房角

OpticalCoherenceTomography

MONISHAE.NONGPIUR,BENJAMINA.HAALAND,SHAMIRAA.PERERA,DAVIDS.FRIEDMAN,

MINGGUANGHE,LISANDROM.SAKATA,MANIBASKARAN,ANDTINAUNGTodevelopascorealongwithanestimated

probabilityofdiseasefordetectingangleclosurebasedonanteriorsegmentopticalcoherencetomography(ASOCT)imaging.

󰀁DESIGN:Cross-sectionalstudy.

󰀁METHODS:Atotalof2047subjects50yearsofageandolderwererecruitedfromacommunitypolyclinicinSingapore.AllsubjectsunderwentstandardizedocularexaminationincludinggonioscopyandimagingbyASOCT(CarlZeissMeditec).Customizedsoftware(Zhong-shanAngleAssessmentProgram)wasusedtomeasureASOCTparameters.Completedatawereavailablefor1368subjects.Datafromtherighteyeswereusedforanalysis.AstepwiselogisticregressionmodelwithAkaikeinformationcriterionwasusedtogenerateascorethatthenwasconvertedtoanestimatedprobabilityofthepresenceofgonioscopicangleclosure,definedastheinabilitytovisualizetheposteriortrabecularmeshworkforatleast180degreesonnonindentationgonioscopy.󰀁RESULTS:Ofthe1368subjects,295(21.6%)hadgonioscopicangleclosure.TheangleclosurescorewascalculatedfromtheshiftedlinearcombinationoftheASOCTparameters.Thescorecanbeconvertedtoanesti-matedprobabilityofhavingangleclosureusingtherela-tionship:estimatedprobability[escore/(1Descore),whereeisthenaturalexponential.Thescoreperformedwellinasecondindependentsampleof178angle-closuresubjectsand301normalcontrols,withanareaunderthereceiveroperatingcharacteristiccurveof0.94.

󰀁PURPOSE:

󰀁CONCLUSIONS:

AscorederivedfromasingleASOCT

image,coupledwithanestimatedprobability,providesanobjectiveplatformfordetectionofangleclosure.(AmJOphthalmol2014;157:32–38.Ó2014byElsevierInc.Allrightsreserved.)

RIMARYANGLE-CLOSUREGLAUCOMAACCOUNTS

AcceptedforpublicationSep11,2013.

FromtheSingaporeEyeResearchInstituteandSingaporeNationalEyeCenter,Singapore,RepublicofSingapore(M.E.N.,S.A.P.,L.M.S.,M.B.,T.A.);theCentreforQuantitativeMedicine,OfficeofClinicalSciencesDuke-NUSGraduateMedicalSchool,Singapore,RepublicofSingapore(M.E.N.,B.A.H.);DepartmentofStatisticsandAppliedProbability,NationalUniversityofSingapore,Singapore,RepublicofSingapore(B.A.H.);WilmerEyeInstitute,DanaCenterforPreventiveOphthalmology,JohnsHopkinsUniversity,Baltimore,Maryland(D.S.F.);theStateKeyLaboratoryofOphthalmology,ZhongshanOphthalmicCenter,SunYat-senUniversity,Guangzhou,China(M.H.);andtheYongLooLinSchoolofMedicine,NationalUniversityofSingapore,Singapore,RepublicofSingapore(T.A.).

LisandroM.SakataisnowaffiliatedwiththeFederalUniversityofParana,Curitiba,Brazil.

InquiriestoTinAung,SingaporeNationalEyeCenter,11ThirdHospitalAvenue,Singapore168751,RepublicofSingapore;e-mail:aung.tin@snec.com.sg

forapproximately50%ofpeoplewithbilateralglaucomablindnessworldwide,1,2withsignificantimpactonindividuals,societies,andhealthcaresystems.Identifyingpeoplewiththeearlyasymptomaticstagesofthediseasemayallowforprophylactictreatmentbylaseriridotomytopreventvisualloss.Therefore,itisdesirabletoaugmenttheefficiencyandeffectivenessofscreeningpracticesaboveandbeyondthecurrentapproaches.3–6Anteriorsegmentopticalcoherencetomography(ASOCT)technologyenablestheacquisitionandvisualizationofhigh-resolutionimagesoftheanteriorsegment(AS)struc-turesinvivo.7Recently,severalnovelASOCTparametershavebeenassociatedwithangleclosure,includingsmalleranteriorchamberwidth,area,andvolume8,9;greateriristhickness,curvature,andarea10;andalargerlensvault.11,12Inadditiontotheirindividualassociationwithangleclosure,itwasshownrecentlythatmorethan80%ofthevariationinanglewidthisexplainedbythesenewlyidentifiedriskfactors,withlensvault,anteriorchamberarea,andvolumebeingthemostimportantdeterminants.13Werecentlyevaluatedthediagnosticabilityofarangeofbiometric,demographic,andASOCT-basedparameterstodetectangleclosure.Astepwiselogisticregressionmodelwithavariablesetconsistingofjust6ASOCT-derivedparameters—namely,lensvault;anteriorchamberwidth,area,andvolume;iristhickness;andirisarea—hadahighdiagnosticcapabilitywithanareaunderthereceiveroper-atingcharacteristiccurve(AUC)ofmorethan0.95iniden-tifyingeyeswithgonioscopicangleclosure.14Webelievethatthesefindingscanbeusefulindevelopingafeasible,imaging-basedtoolforthedetectionofeyeswithangleclosure.

