摘要
随着电子计算机断层摄影、核磁共振成像技术等医学成像技术的产生和发展,人们获得了越来越多可供临床诊断的医学图像,如何有效地利用计算机技术对其进行分类已成为国内外许多学者研究的一个热点问题。另外,在现实环境中,许多医院的医学图像数据是分散存放的,它们之间除了通过网络传递信息外,其它资源全部独立。因此,研究并提出适合于分布式医学图像的分类方法具有较高的学术价值和广泛的应用前景。
论文讨论了分布式医学图像分类挖掘的背景、意义和国内外研究状况。针对医学图像存放的特点及分布式医学图像分类挖掘的实际应用要求,提出了解决分布式医学图像分类的方法。对现有决策树算法存在的问题,提出基于SLIQ的改进算法DMI_SLIQ,较好地解决了分布式分类挖掘问题,初步建立了基于SLIQ的分布式医学图像分类框架。
论文从理论、算法和应用三个方面对分布式医学图像分类方法进行了研究,主要工作包括:
1、本文将分类技术与数字图像处理技术有机结合,研究了分布式医学图像的主要分类方法,并给出了本文实验所用到的医学图像特征。
2、针对SLIQ算法的缺点,提出了DMI_SLIQ算法,从而解决了SLIQ算法搜索空间巨大,耗时多等问题。
3、设计并实现了一个分布式协调器算法,用来初始化和管理Agent的分类活动,管理和控制事务的运行。
4、针对医学图像分布式存放的特点,提出了一个适用于分布式医学图像分类的框架。该框架包括:表示层、处理层、采掘层。
5、实现了分布式医学图像分类实验系统—DMIDM,在系统中实现了本文提出的分布式协调器任务调度算法、全局模型规则生成算法。
With the development of computer tomography, magnetic resonance imaging technology, and other medical imaging technology, People has gained the more and more diagnosis of medical image , how effective use of computer technology to classify so many scholars has become a hot issue both at home and abroad. In reality, medical image data are scattered store in many hospitals. They transfer the information outside between them except passing a network, other all independences of resource. Therefore, the study and propose appropriate to the classification of medical images distributed method has a fairly high academic value and broad application prospects.
The paper take the medicine image data as the research object, from the theory, the algorithm and applied three aspects to conduct there search to the distributional medicine picture classification method. In view of the existing policy-making tree algorithm existence question, we proposed improvement algorithm DMI_SLIQ based on SLIQ, DMI_SLIQ has solved the distributional classified excavation problem. This paper has established distributional medicine image classification frame based on the SLIQ.
The paper from the theory, the algorithm and applied three aspects to conduct the research to the distributional medicine picture classification method, the main work included
1. This article classifies technical and the digital medical image processing technology, and has studied the main method about the distributional medicine picture classified, and has produced the medicine picture characteristic which this article tests uses.
2. In view of the decision tree algorithm SLIQ shortcoming, this paper improved and proposed the DMI_SLIQ algorithm. Thus solves the SLIQ algorithm search space greatly, consumes when are to many and so on the question, and compared with algorithm performance, through analysis experiment result.
3. In view of the medicine picture distributional depositing characteristic, distributional medicine image classification theory frame is proposed suitable for the issue. This frame divides into three parts, we produces each detailed design thought and the principle of work.
4. This paper has realized distributional medicine image classification experiment system DMIDM, and realized the distributional synchronizer duty dispatch algorithm, the overall situation model rule production algorithm in the system which this article proposed. The framework includes: that layer, layer handling, and mining layer.
引文
[1]Gagan Agrawal,Ruoming Jin,and Xiao gang Li.Middleware and compiler support for scalable data mining.In Proceedings of Languages and Compilers for Parallel Computing (LCPC),2006.
[2]瞿爱珍,庄天戈,计算机辅助医学诊断系统的数据挖掘和知识发现研究,国外医学生物医学工程分册,Vol.245(3),2002,pp.97-103.
[3]朱玉全,杨鹤标,孙蕾编著.数据挖掘技术.南京:东南大学出版社,2006.
[4]H.Kargupta,B.Park,E.Johnson,et al.Collective data mining from distributed vertically partitioned feature space.Workshop on distributed data mining.Conference on Knowledge Discovery and Data Mining.New York,USA,2000.
[5]A.L.Prodromidis,P.K.Chan,S.J.Stolfo.Meta-learning in distributed data mining Systems:issues and approaches.Advances in distributed data mining.AAAI Press,1999.
[6]Jiawei Han,Micheline Kamber Data Mining,Second Edition,Second Edition:Concepts and Techniques Morgan Kaufmann.
[7]Jiawei Han,Micheline Kamber,Data Mining Concepts and Techniques,China Machine Press,2001.
