基于D-Unet神经网络的鼻腔鼻窦肿瘤分割算法

doi:10.19678/j.issn.1000-3428.0060120

计算机工程 ›› 2022, Vol. 48 ›› Issue (1): 281-287. doi: 10.19678/j.issn.1000-3428.0060120

基于D-Unet神经网络的鼻腔鼻窦肿瘤分割算法

李富豪¹, 赵希梅^1,2

1. 青岛大学计算机科学技术学院, 山东青岛 266071;
2. 山东省数字医学与计算机辅助手术重点实验室, 山东青岛 266071

收稿日期:2020-11-26 修回日期:2021-01-11 发布日期:2021-01-20
作者简介:李富豪(1996-),男,硕士研究生,主研方向为计算机视觉、医学图像处理;赵希梅(通信作者),副教授。
基金资助:
国家自然科学基金（61303079）。

Nasal Cavity and Paranasal Sinuses Tumor Segmentation Algorithm Based on D-Unet Neural Network

LI Fuhao¹, ZHAO Ximei^1,2

1. College of Computer Science and Technology, Qingdao University, Qingdao, Shandong 266071, China;
2. Shandong Province Key Laboratory of Digital Medicine and Computer Aided Surgery, Qingdao, Shandong 266071, China

Received:2020-11-26 Revised:2021-01-11 Published:2021-01-20

摘要/Abstract

摘要： 鼻腔鼻窦肿瘤为多发性疾病，其CT影像具有形态不规则、分界不均匀等特征，而现有的U-Net分割算法对图片细节不敏感且割裂了图像局部与整体特征的一致性，难以实现精准诊断。提出一种基于D-Unet深度神经网络的改进算法，根据鼻腔鼻窦肿瘤空间形变特点，将可变形卷积融入U-Net网络，并利用可变形卷积能依据目标形态拥有自适应感受野的特点，充分学习图像细节，从而提升算法的特征提取能力。在此基础上，使用损失函数Tversky解决数据集样本失衡问题，从而获得更高的灵敏度和泛化能力。为方便进一步研究，建立鼻腔鼻窦肿瘤分割数据集。实验结果表明，所提算法能有效提高鼻腔鼻窦肿瘤分割精度，相比U-Net、Res-Unet和Attention U-Net算法，分割精度分别提高了5.01%、2.56%和0.48%。

关键词: 鼻腔鼻窦肿瘤, U-Net算法, 目标分割, 可变形卷积网络, Tversky损失函数

Abstract: Nasal cavity and paranasal sinus tumor are a type of multiple disease, and their CT images are characterized by irregular shape as well as uneven boundary.When dealing with these images, the existing U-Net segmentation algorithms are not sensitive to image details, and split the consistency of local and overall features of the picture, so the accuracy of diagnosis is reduced.To address the problem, this paper proposes an improved D-Unet based algorithm.Considering the spatial deformation of nasal cavity and paranasal sinus tumors, deformable convolutions are introduced into U-Net.The convolutions can adapt the receptive field to the target shape, fully learn image details, and enhance the feature extraction ability.On this basis, Tversky loss function is used to address imbalanced sample data sets, so higher sensitivity and generalization ability are obtained.In addition, a data set of nasal cavity and paranasal sinus tumor segmentation is established to facilitate further research.Experimental results show that the proposed algorithm can effectively improve the segmentation accuracy of nasal cavity and paranasal sinus tumors.The accuracy is 5.01% higher than U-Net, 2.56% higher than Res-Unet and 0.48% higher than Attention U-Net.

Key words: nasal cavity and paranasal sinus tumor, U-Net based algorithm, object segmentation, Deformable Convolution Network(DCN), Tversky loss function

中图分类号:

TP391.41

李富豪, 赵希梅. 基于D-Unet神经网络的鼻腔鼻窦肿瘤分割算法[J]. 计算机工程, 2022, 48(1): 281-287.

LI Fuhao, ZHAO Ximei. Nasal Cavity and Paranasal Sinuses Tumor Segmentation Algorithm Based on D-Unet Neural Network[J]. Computer Engineering, 2022, 48(1): 281-287.