Thepurposeofthisstudywastodevelopanangle-closurescoreandprobabilityestimatebasedonASOCTimagingthatwouldallowforobjectiveidentificationofeyeswithgonioscopicangleclosure.Asasecondaryaim,wewantedtodissectoutthecontributionofthevariousanatomicfactorsimagedbyASOCTtoidentifythelikelymecha-nism(s)contributingtoangleclosure.

RIGHTSRESERVED.

P

32

Ó

2014BY

ELSEVIERINC.ALL

0002-9394/$36.00

http://dx.doi.org/10.1016/j.ajo.2013.09.012

METHODS

APPROVALFORTHESTUDYWASGRANTEDBYTHE

SingaporeEyeResearchInstituteInstitutionalReviewBoard.ThestudywasconductedinaccordancewiththetenetsoftheDeclarationofHelsinki,andwritteninformedconsentwasobtainedfromallsubjectsbeforeenrolmentintothestudy.Analysiswasperformedondatacollectedfromacommunity-basedcross-sectionalstudyperformedinSingapore,thedetailsofwhichhavebeendescribedpre-viously.6Inbrief,thesubjectsforthestudywererecruitedfromagovernment-runcommunity-basedclinicfacilitythatprovidesprimaryhealthcareservicesforlocalresi-dents.Thestudyparticipantswere50yearsofageandolder.

󰀁EXAMINATIONANDDEFINITIONOFGONIOSCOPICANGLECLOSURE:Allstudyparticipantsunderwentastan-

dardizedeyeexaminationthatincludedmeasurementsofvisualacuityusingalogarithmoftheminimalangleofreso-lutionchart(Lighthouse,Inc,LongIsland,NewYork,USA),slit-lampexamination(ModelBQ900;Haag-Streit,Bern,Switzerland),stereoscopicopticdiscexaminationwitha78-diopterlens(VolkOptical,Inc,Mentor,Ohio),measure-mentofintraocularpressurebyGoldmannapplanationtonometry(Haag-Streit,Koniz,Switzerland),andgonio-scopyperformedinthedarkusingaGoldmann2-mirrorlensathighmagnification(316)byasingleexaminer.Aneyewasconsideredtohavegonioscopicangleclosureiftheposteriorpigmentedtrabecularmeshworkwasnotvisibleforatleast180degreesonnonindentationgonio-scopywiththeeyeintheprimaryposition.Axiallengthandanteriorchamberdepth(ACD)weremeasuredusingpartialcoherencelaserinterferometrybyIOLMaster(CarlZeissMeditec,Dublin,California,USA).

󰀁ANTERIORSEGMENTOPTICALCOHERENCETOMOGRAPHYIMAGING:ASOCTimageswereacquiredfromallparticipants

(inmillimeters),8andlensvault11wastheperpendiculardis-tancebetweentheanteriorpoleofthecrystallinelensandthehorizontallinejoiningthe2scleralspurs(inmicrometers).Theiristhicknesswasmeasuredat750and2000mmfromthescleralspur.10Theirisareawascalculatedasthecross-sectionalareaofthefulllength(fromspurtopupil)oftheiris(insquaremillimeters).Todetermineiriscurvature,thesoftwaredrawsalinefromthemostperipheraltothemostcentralpointsoftheirispigmentepithelium.Aperpendicularlinethenisextendedfromthislinetotheirispigmentepitheliumatthepointofgreatestconvexity.Thelengthofthisperpendicularlineisameasureoftheiriscur-vature(inmillimeters).10Theanteriorchamberareawasdefinedasthecross-sectionalarea(insquaremillimeters)ofASboundedbythecornealendothelium,theanteriorsur-faceoftheiris,andtheanteriorsurfaceofthelens(withinthepupil).9Toassessanteriorchambervolume,thealgo-rithmfirstplotsaverticalaxisthroughthemidpoint(center)oftheanteriorchamberarea,andbyrotatingtheanteriorchamberarea360degreesaroundthisverticalaxis,theante-riorchambervolume9wascalculated(incubedmillimeters).

