首页 > 论文 > 激光与光电子学进展 > 57卷 > 20期(pp:201008--1)

基于AttentionNet和DenseUnet的脊椎CT分割

Spinal CT Segmentation Based on AttentionNet and DenseUnet

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

在脊椎CT图像分割问题中,由于脊椎与组织对比度过低和噪声的影响,传统分割算法存在分割精度差和自动化程度低等问题。基于此,提出一种通过AttentionNet定位脊椎,然后使用改进的DenseUnet进行脊椎CT分割的方法。首先,对所有脊椎CT样本数据进行裁剪、重采样、灰度值归一化等预处理操作;再次,对样本使用AttentionNet训练得到具有位置信息的Attention图;然后,对传统DenseUnet进行改进,在每个Dense block加入Shuffle操作来增加网络的鲁棒性,在每个Dense block后加入1×1卷积,以降低通道数,减少网络参数量;接着使用改进后的DenseUnet对训练样本进行预训练,得到具有先验信息的预测图;最后,将Attention图、预测图及原始图像融合为三通道的训练样本作为输入,采用改进的DenseUnet训练分割模型,并在测试集上进行验证,最终实现脊椎CT自动分割。实验结果表明,所提方法的分割精度优于传统DenseUnet,是一种有效的脊椎CT自动分割方法。

Abstract

In the spinal computed tomography (CT) image segmentation problem, owing to the low contrast between the spine and tissues, and the influence of noise, the traditional segmentation algorithms have problems such as poor segmentation accuracy and low degree of automation. Aiming at solving the above-mentioned problems, a method of locating the spine through AttentionNet and then using improved DenseUnet to perform spinal CT segmentation is proposed herein. First, preprocessing operations such as cropping, resampling, and normalization of gray values are performed on all spinal CT sample data; the samples are trained using AttentionNet to obtain Attention maps with position information. Second, the traditional DenseUnet is improved, and each Dense block adds the Shuffle operation to increase the network robustness. After each Dense block, a 1×1 convolution is added to reduce the number of channels and network parameters. Third, the training samples are pretrained using the improved DenseUnet to obtain the prediction maps with prior information. Finally, the Attention map, prediction map, and the original images are fused into three-channel training samples as the input, and the improved DenseUnet is used to train segmentation model and is verified against the test set. Consequently, the spinal CT automatic segmentation is realized. The experimental results show that the segmentation accuracy of the proposed method is better than that of the traditional DenseUnet, and the proposed method is an effective automatic segmentation method for spinal CT.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:TP37

DOI:10.3788/LOP57.201008

所属栏目:图像处理

基金项目:国家重点研发项目、国家自然科学基金、陕西省产业创新链项目、陕西省重点研发计划;

收稿日期:2019-12-17

修改稿日期:2020-02-25

网络出版日期:2020-10-01

作者单位    点击查看

田丰源:西北大学信息科学与技术学院, 陕西 西安 710127
周明全:西北大学信息科学与技术学院, 陕西 西安 710127
闫峰:西北大学信息科学与技术学院, 陕西 西安 710127
范力:西北大学信息科学与技术学院, 陕西 西安 710127
耿国华:西北大学信息科学与技术学院, 陕西 西安 710127

联系人作者:耿国华(ghgeng@nwu.edu.cn)

备注:国家重点研发项目、国家自然科学基金、陕西省产业创新链项目、陕西省重点研发计划;

【1】Jung C R, Scharcanski J. Robust watershed segmentation using wavelets [J]. Image and Vision Computing. 2005, 23(7): 661-669.

【2】Litjens G, Kooi T, Bejnordi B E, et al. A survey on deep learning in medical image analysis [J]. Medical Image Analysis. 2017, 42: 60-88.

【3】Kaur G, Chhaterji J. A survey on medical image segmentation [J]. International Journal of Science and Research. 2017, 6(5): 1305-1311.

【4】Sungkhun S, Rasmequan S, Chinnasarn K, et al. Vertebral body segmentation using aggregate superpixels . [C]∥2016 13th International Joint Conference on Computer Science and Software Engineering (JCSSE), July 13-15, 2016, Khon Kaen, Thailand. New York: IEEE. 2016.

【5】Li G, Liu H, Ke S Q. Application of spine images segment and extract information [J]. Electronic Design Engineering. 2017, 25(16): 183-186.
李滚, 刘欢, 柯善群. 脊椎图像分割与椎体信息提取的算法及应用 [J]. 电子设计工程. 2017, 25(16): 183-186.

【6】Liu J P, Yang Y Z, Chen M Y, et al. Image dehazing algorithm based on convolutional neural network and dynamic ambient light [J]. Acta Optica Sinica. 2019, 39(11): 1110002.
刘杰平, 杨业长, 陈敏园, 等. 结合卷积神经网络与动态环境光的图像去雾算法 [J]. 光学学报. 2019, 39(11): 1110002.

