVPR.2019.00326.
[17]	LEE H J, KIM J U, LEE S, et al. Structure boundary preserving segmentation for medical image with ambiguous boundary[C]. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, USA, 2020: 4817–4826. doi: 10.1109/CVPR42600.2020.00487.
[18]	ZHONG Zilong, LIN Zhongqiu, BIDART R, et al. Squeeze-and-attention networks for semantic segmentation[C]. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, USA, 2020: 13065–13074. doi: 10.1109/CVPR42600.2020.01308.
[19]	SZEGEDY C, IOFFE S, VANHOUCKE V, et al. Inception-v4, inception-ResNet and the impact of residual connections on learning[J]. arXiv preprint arXiv: 1602.07261, 2016.
[20]	KINGMA D P and BA J. Adam: A method for stochastic optimization[J]. arXiv preprint arXiv: 1412.6980, 2017.
[21]	AZAD R, ASADI-AGHBOLAGHI M, FATHY M, et al. Attention deeplabv3+: Multi-level context attention mechanism for skin lesion segmentation[C]. The European Conference on Computer Vision, Glasgow, UK, 2020. doi: 10.1007/978-3-030-66415-2_16.
[22]	AZAD R, ASADI-AGHBOLAGHI M, FATHY M, et al. Bi-directional ConvLSTM U-Net with densley connected convolutions[C]. 2019 IEEE/CVF International Conference on Computer Vision Workshop, Seoul, Korea (South), 2019: 406–415, doi: 10.1109/ICCVW.2019.00052.
[23]	ZHOU Zongwei, SIDDIQUEE M M R, TAJBAKHSH N, et al. Unet++: A nested u-net architecture for medical image segmentation[C]. The 4th International Workshop, DLMIA 2018, and 8th International Workshop, ML-CDS 2018, Held in Conjunction with MICCAI 2018, Granada, Spain, 2018: 3–11. doi: 10.1007/978-3-030-00889-5_1.
[24]	HU Jie, SHEN Li, and SUN Gang. Squeeze-and-excitation networks[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, 2018: 7132–7141. doi: 10.1109/cvpr.2018.00745.