󰀁STATISTICALANALYSIS:

indarkroomconditionsbyasingleoperator.Eachscanwascenteredonthepupil,andasinglecross-sectionalhorizontalscan(nasal-temporalanglesat0degreesand180degrees)wasevaluatedforeachsubject.Thebest-qualityimageswereobtainedbyadjustingthesaturationandnoiseandbyoptimizingthepolarizationforeachscanduringtheexamina-tion.Acustomizedsoftware,theZhongshanAngleAssessmentProgram15(Guangzhou,China),wasusedtoprocesstheASOCTimages.Foreachimage,theonlyobserverinputwastodeterminethelocationofthe2scleralspurs,whichisdescribedastheinwardprotrusionofthesclerawithachangeincurvatureofitsinnersurface.Thealgorithmthenautomaticallycalcu-latedthevariousASparameters.ThereproducibilityoftheASparameterswasexcellent,withtheintraclasscorrela-tioncoefficientrangingfrom0.88to0.97.8,10,12󰀁DEFINITIONOFANTERIORSEGMENTOPTICALCOHERENCETOMOGRAPHYPARAMETERS:Theparametersasmeasuredby

Developmentoftheangleclosure

scoreandestimatedprobability.Datafromrighteyeswereusedforanalysis.Afteranevaluationofthediagnosticaccu-racyofseveralclassificationalgorithmstodetectgonioscopicangleclosure,wefoundthatthestepwiselogisticregressionmodelwithAkaikeinformationcriterionconsistingofonly6ASOCTparameterswasboththemostaccurateandleastvariable.14Akaikeinformationcriterionmeasuresboththeaccuracyandcomplexityofamodel,andinthecontextofagivenproblem,amodelwithlowerAkaikeinformationcriterionoftenwillhavebettergeneralizability.Stepwiselogisticregressionwasbegunwithnoneofthe6variablesinthemodel.Ateachstage,thepossibilitiesof(1)addingeachvariablenotyetincluded,(2)removingeachofthecurrentlyincludedvariables,and(3)leavingthemodelunchangedwereconsideredinturn.TheactionwithminimumAkaikeinformationcriterionwastakenwiththealgorithmterminatingwhenleavingthemodelunchangedhadminimumAkaikeinformationcriterion.Thelogisticregressionyieldsascorethatisashiftedlinearcombinationoftheselectedvariables.Thisscorecanbeconvertedtoanestimatedprobabilityofangleclosureusingtherelationship:estimatedprobability¼escore/(1þescore),whereeisthenaturalexponential.

Determinationoftherelativecontributionsoftheanteriorsegmentopticalcoherencetomographyparametersandtheirgraphicalrepresentation.Anassessmentoftherelativecon-tributionsofthevariablestotheangleclosurescoreforagivensubjectwasachievedfirstbystandardizingeachofthevariables,sothateachhadameanof0andastandarddeviationof1.Thiswasachievedbysubtractingeachvari-able’smeanandthendividingeachvariablebyitsstandarddeviation.Then,theshiftedlinearcombinationofselected

33

ASOCT(Figure1)weredefinedasfollows.Anteriorchamberwidthwasdefinedasthehorizontalscleralspur-to-spurdistanceVOL.157,NO.1

ANAUTOMATEDSCOREFORDETECTIONOFANGLECLOSURE

TABLE1.MeansandStandardDeviationsfrom1368Angle-ClosureandNormalSubjectsUsedforStandardizationoftheAnteriorSegmentOpticalCoherenceTomographyParametersintheScoreforDetectingAngleClosureVariable

Mean

StandardDeviation

FIGURE1.Illustrationoftheparametersusedinthescorefordetectingangleclosureasmeasuredfromanteriorsegmentopti-calcoherencetomographyimagesaccordingtotheZhongshanangleassessmentprogram.Boldarrowsindicatethelocationofthescleralspurs.ACA[anteriorchamberarea;ACV[ante-riorchambervolume;ACW[anteriorchamberwidth;IT750[iristhicknessat750mmfromthescleralspur;LV[lensvault.

Anteriorchambervolume(mm3)Anteriorchamberwidth(mm)Iristhickness(mm)Irisarea(mm2)

Anteriorchamberarea(mm2)Lensvault(mm)

132.2311.760.451.6019.8547030.650.390.080.223.288

variablesthatconstitutesthescorewasre-expressedintermsofthestandardizedvariables.Foraparticularsubjectandtheirstandardizedvariablevalues,theproductofeachstandardizedvariablevalueanditsassociatedcoefficientrepresentedtherelativecontributionofthatvariabletoahighorlowangle-closurescore.Acrossallsubjects,thesumofthesecontributionswasapproximately0.Apositiverelativecontributionindicatedapushtowardhigherlikelihoodofangleclosure,whereasanegativerelativecontributionindicatedapushtowardalowerlikelihoodofangleclosure.