[8]M.L.Liu.Distributed Computing:Principles and Applications.顾铁成,王亚丽,叶保留.清华大学出版社,2004.
[9]Rachid Guerraoui,Mohamed E.Fayad.OO distributed programming is not distributed OO programming.Communications of the ACM.1999,42(4).
[10]C.E.Shannon.A Mathematical Theory of Communication.B.S.T.J.vol.27,July and Oct.1948.
[11]藤田宏一,魏鸿骏,陈尚勤泽。 基础信息论。 国防工业出版社,1982
[12]Manish Mehta,Rakesh Agrawal and Jorma Rissanen.SLIQ:A Fast and Scalable Classifier for Data Mining.IBM Alma den Research Center,1996
[13]Ed Wilson.The Knowledge Discovery Process,A Problem Solving Methodology.Computer Associates International,Inc.1998
[14]http://define.cnki.net/define_result.aspx?searchword=%E7%81%B0%E5%BA%A6%E7%89%B9%E5%BE%81
[15]http://define.cnki.net/define_result.aspx?searchword=%E5%87%A0%E4%BD%95%E7%89%B9%E5%BE%81
[16]http://define.cnki.net/define_searchTable.aspx?searchword=%E7%BA%B9%E7%90%86%E 7%89%B9%E5%BE%81
[17]http://define.cnki.net/define_searchTable.aspx?searchword=%E7%BA%B9%E7%90%86%E 7%85%87%A0%E4%BD
[18]Adonis,J.M.Kolluri,V.Provost,F.J et al.The World:Knowledge discovery from Multiple distributed database.Technical Report ISL-96-6.Department of Computer Science,University of Pittsburgh.1996
[19]T.Hilbel,M.A.Reiter.Advantages of a cardiac DICOM network server writer for viewing and permanent CD-R archiving Of cardiovascular X-ray angiography images.Computers in Cardiology,2000.649-652
[20]Mao J.and Jain A.K,A Self-organizing network for hyper ellipsoidal clustering[J].IEEE Transactions on Neural Networks.1996;7(1):16-29.
[21]J.ERedlich.Distributed object technology for networking.IEEE Communication Magazine.1998,26(8).
[22]BEA.BEA SOA域模型:实现面向服务的架构的业务优势.BEA系统公司,2005:5-24T.Hilbel,M.A.Reiter.Advantages of a cardiac DICOM network server writer for viewing and permanent CD-R archiving of cardiovascular X-ray angiography images.Computers in Cardiology,2000.649-652.
[23]R.Guerraoui Distributed programming abstractions.ACM Computing Surveys.1996,28(4es).
[24](美)Jeff Garland,Richard Anthony.大型软件体系结构:使用UML实践指南.叶俊民,汪望珠等译.电子工业出版社,2004:235-237。
[25]Ping Zhang,Brijesh Venma,Kuldeep Kumar,A neural-genetic algorithm for feature selection and breast abnormality classification in digital mammography,Proceedings of the 2004 IEEE International Joint Conference on Neural Networks,2004,2303-2308.
[26]M.Kandemir,A.Choudhary,J.Ramanujam,and M.A..Kandaswamy.A unified framework for optimizing locality,parallelism,and communication in out-of-core computations.IEEETransactions on Parallel and Distributed Systems,11(9):648-662,2000.
[27]K.Yamanishi,J.Takeuchi,G.Williams,and P.Milne.Online unsupervised outlier detection using finite mixtures with discounting learning algorithms.In Proceedings of the Sixth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,pages 320-324,2000.
[28]ACR-NEMA联合委员会.医学数字影像和通讯标准.美国放射学会.2003-10
[29]宾宁,李宏,陈松乔.基于SPRINT分类算法的异构分布式数据挖掘研究.计算机测量与控制,2005,1(13)
[30]Yingwu Fang,Guangpeng Zhang,Research on distributive data mining calculating process-DDCP algorithm,Journal of university of electronic science and technology,Vol 32,No.1,pp.80-84,Feb.2003.
[31]UIUC National Center for Data Mining.Predictive Model Markup Language,1999.See Web site at http:///www.ncdm.uic.edu/.
[32]R.Olejnik,A.Bouchi,B.Toursel.A java objects policy for load balancing.PDPTA:The international Conference on Parallel and Distributed Processing Techniques and Applications.Vol.2,pp.816-821.Las Vegas,USA,June 2002.
[33]V.Fiolet and B.Toursel,Progressive Clustering for Database Distribution on a Grid,In Proc.of ISPDC 2005,IEEE Computer Society,July 2005,pp.282-289.
[34]蒋良孝,蔡之华.分布式数据挖掘研究[J].计算机与现代化,2002,85(9):4-7.
[35]郭威 北京工业大学 分布式系统的进程间通信技术的研究和开发 2006-5。