AppliedSoftComputing22(2014)94–100
ContentslistsavailableatScienceDirect
AppliedSoftComputing

journalhomepage:doc.guandang.net


c
RetinalvesselsegmentationemployingANNtechniquebyGaborandmomentinvariants-basedfeatures
S.WilfredFranklina,S.EdwardRajanb,
ab
CSIInstituteofTechnology,Thovalai,TamilNadu,India
MepcoSchlenkEngineeringCollege,Sivakasi,TamilNadu,India
article
info
abstract
Articlehistory:
Received16March2012
Receivedinrevisedform4July2012Accepted21April2014
Availableonline21May2014
Keywords:
Diabeticretinopathy
RetinalvesselsegmentationArti cialneuralnetworksRetinalimages
Retinalvasculature
Diabeticretinopathy(DR)isthemajorophthalmicpathologicalcauseforlossofeyesightduetochangesinbloodvesselstructure.Theretinalbloodvesselmorphologyhelpstoidentifythesuccessivestagesofanumberofsightthreateningdiseasesandtherebypavesawaytoclassifyitsseverity.Thispaperpresentsanautomatedretinalvesselsegmentationtechniqueusingneuralnetwork,whichcanbeusedincomputeranalysisofretinalimages,e.g.,inautomatedscreeningfordiabeticretinopathy.Furthermore,thealgorithmproposedinthispapercanbeusedfortheanalysisofvascularstructuresofthehumanretina.Changesinretinalvasculatureareoneofthemainsymptomsofdiseaseslikehypertensionanddiabetesmellitus.Sincethesizeoftypicalretinalvesselisonlyafewpixelswide,itiscriticaltoobtainprecisemeasurementsofvascularwidthusingautomatedretinalimageanalysis.Thismethodsegmentseachimagepixelasvesselornonvessel,whichinturn,usedforautomaticrecognitionofthevasculatureinretinalimages.Retinalbloodvesselsareidenti edbymeansofamultilayerperceptronneuralnetwork,forwhichtheinputsarederivedfromtheGaborandmomentinvariants-basedfeatures.Backpropagationalgorithm,whichprovidesanef cienttechniquetochangetheweightsinafeedforwardnetworkisutilizedinourmethod.TheperformanceofourtechniqueisevaluatedandtestedonpubliclyavailableDRIVEdatabaseandwehaveobtainedillustrativevesselsegmentationresultsforthoseimages.
©2014PublishedbyElsevierB.V.
Introduction
Diabeticretinopathy(DR)isaresultoflong-termdiabetesanditisasevereandthemostcommonsightthreateningeyedis-ease,whichcausesblindnessamongworking-agepeoplearoundtheworld[1].MajorvisionlossduetoDRishighlypreventablewithproperscreeningandtimelydiagnosisattheearlierstages.However,DRisnotpainfulandthediabeticpatientsperceivenosymptomuntilvisuallossoccursandhencetheyneedperiodicaleye-fundusexaminationtoensurethattreatmentisreceivedintime.Evaluationofthecharacteristicsofretinalbloodvesselsplaysanimportantroleinthediagnosisofdiseasesbasedonvascularpathology.Thevariousfeaturesofretinalvesselssuchaslength,width,tortuosityandbranchingpatternprovidenewtechniquestodiagnosevariousdiseaseslikediabetes,arteriosclerosis,hyper-tension,cardiovasculardiseaseandstroke.Retinalimagesprovidevaluableinformationrelatedtohumaneye,bywhichthevascular
Correspondingauthor.Tel.:+9104562235308.
E-mailaddresses:swfranklin@gmail.com(S.W.Franklin),sedraj@yahoo.com(S.E.Rajan).
conditioncanbeaccuratelyobservedandanalyzed[2].Theonlypartofthecentralcirculationthatcanbevieweddirectlyandana-lyzedistheretinalvessel.Changesinbloodvesselstructureandvesseldistribution,causedbydiabeticretinopathycanleadtonewvesselgrowth,whichinturninstigatesvisionimpairment.Henceretinalvesselsegmentationbecomesanessentialtoolforthedetec-tionofanyvariationsthatoccursinbloodvessels.Retinalbloodvesselsegmentationgivesthedetailedinformationabouttheloca-tionofvesselswhichhelpsinthescreeningofdiabeticretinopathy,e.g.,inthedetectionofmicro-aneurysmsithelpstoreducethenumberoffalsepositives[3,4].Observationsbasedonretinalves-selsegmentationarequietcomplexandhencethetraditionalwayofdiagnosisinwhichtheophthalmologistidenti estheanomaliespresentintheretinalimagesbyexaminingit,becomestiresomeorevenimpossible.Astheactualsizeofatypicalbloodvesselinhumanretinaisverysmall,justafewpixelswide,measurementofvascularwidthusingretinalimagesegmentationprocessbecomescriticalandchallenging.Oneofthepossiblesolutionsforthesemea-surementsistheuseofcomputerbasedautomatedanalysisofopticfundus,atechniquewidelyacceptedbythemedicalcommunity.Thiscomputerbasedanalysishelpsthemedicalpersonneltodetectthechangesinblood owandvesseldistribution.Italsoenablesthe
doc.guandang.net.
ICPR06_A new approach to automated retinal vessel segmentation using ...Gabor filters and scale production, and a threshold probing technique ...
热门文档