Foreachsubject,theirASOCT-derivedparameterswereenteredintotheformulafortheangleclosurescore,andthecorrespondingscorewascalculated.ThesescoreswereanalyzedalongsidetheknownangleclosurestatustocalculatetheAUC,byvaryingthediagnosticthresholdfromÀNtoNtoobtainallpossiblecombinationsofsensitivityandspecificity.Toaidintheidentificationofimportantvariablesassoci-atedwithangleclosureforaparticularpatient,standardizedparameterswereused,andthesearedenotedwithanasteriskasanteriorchambervolume*,anteriorchamberwidth*,iristhicknessat750mm*,irisarea*,anteriorcham-berarea*,andlensvault*(Table1).Theformulaforthescorethenwasre-expressedintermsofthestandardizedvar-iables.Foraparticularsubject,thevaluesoftheproductoftheirstandardizedvariablesandthecorrespondingcoeffi-cientsinthere-expressedscorecanbeplottedtoassesstherelativecontributionsofeacheyebiometrymeasure-menttotheangle-closurescore.Becauseanteriorchamberareaandanteriorchambervolumearecorrelatedstrongly,theireffectsarecloselylinked.Toenableinterpretation,thevalueofthesumofthecorrespondingproductscanbeplottedtoassessthejointeffectofanteriorchamberareaandvolume.Ifaparticularcomponentofthescoreislargerelativetotheothercomponents,thenthecorrespondingvariableisasubstantialcontributingfactortotheirrisk.34

Validationoftheangleclosurescore.Tovalidatetheangleclosurescore,wetesteditonanindependentgroupconsist-ingof178hospital-basedsubjectswithgonioscopicangleclosure(asdefinedpreviously)recruitedaspartofanongoingrandomizedcontrolledtrial(NCT00347178,clinicaltrials.gov)16and301normalcontrolsubjectswhowererecruitedfromanongoingpopulation-basedstudy.17Thecriteriafornormalswereanintraocularpressureoflessthan21mmHgwithopenangles,healthyopticnerves,normalvisualfields,andnopreviousintraocularsurgery.ASOCTimagingwasperformedbeforelaserperipheraliridotomyintheangle-closuresubjects.

RESULTS

ATOTALOF2047SUBJECTSWERERECRUITEDINTOTHE

study,ofwhom679(33.2%)wereexcludedforthefollowingreasons:11(0.54%)couldnotundergogonio-scopy,97(4.7%)hadincompletedemographicorbiometricmeasurements,62(3.0%)couldnotcompletetheASOCTexaminationorhadpoorqualityASOCTimages,42(2.1%)hadsoftwaredelineationerrors,and467(22.8%)hadscleralspursthatwerenotclearlyvisibleonASOCTimages.Therefore,datafrom1368subjects(66.8%)wereincludedintheanalysis.Ofthese,750(54.8%)werewomen,and1232(90.0%)wereChinese.Themeanage6standarddeviationwas62.467.7years.Therewerenosignificantdifferencesinrace,sphericalequivalentrefraction,intraocularpressure,oraxiallength,butincludedparticipantswereyounger(P<.001),hadshallowerACD(P¼.02),andweremorelikelytobewomen(P¼.03).Therewere295subjects(21.6%)withgonioscopicangleclosure.Personswithangleclosurewereolder;hadsmallerACD;hadsmalleranteriorchamberwidth,area,andvolume;hadgreaterlensvault;andhadthickerirides(Table2).FindingsweresimilarwhenonlyChineseper-sonswereanalyzed.

󰀁ANGLECLOSURESCOREANDESTIMATEDPROBABILITY:

TheshiftedlinearcombinationoftheASOCT-derived

JANUARY2014

AMERICANJOURNALOFOPHTHALMOLOGY

TABLE2.SummariesandComparisonsbetweenAngle-ClosureandNormalSubjectsofDemographicandAnteriorSegmentOpticalCoherenceTomographyParametersPotentiallyUsefulinDetectingAngleClosureAngleClosure(n¼295)

Normal(n¼1073)

PValue

Age(y)

MalegenderEthnicityChineseOtherACD(mm)AL(mm)IT750(mm)IT2000(mm)IArea(mm2)ICurv(mm)ACW(mm)ACA(mm2)ACV(mm3)LV(mm)

