神经网络数据集的Otsu阈值反投影建模(IJIGSP-V6-N7-7)

I.J.ComputerNetworkandInformationSecurity,2014,7,53-60PublishedOnlineJune2014inMECS()DOI:10.5815/ijcnis.2014.07.07Copyright©2014MECSI.J.Image,GraphicsandSignalProcessing,2014,7,53-60OTSU’sThresholdingwithBackProjectionModelingforNeuralNetworkDataSetsS.AsifHussainDepartmentofECE,A.I.T.S,Rajampet,AndhraPradesh,IndiaE-mail:[e-mail]fECE,R.G.M.C.E.T,Nandyal,AndhraPradesh,IndiaE-mail:[e-mail]E,J.N.T.U.A,Anantapuramu,AndhraPradesh,IndiaE-mail:[e-mail]—ForTrackinginterfacesandshapeswhichdependsontheregionsofpixelintensityisachallengingtaskinimagesegmentation.Manylevelsetmethodshavebeenformulatedforregionbasedandedgebasedmodelsincomputeraideddiagnosissystems.Inordertoprovideaccuratemodelinginvolvingnumericalcomputations,contours,lesionsandbiasvariancewhichoftenrelyonpixelintensityvariationsfortheregionofInterest.Theproposedmethodinvolvestheformulationbyderivingaglobalcriterionfunctionintermsofneighborhoodpixelstorepresentdomainfieldandbiasvariancecharacteristics.Gaussianimpulseisusedforsmootheningsharpedges.Computationalneuralnetworksprovidetheintegralpartofmostlearningalgorithmsasimagesconsistsofredundantattributesofdatawhichhaveredundantnetworkconnectionswithdifferentinputpatternsofsmallweightsformanetworktrainingprocessforminimizingtheenergyandtoestimatethebiasfieldcorrectionforvariousimagingmodalities.ThePETandCTimagesareusedasinputswhichareaffectedwithcancer;inordertoextractthefeatures,proposedmethodisusedforeasydiagnosis.TheresultshowstheimprovedperformancewithNeuralNetworksandprovidesvaluablediagnosticinformation.IndexTerms—Contours,NeuralNetworks,Gaussianimpulse,Redundantattributes.I.INTRODUCTIONThescopeofbiomedicalimagingcoversdataacquisition,imagereconstruction,andimageanalysis,involvingtheories,methods,systems,andapplications.Whiletomographyandpost-processingtechniquesbecomeincreasinglysophisticated,traditionalandemergingmodalitiesplaymoreandmorecriticalrolesinanatomical,functional,cellular,andmolecularimaging.TheoverallgoaloftheInternationalJournalofBiomedicalImagingistopromoteresearchanddevelopmentofbiomedicalimagingbypublishinghigh-qualityresearcharticlesandreviewsinthisrapidlygrowinginterdisciplinaryfield.Imagesofthehumanbodyaregeneratedforclinicaluseormedicalsciencebyamethodcalledmedicalimaging.Normallythemedicalimagingisnotusedtorefertotheprocedureofimagingremovedorgansandtissuesalthoughtheycanbeimagedformedicalreasons.Severalimagingtechniquesarebasedonreconstructinganimageasasetofprojectionsoftheimagedobjectfromdataeitherbydirectinterpretationofdataorsubsequenttocertainpreprocessing.Anactiveresearchfieldinmoderntimesisthetomographyimagereconstruction.Computerizedtomography(CT),positronemissiontomography(PET),andmagneticresonanceimaging(MRI)andseveralothersemployiterativereconstructionasacollectionofmethodsforreconstructingtwodimensionalandthreedimensionalimagesfromtheprojectionsoftheobject.Multipleviews[1]oftheinsideofthebodyaregeneratedbyComputedTomography(CT)scanningusingspecialx-rayequipment’s.Computersystemsareusedtoreconstructthesemultiplex-rayviewsintocross-sectionalimagesofthebody.Filteredback-projection(FBP)isusedforpermittingimagereconstructionofgatedmyocardialperfusionstudiesinallgeneralSPECTcamerasoftwarepackages.Medicalimaginghasbeenundergoingarevolutioninthepastdecadewiththeadventoffaster,moreaccurate,andlessinvasivedevices.Thishasdriventheneedforcorrespondingsoftwaredevelopmentwhichinturnhasprovidedamajorimpetusfornewalgorithmsinsignalandimageprocessing.Manyofthesealgorithmsarebasedonpartialdifferentialequationsandcurvaturedrivenflowswhichwillbethemaintopicsofthissurveypaper.Mathematicalmodels[2]arethefoundationofbiomedicalcomputing.Basingthosemodelsondataextractedfromimagescontinuestobeafundamentaltechniqueforachievingscientificprogressinexperimental,clinical,biomedical,andbehavioralresearch.Today,medicalimagesareacquiredbyarange54OTSU’sThresholdingwithBackProjectionModelingforNeuralNetworkDataSetsCopyright©2014MECSI.J.Image,GraphicsandSignalProcessing,2014,7,53-60oftechniquesacrossallbiologicalscales,whichgofarbeyondthevisiblelightphotographsandmicroscopeimagesoftheearly20thcentury.Modernmedicalimagesmaybeconsideredtobegeometricallyarrangedarraysofdatasampleswhichquantifysuchdiversephysicalphenomenaasthetimevariationofhemoglobindeoxygenationduringneuronalmetabolism,orthediffusionofwatermoleculesthroughandwithintissue.Thebroadeningscopeofimagingasawaytoorganizeourobservationsofthebiophysicalworldhasledtoadramaticincreaseinourabilitytoapplynewprocessingtechniquesandtocombinemultiplechannelsofdataintosophisticatedandcomplexmathematicalmodelsofphysiologicalfunctionanddysfunction.ThispaperprovidesdetailsofaccuratemodelingforimagesegmentationbyinvolvingOTSU’smethodinaccuratediagnosis.FurtherdetailsofneuralnetworkdatasetsareusedindetectionofcancerLesionwhichcanbefoundin[8].II.BACKGROUNDA.ExistingSystemInthisExistingmethodologythemethodproposedbyZhangetal[5]