基于关联图模型的医学图像Top-k查询方法.pdfVIP

计算机研究与发展 7544／issnlOOO一1239．2015 JournalofComputerResearchandDevelopment 52(9)：2033-2045，2015 基于关联图模型的医学图像 Top—k查询方法 李鹏远 潘海为 李 青 韩启龙 谢晓芹 张志强 (哈尔滨工程大学计算机科学与技术学院 哈尔滨 150001) (香港城市大学计算机工程与信息技术系 香港) (pengyuan@live．cn) Top。kMedicalImagesQueryBasedonAssociationGraph LiPengyuan ，PanHaiwei，LiQing ，HanQilong ，XieXiaoqin ，andZhangZhiqiang (CollegeofComputerScienceandTechnology，HarbinEngineeringUniversity，Harbin150001) 。(DepartmentofComputerEngineeringandInformationTechnology，CityUniversityofHongKong，HongKong) Abstract Patient—to—patientcomparison，especially image—to—image comparison plays an important rolein themedicaldomain since doctors invariablymake diagnosesbasedon prior experiencesof similarcases．Itisvery significantfordoctorsto find similarmedicaIimagesfrom thedatabaseas similarpathologicalchangesinpriorpatients’imagesandcorrespondingreportscan assistdoctorsto makediagnosesfor currentpatients． Therefore，advanced medicalimageretrievaltechniqueshave beenwidely studied to improve theaccuracy in recentyears． However，theprocessing timehas becomeanotherproblem inmedicalimageretrievaldomainbecauseoftheincreasingnumberofmedical images．ASdoctorsareonly interested in themostsimilar是results，anovelmodelofassociation graphisproposedformedicalimagetop—kquery in thispaper．Thefuzzyexpression inaassociation graph can describe the similarity between images effectively．Moreover， a seriesofcorrelation measurementsareproposed forsimilarityreasoning．Then themedicalimagetop～足 querymethod is representedbasedon thecharactersofcorrelationmeasurements．Furthermore，fourwalk strategies arestudiedtoaccelerateandstabilizethetop-kprocess．Experimentalresultsshow thatitsefficiency and effectivenessarehigherincomparisonwith stateoftheart． Keywords associationgraph；Top—kquery；walkstrategy；imageretrieval；medicalimage 摘 要 找到与病人具有相似纹理特征的医学图像，有助于医