SE(diopters)

.56(7.42)34.6%96.3%3.7%2.68(0.22)23.07(0.88)0.47(0.08)0.48(0.07)1.61(0.21)0.39(0.09)11.57(0.37)15.66(1.90)97.70(15.30)774(188)þ2.04(1.78)

62.76(7.43)48.1%88.3%11.7%3.20(0.33)24.15(1.36)0.45(0.07)0.47(0.07)1.60(0.23)0.27(0.12)11.81(0.37)21.00(3.11)141.70(26.73)386(252)þ0.37(2.78)

<.0001<.0001<.0001

smalleranteriorchamberareaandvolumeandgreaterlensvault.Thebottomrightplotwithanestimatedproba-bilityof0.45illustratesanequivocalsituation.Atalowerdiagnosticthreshold,whichachievesahighersensitivity,thissubjectwouldhavebeenclassifiedashavingangleclosure,butahigherdiagnosticthresholdindicatesthatthesubjecthasnoangleclosure.

Theestimatedproba-bilityandtherelativecontributionofthevariousparame-tersalsowereinvestigatedinthevalidationdatasetofhospital-basedsubjectswithangleclosureandnormalcon-trols.Ofthe178hospital-basedsubjectswithangleclosure,thescleralspurcouldnotbeidentifiedin2,leaving176(98.8%)subjectsinthefinalanalysis.Onehundredthirty-fivesubjects(76.7%)werewomen,and168wereChinese(95.5%).Themeanageofthesubjectswas63.167.4years.Meanageofthecontrolswas56.069.2years.TheAUCoftheangleclosurescoreasestimatedbythisvalidationdatasetwas0.935(95%confidenceinterval,0.913to0.957).Todemonstratetherelativecontributionsofthevariables,weselected3representativesubjectsfromamongthe176angle-closuresubjectsandplottedtheirpro-fileswiththecorrespondingASOCTimages(Figure3).Byfarthelargestcontributorwassmalleranteriorchamberareaandvolume(all3rows).AsdepictedinFigure3(MiddlerowandBottomrow),largerlensvaultandsmalleranteriorchamberwidth(Figure3,Middlerow)andgreateriristhickness(Figure3,Bottomrow)hadlessimpactthananteriorchamberareaandvolume.Theestimatedproba-bilityrangedfrom0.50to1.

󰀁VALIDATIONINSECONDSAMPLE:

<.0001<.0001<.0001.041.51<.0001<.0001<.0001<.0001<.0001<.0001

ACA¼anteriorchamberarea;ACD¼anteriorchamberdepth;ACV¼anteriorchambervolume;ACW¼anteriorchamberwidth;AL¼axiallength;IArea¼irisarea;ICurv¼iriscurvature;IT¼iristhickness;LV¼lensvault;SD¼standarddeviation;SE¼sphericalequivalent.Dataaremean(SD).parametersusedtocalculatetheangle-closurescorewasgivenby:

Score¼À28.986879À0.3399103(anteriorchambervolume)þ3.2235063(anteriorchamberwidth)þ7.2966543(iristhicknessat750mm)À2.2028243(irisarea)þ1.5345223(anteriorchamberarea)þ0.003242(lensvault)

Theestimatedprobabilitywascomputedasescore/(1þescore).Toachievespecificityof0.96andsensitivityof0.75,theestimatedprobabilitythresholdfordiagnosisofangleclosurewasdeterminedtobe0.50.Atalowerdiag-nosticthresholdof0.26,thespecificitydropsto0.,whereasthesensitivityrisesto0..

󰀁RELATIVECONTRIBUTIONSOFTHEANTERIORSEGMENTOPTICALCOHERENCETOMOGRAPHYPARAME-TERS:Asexamples,werandomlyselected6subjectswhose

DISCUSSION

THEPRESENTSTUDYREPRESENTSANEWAPPROACHTOTHE

profilesareplottedinFigure2.Depictedforeachsubjectaretheestimatedprobability,gonioscopicangle-closurestatus,andrelativecontributionsofthevariables.Positivevaluesintheplotsindicateanaffirmativecontributionofthevariabletothescore,whereasnegativevaluesindicateanegativecontributionofthevariabletothescore.Ascanbeseeninthefigure,theestimatedprobabilityforthosewithoutangleclosurewaslowerthanforthosewithangleclosure(asonewouldexpect).Additionally,forthosewithangleclosure,thelargestcontributorswereoftenVOL.157,NO.1