【7】Feng Y, Yi B S, Wu C Y, et al. Pulmonary nodule recognition based on three-dimensional convolution neural network [J]. Acta Optica Sinica. 2019, 39(6): 0615006.
冯雨, 易本顺, 吴晨玥, 等. 基于三维卷积神经网络的肺结节识别研究 [J]. 光学学报. 2019, 39(6): 0615006.

【8】Chen Y, Guo H G, Ai Y P. Single image dehazing method based on multi-scale convolution neural network [J]. Acta Optica Sinica. 2019, 39(10): 1010001.
陈永, 郭红光, 艾亚鹏. 基于多尺度卷积神经网络的单幅图像去雾方法 [J]. 光学学报. 2019, 39(10): 1010001.

【9】Shelhamer E, Long J, Darrell T. Fully convolutional networks for semantic segmentation [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2014, 39(4): 640-651.

【10】Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation [M]. ∥Navab N, Hornegger J, M.Wells W, et al. Medical image computing and computer-assiated intervention-MICCAI 2015. Lecture notes in computer science. Cham: Springer. 2015, 9351: 234-241.

【11】Liu Z L, Chen G, Shan Z Y, et al. Segmentation of spine CT image based on deep learning [J]. Computer Applications and Software. 2018, 35(10): 200-204, 273.
刘忠利, 陈光, 单志勇, 等. 基于深度学习的脊柱CT图像分割 [J]. 计算机应用与软件. 2018, 35(10): 200-204, 273.

【12】Wang Y G, Wang M, Han J G, et al. R-U-Net neural network for automatic segmentation of femur area [J]. Mini-Micro Systems. 2019, 40(4): 839-844.
王亚刚, 王萌, 韩俊刚, 等. 一种自动分割股骨区域的R-U-Net神经网络 [J]. 小型微型计算机系统. 2019, 40(4): 839-844.

【13】Feng J W, Zhang L M, Deng X Y. Image segmentation based on multi-source fusion FCN [J]. Application Research of Computers. 2018, 35(9): 2877-2880.
冯家文, 张立民, 邓向阳. 基于多源融合FCN的图像分割 [J]. 计算机应用研究. 2018, 35(9): 2877-2880.

【14】Sekuboyina A, Kukacka J, Kirschke J S, et al. Attention-driven deep learning for pathological spine segmentation [M]. ∥Glocker B, Yao J H, Vrtovec T, et al. Computational methods and clinical applications in musculoskeletal imaging. Lecture notes in computer science. Cham: Springer. 2018, 10734: 108-119.

【15】Li X M, Chen H, Qi X J, et al. H-DenseUNet: hybrid densely connected UNet for liver and tumor segmentation from CT volumes [J]. IEEE Transactions on Medical Imaging. 2018, 37(12): 2663-2674.

【16】Ibragimov B, Likar B, Pernu? F, et al. Shape representation for efficient landmark-based segmentation in 3-D [J]. IEEE Transactions on Medical Imaging. 2014, 33(4): 861-874.

【17】Korez R, Ibragimov B, Likar B, et al. A framework for automated spine and vertebrae interpolation-based detection and model-based segmentation [J]. IEEE Transactions on Medical Imaging. 2015, 34(8): 1649-1662.

【18】Liu G L, Shih K J, Wang T C, et al. -11-28)[2019-12-16] . https:∥arxiv. 2018, org/abs/1811: 11718.

【19】Yu F. -04-30)[2019-12-16] . https:∥arxiv. 2016, org/abs/1511: 07122.

【20】Wang P Q, Chen P F, Yuan Y, et al. -06-01)[2019-12-16] . https:∥arxiv. 2018, org/abs/1702: 08502.

【21】Koltchinskii V, Panchenko D. Complexities of convex combinations and bounding the generalization error in classification [J]. Annals of Statistics. 2005, 33(4): 1455-1496.

【22】Huang G, Liu Z, van der Maaten L, et al. -01-28)[2019-12-16] . https:∥arxiv. 2018, org/abs/1608: 06993.

【23】Kingma D P. -01-30)[2019-12-16] . https:∥arxiv. 2017, org/abs/1412: 6980.

引用该论文

Tian Fengyuan,Zhou Mingquan,Yan Feng,Fan Li,Geng Guohua. Spinal CT Segmentation Based on AttentionNet and DenseUnet[J]. Laser & Optoelectronics Progress, 2020, 57(20): 201008

田丰源,周明全,闫峰,范力,耿国华. 基于AttentionNet和DenseUnet的脊椎CT分割[J]. 激光与光电子学进展, 2020, 57(20): 201008

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF