|本期目录/Table of Contents|

[1]宋云霞,强 彦*,赵涓涓,等.基于视觉信息与征象标签的肺结节CT图像检索[J].南京大学学报(自然科学),2017,53(6):1043.[doi:10.13232/j.cnki.jnju.2017.06.006]
 Song Yunxia,Qiang Yan*,Zhao Juanjuan,et al.Pulmonary nodule CT images retrieval based on visual information and signs label[J].Journal of Nanjing University(Natural Sciences),2017,53(6):1043.[doi:10.13232/j.cnki.jnju.2017.06.006]
点击复制

基于视觉信息与征象标签的肺结节CT图像检索()
     

《南京大学学报(自然科学)》[ISSN:0469-5097/CN:32-1169/N]

卷:
53
期数:
2017年第6期
页码:
1043
栏目:
出版日期:
2017-12-01

文章信息/Info

Title:
Pulmonary nodule CT images retrieval based on visual information and signs label
作者:
宋云霞1强 彦1*赵涓涓1唐笑先2田 奇3
1.太原理工大学计算机科学与技术学院,太原,030024;
2.山西省人民医院CT室,太原,030012;
3.美国德克萨斯大学圣安东尼奥分校计算机科学系,圣安东尼奥,TX 78249,美国
Author(s):
Song Yunxia1Qiang Yan1*Zhao Juanjuan1Tang Xiaoxian2Tian Qi3
1.School of Computer Science and Technology,Taiyuan University of Technology,Taiyuan,030024,China;
2.Shanxi Provincial People’s Hospital,CT Room,Taiyuan,030012,China;
3.Department of Computer Science,University of Texas at San Antonio,San Antonio,TX 78249,USA
关键词:
肺结节图像多特征概率超图哈 希图像检索
Keywords:
pulmonary nodule imagesmulti-featureprobability hypergraphhashing methodimage retrieval
分类号:
TP391
DOI:
10.13232/j.cnki.jnju.2017.06.006
文献标志码:
A
摘要:
肺结节CT图像的相似性检索是计算机辅助诊断系统中最重要的部分,目前常用的检索方法通常匹配精度低,检索速度慢.针对上述问题,提出一种新的基于视觉信息与征象标签的双概率超图哈希算法,使用两层结构提高肺结节图像的检索精度:在第一层,将肺结节影像视觉信息和标签信息分别构建概率超图,最优划分概率超图得到哈希码;在第二层,使用结节图像的视觉特征、标签特征和第一层得到的哈希码来训练哈希函数.在检索时,对待检图像通过训练好的哈希函数进行0,1编码,与数据集中图像比较汉明距离,返回相似结节图像.对9种不同征象类型的3422张肺结节CT图像进行实验,并与不同哈希算法进行比较,结果表明,提出的方法在哈希码长为32位时可以达到最高精度90.18%,有效提高了检索精度,可以给医生提供客观的辅助诊断.
Abstract:
The similarity search of CT images of pulmonary nodules is the most important part of computer aided diagnosis system.At present,commonly used retrieval methods have problems of low matching accuracy and slow retrieval speed.Aiming at these problems,we propose a new hashing algorithm used double probabilistic hypergraphs based on visual information and signs label,using two-layer structure to improve retrieval precision.On the first layer,we utilize visual information and label information to construct probabilistic hypergraph respectively,and hashing codes can be obtained by the optimal partition of probabilistic hypergraph.On the second layer,we use visual features,labels and hashing codes obtained from the first layer to train hashing functions.On retrieval stage,we encode the query image by training hashing functions into 0 or 1,comparing hamming distance with the database images,and returning similar nodule images.In this paper,experiments were performed on 3422 pulmonary nodule CT images with 9 different types of signs.Compared with different hashing algorithms,the results show that the proposed method can achieve higher accuracy of 90.18% when the length of hashing code is 32-bit,and effectively improves the retrieval accuracy,which can provide the doctors with objective auxiliary diagnosis.

参考文献/References:

[1] Soltani T,Salari R,Ferdousi R.Make a good diagnosis on clinical images by ubiquitous decision support tools.Iranian Imaging Informatics Conference,2015:37.
[2] 潘 玲,杜晓平,赵涓涓.基于有监督哈希的肺结节CT图像检索.计算机应用研究,2017,34(9):2838-2842.(Pan L,Du X P,Zhao J J.Lung nodules CT image retrieval based on supervised hashing.Application Research of Computers,2017,34(9):2838-2842.)
[3] Schlkopf B,Platt J,Hofmann T.Learning with hypergraphs:Clustering,classification,and embedding.In:2006 Conference on Advances in Neural Information Processing Systems.Vancouver,Canada:MIT Press,2007:1601-1608.
[4] Liu Y,Shao J,Xiao J,et al.Hypergraph spectral hashing for image retrieval with heterogeneous social contexts.Neurocomputing,2013,119:49-58.
[5] Weiss Y,Torralba A,Fergus R.Spectral hashing.In:21st International Conference on Neural Information Processing Systems.Vancouver,Canada:Curran Associates Inc.,2009:1753-1760.
[6] Zhu L,Shen J L,Xie L,et al.Unsupervised topic hypergraph hashing for efficient mobile image retrieval.IEEE Transactions on Cybernetics,2017,47(11):3941-3954.
[7] Huang Y C,Liu Q S,Zhang S T,et al.Image retrieval via probabilistic hypergraph ranking.In:2010 IEEE Conference on Computer Vision and Pattern Recognition.San Francisco,CA,USA:IEEE,2010:3376-3383.
[8] Armato III S G,McNitt-Gray M F,Reeves A P,et al.The Lung Image Database Consortium(LIDC):An evaluation of radiologist variability in the identification of lung nodules on CT scans.Academic Radiology,2007,14(11):1409-1421.
[9] Qiang Y,Zhang X H,Ji G H,et al.Automated lung nodule segmentation using an active contour model based on PET/CT images.Journal of Computational and Theoretical Nanoscience,2015,12(8):1972-1976.
[10] Ouyang J L,Liu Y Z,Shu H Z.Robust hashing for image authentication using SIFT feature and quaternion Zernike moments.Multimedia Tools and Applications,2017,76(2):2609-2626.
[11] Bengio Y,Delalleau O,Le Roux N,et al.Learning eigenfunctions links spectral embedding and kernel PCA.Neural Computation,2004,16(10):2197-2219.
[12] Zhuang Y T,Liu Y,Wu F,et al.Hypergraph spectral hashing for similarity search of social image.In:19th ACM International Conference on Multimedia.Scottsdale,AR,USA:ACM,2011:1457-1460.
[13] Zhang D,Wang J,Cai D,et al.Self-taught hashing for fast similarity search.In:33rd International ACM SIGIR Conference on Research and Development in Information Retrieval.Geneva,Switzerland:ACM,2010:18-25.
[14] Adankon M M,Cheriet M.Support vector machine.In:Li S Z,Jain A.Encyclopedia of biometrics.Springer New York,2015,1303-1308. 
[15] Andoni A,Indyk P.Near-optimal hashing algorithms for approximate nearest neighbor in high dimensions.Communications of the ACM,2006,51(1):117-122.

相似文献/References:

备注/Memo

备注/Memo:
 基金项目:国家自然科学基金(61373100),山西省回国留学人员科研资助项目(2016-038),虚拟现实技术与系统国家重点实验室开放基金(BUAA-VR-17KF-14,BUAA-VR-17KF-15)
收稿日期:2017-10-28
*通讯联系人,E-mail:qiangyan99@foxmail.com
更新日期/Last Update: 2017-11-26