developmentofatoolfordetectingangleclosurewithASOCTimaging.Basedprimarilyonanatomicfeaturesoftheeye,theangle-closurescoringsystemcanbeintegratedeasilyintoASOCTimageanalysissoftwaretoalertclini-cianstothepresenceofgonioscopicangleclosure.Thegraphicaloutput,whichrepresentstherelativecontribu-tionsofthedifferentvariablestothescore,providesanobjectiveassessmentofimportantanatomicfactorsintheeye,whichcanaidindecidingontreatmenttargetingspe-cificfactors.Forexample,forsubjectswithasignificantlensvault,whichisnoteasilydiscernibleclinically,thealgo-rithmmayhighlightthosecasesthataremorelikelytobenefitfromcataractextraction,althoughdatasupportingthisarelimited.18Gonioscopy,thereferencestandardfordiagnosingangleclosure,islimitedasascreeningtoolbecauseitrequiresconsiderableexpertise,isdependentonsingle-observerinterpretation,andissubjectedtovariabilityinducedbythetypeoflensandilluminationconditions.4,19,2035

ANAUTOMATEDSCOREFORDETECTIONOFANGLECLOSURE

FIGURE2.Profileplotsof6subjectsillustratingtheestimatedprobabilityofangleclosurebasedonthescorefordetectingangleclosure,actualgonioscopicangle-closurestatus,andtherelativecontributionsoftheanteriorsegmentopticalcoherencetomographyvariablesusedinthescore.Positivevaluesintheprofileplotindicateanaffirmativecontributionofthevariabletothepatient’sscore,whereasnegativevaluesindicateanegativecontributionofthevariabletothepatient’sscore.Darkershadedplotsindicateactualgonioscopicangleclosureandlightershadedplotsindicateactualopenanglesornogonioscopicangleclosure.ACA[anteriorchamberarea;ACV[anteriorchambervolume;ACW[anteriorchamberwidth;IArea[irisarea;IT750[iristhicknessat750mmfromthescleralspur.

WhereastheperformanceofACDandlimbalACDasascreeningtoolforangleclosureissatisfactorywithanAUCvaryingbetween0.81and0.86,6,9,21thecurrentangle-closurescoreincorporatingASOCT-derivedparam-etersfaredbetter,withanAUCof0.94to0.97.Anaddedadvantageofthescoreistheabilitytoassessgraphicallytherelativecontributionsofthevariablestothescore.Mea-surementsofanglewidthsuchastrabecular–irisspaceareaandangleopeningdistancearerelativelysmallandareinfluencedbytheiriscontour22;theythereforearesub-jecttogreatervariationaroundthecircumferenceoftheeye.Byincorporatingnonangleparameters,theformuladerivedinthisstudyislesslikelytovaryacrossmeridionalscans,andthereforemoreeffectivelysummarizestheASOCTfindings.However,animportantlimitationofthecurrentstudyisthatalmostonequarteroftheASOCTimageswerenotgradablebecauseofthepoorvisualizationofthescleraspur.

Byusingcustomizedanalysissoftware,suchastheZhongshanAngleAssessmentProgram,15theAnteriorSegmentAnalysisProgram,23orimageJplatforms,thevariousASOCT-basedparameterscanbeobtainedeasilyfromasingleASOCTscanaftermanualidentificationofthescleralspurs.ThereproducibilityofASOCTparame-tershasbeenwellstudied,withinterclassandintraclasscorrelationcoefficientsrangingfrom0.88to0.97.8,10,24TheASOCT-basedangle-closurescoringsystemwedescribeisobjectiveandcanbeoperatedbyatechnician.36

Thesefactors,inadditiontoeaseofoperationandrapidimageacquisition,maketheASOCTappealingasapotentialtoolforscreeningforangleclosure,althoughthehighcostoftheinstrumentmaylimitthefeasibilityofusingthisdeviceforpopulation-basedscreening.Becausethediagnosisthresholdcanbevaried,thescoringsystemcanbeappliedforeitheruniversalpopulation-basedscreeningwherehighspecificitylevelsarepreferable,orforopportunisticclinic-basedcaseidentification,whereahighersensitivityismoreappropriate.Thepositivelyiden-tifiedsubjectsthencanundergofurtherdiagnosticinvesti-gationsandevaluationsbyaglaucomaspecialist.Targetedoropportunisticscreeningofat-riskpopulationssuchasolderindividuals,females,peopleofAsianancestry,andthosewithafamilyhistoryofprimaryangle-closureglau-comacanincreasecasedetectionrates.

Asnotedabove,themainlimitationsofourstudyweretheuseofsemiautomatedimageanalysissoftware,whichrequiredmanuallocalizationofthescleralspur,andtheexclusionofapproximatelyonequarteroftheimagesmainlybecauseoftheinabilitytoidentifythescleralspur.Useofhigher-resolutionimagesandfullyautomatedimageanalysissoftwaremayeliminatethisissue.BecausemostofoursubjectswereChinese,cautioniswarrantedwhentryingtoextrapolatethefindingstootherethnicgroups.