http://www.ecice06.com/CN/Y2022/V48/I1/281

图/表 10

20220108123952

20220108123955

20220108123959

20220108124003

20220108124006

20220108124010

20220108124014

20220108124018

20220108124021

20220108124025

参考文献

[1] 李文东, 刘文君.鼻腔鼻窦肿瘤的临床特征与病理组织分类及其影响因素分析[J].中国实用医药, 2015(5):60-61. LI W D, LIU W J.Clinical features, pathological classification and influencing factors of tumors in nasal cavity and paranasal sinuses[J].China Practical Medicine, 2015(5):60-61.(in Chinese)
[2] 马芡, 姚选军, 钱斌.CT断层扫描联合CD24检测对早期鼻窦癌的诊断价值[J].CT理论与应用研究, 2018, 27(4):537-542. MA Q, YAO X J, QIAN B.Diagnostic value of CT tomography combined with CD24 detection in the early sinus carcinoma[J].CT Theory and Applications, 2018, 27(4):537-542.(in Chinese)
[3] 李根柱, 石木兰.鼻腔和鼻窦腺样囊性癌的CT诊断[J].中华放射学杂志, 1998(3):3-5. LI G Z, SHI M L.CT diagnosis of adenoid cystic carcinoma of the nasal cavity and paranasal sinus[J].Chinese Journal of Radiolog, 1998(3):3-5.(in Chinese)
[4] SCHECHTMAN S A, WERTZ A P, SHANKS A, et al.Preoperative β-blockade and hypertension in the first hour of functional endoscopic sinus surgery[J].Laryngoscope, 2017, 127(7):1496-1505.
[5] SHIN H, ROTH H R, GAO M, et al.Deep convolutional neural networks for computer-aided detection:CNN architectures, dataset characteristics and transfer learning[J].IEEE Transactions on Medical Imaging, 2016, 34(3):1285-1298.
[6] TAO C, TU S X, WANG H, et al.Computer-aided diagnosis of gallbladder polyps based on high resolution ultr asonography[J].Computer Methods and Programs in Biomedicine, 2020, 185(7):105-118.
[7] YANG J F, QIAO P R, LI Y M, et al.Review of machine learning classification problems and algorithms research[J].Statistics and Decision, 2019, 35(6):36-40.
[8] NANDA A S, SWARNKAR T.Deep learning approach for image classification[C]//Proceedings of the 2nd International Conference on Data Science and Business Analytics.Washington D.C., USA:IEEE Press, 2018:97-101.
[9] PASSERA K.A fuzzy-C-means cluster ing algorithm for a volumetr ic analysis of paranasal sinus and nasal cavity cancers[C]//Proceedings of International Conference of the IEEE Engineering in Medicine and Biology Society.Washington D.C., USA:IEEE Press, 2006:3078-3081.
[10] PASSERA K M, POTEPAN P, BRAMBILLA L, et al.ITAC volume assessment through a gaussian hidden markov random field model-based algorithm[C]//Proceedings of International Conference of the IEEE Engineering in Medicine and Biology Society.Washington D.C., USA:IEEE Press, 2008:1218-1221.
[11] RONNEBERGER O, FISCHER P, BROX T.U-Net:convolutional networks for biomedical image segmentation[C]//Proceedings of International Conference on Medical Image Computing and Computer-Assisted Intervention.Berlin, Germany:Springer, 2015:234-241.
[12] DAI J F, QI H Z, XIONG Y W, et al.Deformable convolutional networks[C]//Proceedings of the IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:764-773.
[13] SALEHI S S M, ERDOGMUS D, GHOLIPOUR A.Tversky loss function for image segmentation using 3D fully convolutional deep networks[C]//Proceedings of International Workshop on Machine Learning in Medical Imaging.Berlin, Germany:Springer, 2017, 379-387.
[14] SHAN H, ZHANG Y, YANG Q, et al.3D convolutional encoder-decoder network for low-dose CT via transfer learning from a 2D trained network[J].IEEE Transactions on Medical Imaging, 2018, 37(6):1522-1533.
[15] HUANG G, LIU Z, MAATEN L V D, et al.Densely connected convolutional networks[J].IEEE Computer Society, 2017, 34(5):2261-2269.
[16] SHELHAMER E, LONG J, DARRELL T.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(4):640-651.
[17] SHAN H, ZHANG Y, YANG Q, et al.3D convolutional encoder-decoder network for low-dose CT via transfer learning from a 2D trained network[J].IEEE Transactions on Medical Imaging, 2018, 37(6):1522-1532.
[18] HUANG G, LIU Z, MAATEN L V D, et al.Densely connected convolutional networks[EB/OL].[2020-10-06].http://en.cnki.com.cn/Article_en/CJFDTotal-JSJS201810019.htm.
[19] HEESWIJK M M, LAMBREGTS D M, GGIETHUYSEN J J, et al.Automated and semiautomated segmentation of rectal tumor volumes on diffusion-weighted MRI:can it replace manual volumetry?[J].International Journal of Radiation Oncology Biology Physics, 2016, 94(4):824-831.
[20] ABADI M, AGARWAL A, BARHAM P, et al.Tensorflow:large-scale machine learning on heterogeneous distributed systems[J].Preliminary White Paper, 2016(3):25-38.
[21] KINGMA D P, BA J.Adam:a method for stochastic optimization[EB/OL].[2020-10-06].https://arxiv.org/abs/1412.6980v8.

选择文件类型/文献管理软件名称

选择包含的内容

基于D-Unet神经网络的鼻腔鼻窦肿瘤分割算法

Nasal Cavity and Paranasal Sinuses Tumor Segmentation Algorithm Based on D-Unet Neural Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献

相关文章 7

编辑推荐

Metrics

本文评价

[1]	张华悦, 张顺利, 张利. 基于双阶段网络的交互式目标分割算法[J]. 计算机工程, 2021, 47(2): 300-306.
[2]	侯旺, 钟立军, 张小虎, 雷志辉. 基于类内一致性的红外背景弱目标检测方法[J]. 计算机工程, 2013, 39(5): 209-211,217.
[3]	陈舒, 刘秉瀚. 基于角点特征的粘连车辆分割方法[J]. 计算机工程, 2013, 39(4): 210-213,218.
[4]	刘青;汪同庆;李宏友. 基于水平集分割的3DOGHM检测算法[J]. 计算机工程, 2010, 36(7): 230-232.
[5]	邓颖娜;朱虹;刘薇;钱慧芳. 基于姿态模型的人群目标分割方法[J]. 计算机工程, 2010, 36(7): 195-197.
[6]	徐秋平;郭敏*. 基于变宽邻域图割和活动轮廓的目标分割方法[J]. 计算机工程, 2009, 35(8): 233-237.
[7]	何毅;陆淑娟;梅雪. 基于ROI提取的多目标图像水平集分割[J]. 计算机工程, 2009, 35(23): 214-216.

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于D-Unet神经网络的鼻腔鼻窦肿瘤分割算法

Nasal Cavity and Paranasal Sinuses Tumor Segmentation Algorithm Based on D-Unet Neural Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献

相关文章 7

编辑推荐

Metrics

本文评价