Insummary,ourstudydescribesanangle-closurescorefordetectingthepresenceofgonioscopicangleclosure,derivedfromASOCT-basedparameters.Afterbeing

JANUARY2014

AMERICANJOURNALOFOPHTHALMOLOGY

FIGURE3.Illustrationoftheanteriorsegmentopticalcoherencetomographyimagesandthecorrespondingprofileplotsof3subjectsdemonstratingtherelativecontributionsofthevariablesusedinthescorefordetectingangleclosure.Theestimatedprobabilityofangleclosurebasedonthescorefordetectingangleclosureandgonioscopicangleclosurestatusareindicatedwithineachprofileplot.Positivevaluesintheprofileplotindicateanaffirmativecontributionofthevariabletothepatient’sscore,whereasnegativevaluesindicateanegativecontributionofthevariabletothepatient’sscore.Thelargestcontributorwassmalleranteriorchamberareaandvolume(all3rows),whereasalargerlensvaultandsmallerACW(Middlerow)andgreateriristhickness(Bottomrow)hadasmallerimpact.ACA[anteriorchamberarea;ACV[anteriorchambervolume;ACW[anteriorchamberwidth;IArea[irisarea;IT750[iristhicknessat750mmfromthescleralspur.

integratedintoASOCTdevices,thismayleverageonitsstrengthsofbeingsimple,objective,andclinicianindepen-dentandmayhavepotentialuseasascreeningtoolforangleclosure.Moreover,thealgorithmoutputisabletocategorizetheimportanceofdifferentanatomicfactorssuchasthelensvaultoranteriorchamberarea.

ALLAUTHORSHAVECOMPLETEDANDSUBMITTEDTHEICMJEFORMFORDISCLOSUREOFPOTENTIALCONFLICTSOFINTERESTandthefollowingwerereported.Dr.Aunghasreceivedresearchfunding,travelsupport,andhonorariafromCarlZeissMeditec.DrFriedmanhasreceivedaninstrumentloanfromCarlZeissMeditec.SupportedinpartbygrantsfromtheNationalMedicalResearchCouncil,Singapore,andtheNationalResearchFoundation,Singapore,RepublicofSingapore.InvolvedinDesignandconductofstudy(M.E.N.,B.A.H.,D.S.F.,T.A.);Datacollection(M.E.N.,L.M.S.,M.B.);Statisticalanalysisandinterpretation(B.A.H.,M.E.N.,D.F.,T.A.);Manuscriptpreparationandapproval(M.E.N.,B.A.H.,S.A.P.,D.S.F.,M.H.,L.M.S.,M.B.,T.A.);andProvisionofpatientsorresources(S.A.P.,M.H.,M.B.,T.A.).Theangleclosurecasesofthevalidationgroupwererecruitedfromanon-goingrandomizedcontrolledtrialregisteredatclinicaltrials.govundernumberNCT00347178.

VOL.157,NO.1ANAUTOMATEDSCOREFORDETECTIONOFANGLECLOSURE37

REFERENCES

1.QuigleyHA,BromanAT.Thenumberofpeoplewithglau-comaworldwidein2010and2020.BrJOphthalmol2006;90(3):262–267.

2.HuZ,ZhaoZL,DongFT.AnepidemiologicalinvestigationofglaucomainBeijingandShun-Yicounty.ChinJOphthalmol19;25:115–118.

3.CongdonNG,QuigleyHA,HungPT,WangTH,HoTC.Screeningtechniquesforangle-closureglaucomainruralTaiwan.ActaOphthalmolScand1996;74(2):113–119.

4.FosterPJ,DevereuxJG,AlsbirkPH,etal.Detectionofgonio-scopicallyoccludableanglesandprimaryangleclosureglau-comabyestimationoflimbalchamberdepthinAsians:modifiedgradingscheme.BrJOphthalmol2000;84(2):186–192.5.DevereuxJG,FosterPJ,BaasanhuJ,etal.Anteriorchamberdepthmeasurementasascreeningtoolforprimaryangle-closureglaucomainanEastAsianpopulation.ArchOphthalmol2000;118(2):257–263.6.LavanyaR,FosterPJ,SakataLM,etal.ScreeningfornarrowanglesintheSingaporepopulation:evaluationofnewnoncon-tactscreeningmethods.Ophthalmology2008;115(10):1720–1727.7.RadhakrishnanS,RollinsAM,RothJE,etal.Real-timeopticalcoherencetomographyoftheanteriorsegmentat1310nm.ArchOphthalmol2001;119(8):1179–1185.

8.NongpiurME,SakataLM,FriedmanDS,etal.Novelassoci-ationofsmalleranteriorchamberwidthwithangleclosureinSingaporeans.Ophthalmology2010;117(10):1967–1973.

9.WuRY,NongpiurME,HeMG,etal.Associationofnarrowangleswithanteriorchamberareaandvolumemeasuredwithanteriorsegmentopticalcoherencetomography.ArchOphthalmol2011;129(5):569–574.

10.WangBS,SakataLM,FriedmanDS,etal.Quantitativeiris

parametersandassociationwithnarrowangles.Ophthal-mology2010;117(1):11–17.

11.NongpiurME,HeMG,AmerasingheN,etal.Lensvault,

thicknessandpositioninChinesesubjectswithangleclosure.Ophthalmology2011;118(3):474–479.

12.TanGS,HeM,ZhaoW,etal.Determinantsoflensvaultand

associationwithnarrowanglesinpatientsfromSingapore.AmJOphthalmol2012;154(1):39–46.

13.FooLL,NongpiurME,AllenJC,etal.Determinantsofangle

widthinChineseSingaporeans.Ophthalmology2012;119(2):278–282.

14.NongpiurME,HaalandBA,FriedmanDS,etal.Classifica-tionalgorithmsbasedonanteriorsegmentopticalcoherencetomographymeasurementsfordetectionofangleclosure.Ophthalmology2013;120(1):48–54.

15.ConsoleJW,SakataLM,AungT,FriedmanDS,HeM.

Quantitativeanalysisofanteriorsegmentopticalcoherencetomographyimages:theZhongshanAngleAssessmentProgram.BrJOphthalmol2008;92(12):1612–1616.

16.HowAC,BaskaranM,HeM,etal.Changesinanterior

segmentmorphologyafterlaserperipheraliridotomy:ananteriorsegmentopticalcoherencetomographystudy.Ophthalmology2012;119(7):1383–1387.

17.LavanyaR,JeganathanVS,ZhengY,etal.Methodologyof

theSingaporeIndianChineseCohort(SICC)eyestudy:quantifyingethnicvariationsintheepidemiologyofeyedis-easesinAsians.OphthalmicEpidemiol2009;16(6):325–336.18.Azuara-BlancoA,BurrJM,CochranC,etal.Effectivenessin

Angle-closureGlaucomaofLensExtraction(EAGLE)StudyGroup.Theeffectivenessofearlylensextractionwithintra-ocularlensimplantationforthetreatmentofprimaryangle-closureglaucoma(EAGLE):studyprotocolforarandomizedcontrolledtrial.Trials2011;12:133.

19.FosterPJ,OenFT,MachinD,etal.Theprevalenceofglau-comainChineseresidentsofSingapore:across-sectionalpopulationsurveyoftheTanjongPagardistrict.ArchOphthalmol2000;118(8):1105–1111.

20.HeM,FosterPJ,GeJ,etal.GonioscopyinadultChinese:the

LiwanEyeStudy.InvestOphthalmolVisSci2006;47(11):4772–4779.

21.NolanWP,AungT,MachinD,etal.Detectionofnarrow

anglesandestablishedangleclosureinChineseresidentsofSingapore:potentialscreeningtests.AmJOphthalmol2006;141(5):6–901.

22.RadhakrishnanS,HuangD,SmithSD.Opticalcoherence

tomographyimagingoftheanteriorchamberangle.Ophthal-molClinNorthAm2005;18(3):375–381,vi.

23.ZhengC,CheungCY,NarayanaswamyA,etal.Pupil

dynamicsinChinesesubjectswithangleclosure.GraefesArchClinExpOphthalmol2012;250(9):1353–1359.

¨rksenE,etal.Anteriorchamber24.Mu¨llerM,DahmenG,Po

anglemeasurementwithopticalcoherencetomography:intraobserverandinterobservervariability.JCataractRefractSurg2006;32(11):1803–1808.

38AMERICANJOURNALOFOPHTHALMOLOGYJANUARY2014

Biosketch

MonishaEstherNongpiur,MDisaSeniorClinicalResearchFellowattheSingaporeEyeResearchInstitute.ShecompletedherOphthalmologyresidencytrainingattheAllIndiaInstituteofMedicalSciences(AIIMS),NewDelhi,India.Shehasauthoredseveralclinicalresearchpapersinthefieldofglaucomaandherresearchinterestsincludeocularimagingandophthalmicgenetics.

VOL.157,NO.1ANAUTOMATEDSCOREFORDETECTIONOFANGLECLOSURE38.e1