retinopathy of prematurity is a potentially blinding：早产儿视网膜病变是一种潜在的致盲

1/1retinopathyofprematurityisapotentiallyblinding：早产儿视网膜病变是一种潜在的致盲RetinopathyofPrematurityPulseOximeterAnalysisSoftwarePayalSehgalTrentRoszellJeanPappalardoAdvisors:Dr.JohnPennDr.KendallGrahamDr.PaulKingRELEVANTTERMSROP:RetinopathyofPrematuritySa02:Thebloodoxygenleveloftheinfant.FiO2:Fractionofinspiredoxygen.Thepercentageofoxygenadministeredtotheinfantinaclinicalsetting.RGM:RespiratoryGasMonitor,PulseOximeter,weusedDatexOhmedaABSTRACTRetinopathyofPrematurity(ROP)alsoknownasRetrolentalFibroplasiaisadiseasethataffectsthevasculatureofretinasininfantsandlowbirthweightnewborns.Whenaninfantisbornprematurely,hisorherretinaisincompletelyvascularizedandasaconsequence,abnormalbloodvesselsmaydevelopwhichcanleadtoleaking,scartissueformation,andretinaldetachment.ItisexcessivefluctuationinSaturatedOxygen(SaO2)levelsintheinfant,especiallyafterthefirstweekofbirth,whichisstronglylinkedwiththeonsetofblindnessinlaboratorymice.Duetotheseresultsfromtheanimalmodels,controlofoxygendeliverytoventilatednewbornshasbecomeapriorityinaneonatalintensivecare.WedecidedtohelpnursesoutbydevelopingsoftwarethatwillreadinandrecognizetrendsinSaO2levelsandthensuggestacorrelatinginputoxygen(FiO2)levelfortheinfant.OuranalysisofthefiveprematureinfantdatasetsgatheredbythepulseoximetershowedthatidealFiO2levelsvarygreatlyfromoneinfanttothenext,andthereisnoclearcutequationtorelateFiO2levelswithSaO2levels.Ourmajordesigncriteriawastodevelopanoxygenmonitoringsystem,whichcouldtakeindatafromapulseoximeterandmakearecommendedFiO2changeinordertohelpdecreaseSaO2variability.WealsowantthissystemtobeabletomakeitscalculationsatanyFiO2inputlevel.OurprogramdoesindeedreachthesedesigncriteriausingaseriesofBooleanlogic.Theendresultofourprogrammingisafrontpanelthatisremarkablyeasytounderstand.ItprovidesataglanceinformationaboutthebabysSa02level,thedifferencebetweenitandthedesiredlevel,aswellascategorizationinoneoffivelevelsofdistress.INTRODUCTIONRetinopathyofPrematurity(ROP)alsoknownasRetrolentalFibroplasiaisadiseasethataffectsthevasculatureofretinasininfantsandlowbirthweightnewborns(1).ROPcanbemildwithnovisualdefectoritcanbesevereandleadtoretinaldetachmentandeventuallylifelongblindness.Itisreportedthatoutofthe3.9millionbabiesborneachyear,11.5%arebornprematurely,thatisapproximately1inevery8childrenarebornprematurely(2).ROPdevelopsin84%ofprematurebabiesbornat28weeksgestation(1).Theyounger(28weeksgestation)andsmaller(1250grams)thebaby,themoreimmaturetheretinalvessels,thehigherincidenceofROP(1).Whenaninfantisbornprematurely,hisorherretinaisincompletelyvascularizedandasaconsequence,abnormalbloodvesselsmaydevelopwhichcanleadtoleaking,scartissueformation,andretinaldetachment.Inthe1940sand1950sprematureinfantswereplacedonoxygentanksduetotheirweaklungs.In1942ROPwasfirstreported(3).In1951apediatricianinAustraliadiscoveredthelinkbetweenoxygenandROP(3).TherehavebeenseveralstudiesfollowingthediscoverythathaveshownarelationshipbetweentheadministrationofsupplementaloxygenandthestartofROP.Dr.JohnPennofVanderbiltUniversityMedicalCenterhasshownthatratsplacedinincubatorsinwhichoxygenadministrationvariedgreatlyalmostalways-developedROP.Infact,hisstudiesshowthatfluctuatingoxygenlevelsinthebodyweremoredamagingthanconstantexposure,evenwhenthecombinedoxygendeliveredtotheratwassubstantiallyless(3).ItisexcessivefluctuationinSaturatedOxygenlevels(SaO2),especiallyafterthefirstweekofbirth,whichisstronglylinkedwiththeonsetofblindnessinlaboratorymice(1).Duetotheseresultsfromtheanimalmodels,controlofoxygendeliverytoventilatednewbornshasbecomeapriorityinaneonatalintensivecare.Neonateoxygencontrol,atthepresenttime,islargelyamanualoperation,withanursemonitoringtheoutputofapulseoximeter(figure1),andadjustingtheinputoxygendelivery(theFiO2level)unittoyieldadesiredSaO2reading.OurstudyofthecurrenttreatmentpracticesledustotheVanderbiltUniversityMedicalCenterNICU.Dr.Graham,aneonatologistwhoiscurrentlyintheVanderbiltfellowshipprogram,explainedthebasicmanneroftreatmentofprematureinfants.Newborns,especiallythosebornatlessthan28weeksgestation,areextremelyfragileandhavemanyunderdevelopedorgans.Thetwomostrelevantofthesearethenewbornslungsandeyes.Thesebabiesareplacedinincubators,whichareenvironmentallycontrolledunitlargeenoughtosafelyholdtheprematureinfant,andfedinsupplementaloxygen.Pulseoximetersarethenusedtomonitortheoxygenlevelsoftheinfant.AlarmsaresetatminimumandmaximumleveloxygenintheoximetertoalertthenursesthattheinfanteitherneedsanincreaseordecreaseinFiO2administration.Theselevelsaremostoftenmeasuredbyapulseoximeter,adeviceweareusinginourproject.TherearetwomethodsofmeasuringtheSaO2levelofthenewborn.Itispossibletomeasurebloodoxygenwithininvasiveprobe,placedinthesubjectsbloodstream(4).Thismethodisalmostneverused,asitispainful,dangerous,andcanpotentiallycomplicateanewbornsalreadytenuoussituation.Theuniversallyacceptedwayofmeasuringtheinfantsbloodoxygenisthroughadeviceknownasapulseoximeterorarespiratorygasmonitor(RGM)(figure1).Pulseoximetershavemanyuniquefeatures.Theyarenoninvasive,needingonlyacleanpatchofskintogatheraccurateSaO2readings.Pulseoximetersaremostoftenattachedtofingersorthetoesthroughaplethysmograph(figure2),aninstrumentthatmeasuresvariationsinthesizeofafingerpartonFigure1:PulseOximeterthebasisoftheamountofbloodpassingthroughorpresentinthepart,asthisareaisblood-richandrelativelythin.Pulseoximetersmeasurebloodoxygenbytransmittingredandinfraredlightthroughthesubjectsskinandmeasuringthewavelengthsoflightthatpassthrough.Letusreferdeoxygenatedbloodasblueandoxygenatedbloodasred,theamountofredandbluelightpassingthroughthesubjectindicatethelevelofbloodoxygenation.Themaindrawbackstopulseoximitrypertaintoitsusageoflightasameasuringtool.Motionthrowsofthesensor,ifababymovesthisleadstomotionofvenousblood,andtheRGMcanhavedifficultyindeterminingthedifferencesbetweenvenousandarterialblood,whichcangenerateinaccurateorunreadablemeasurements(5).Ambientlightalsoaffectsreadings,thoughmostoximeterclampsarecarefullyshieldedtopreventoutsidelightfromaffectingreadings.Weweresuppliedapulseoximeter,theDataexOhmeda5250,touseinourproject.ThepulseoximeterisusedintheVanderbiltUniversityNICUandwedecidedthatitwouldbebesttoworkwithpulseoximeteraswelltomeasureSa02levels.Prematureinfants,duetoweaklungs,oftenbreatheirregularly.InfactthisirregularbreathingleadstovariabilityofSaO2andasmentionedbefore,thisvariabilityofSaO2hasastrongcorrelationtoROP.ThereisdifficultymaintainingaregularlevelofSa02.BasedonthedataofDr.Graham,therearetwopointsinwhichthebabysSa02levelwilltriggeruserintervention.ThesetwopointsareatSaO2levelsabove98%orbelow80%.Thesepointsaredesignedtosavethebabyslife,notmaintainoptimumbloodoxygenlevelsorensurepropereyehealth.ShouldSa02levelsrisetoohigh,analarmsoundsfromthepulseoximeterandanurseintheneonatalwingwillcomeandthenmanuallylowersoxygenlevel.Similarly,Sa02levelsbelow80%triggerthealarm,promptinganursetoincreaseinputoxygen.TheproblemwithFigure2:Plethysmographthiscurrentmethodisthatittakesnursessometimetocomeoverandmanuallyadjusttheknob,thistimeiscrucial,theinfantisstilldroppingorincreasinginoxygen.Also,thenursehastoestimatebyeyehowmuchoxygentodeliverintotheinfant.Thereisnoactualsuggestednumberthattellsthenurseshowmuchorhowlittletoturntheinputoxygenknob.UnderstandingthatthecurrentproblemismaintainingaconstantSaO2levelinprematureinfants,wedecidedtoimprovethiscurrentsystemofcare.Wewentthroughmanydifferentexperimentaldesignsbeforewedecidedonourfinalapproach.WorkingundertheassumptionthatvariabilityisthemaincauseofROP,ourgroupsetouttohelpminimizetheproblemoffluctuatingSa02levelsinnewborninfants.WedecidedtohelpnursesoutbydevelopingsoftwarethatwillreadinandrecognizetrendsinSaO2levelsandthensuggestacorrelatingFiO2levelfortheinfant.Realizingthatweneededasystemthatmeasuredpulseoximeterdata,wedecidedonusingLabVIEWprogramminglanguageasouranalysissoftware.METHODOLOGYUniversalControlTheoryiscenteredonfeedbackcontrol.Thesystemsetupisshowninfigure3.Inthebroadestform,acontrolloopoperatesontheprinciplesofnegativefeedback.Theeffectsofanactionarereportedtothecontrollerthroughaninformationchannel.ThecontrollerstrivestominimizetheFigure3:FeedbackControlDiagramerrordifferencebetweenthemeasuredandthedesiredbehaviorsandcommandsthenextaction.Astraight-forwardexampleofthiscontrolconceptisthecruisecontrolfeatureofacar:ifthemeasuredspeedofthecardropsbelowthesetspeed(becauseofanuphillstretch),thecruisecontrollerwillaccelerate;ifthecarrollstoofast,thecontrollerwillshiftdowninanattempttominimizethediscrepancybetweenthedesiredandthemeasuredspeeds.(6)Unfortunately,thereareseveraldrawbackstousinguniversalcontroltheoryinourdesign.Firstofall,ifweendedupapplyingcontroltheorytooursystem,theFiO2levelswouldendupasadiminishingharmonicsimilartothespeeddatashowninFigure4.SinceitispossiblethattherapidlargefluctuationofFiO2levels,whichoccurinthefirstcouplecycles,contributetoincreasedchancesofretinopathy,thissystemmightendupdoingmoreharmthangood.Second,thereisthecomplexnatureofthetheoryitself.Wespentseveralmonthstryingtoapplytheadvancedmathematicalformulasbehindthetheorytoourproject.ThekeyproblemwassinceSa02isafairlyrandomsignal,whichisverydifficulttomodel,akeyelementtoapplyingcontroltheory.Afterseveralfailedattempts,anddiscussionwithcontroltheoryProfessorKazuhikoKawamura,itwasdecidedthatanotherapproachwouldbeinorder.Thesecondmethodweresearchedwasfuzzylogic.Thebasicsbehindfuzzylogicisthatyoudivideagivendatarangeintovarioussets.TheinputFigure4:SpeedDataFigure5:FuzzyLogicParameterswillthenbeassignedatruth-valuedependingonwhereitfallsintheset.Ofteninfuzzylogic,youwillseegraphssimilartofigure5.Theredatashownhasthreesets,slow,medium,fast.Sayyouhaveaninputspeedof55mph;thiswouldgetassignedavalueofzeroslow,1medium,andzerofast.Whilea60mphspeedwillbezeroslow,0.5medium,0.5fast.Afuzzycontrolsystemwilltakethisdataandadjustspeedaccordingtotheparameters,whichdefineeachset.Thisconcepthasmanyadvantageswhenappliedtocontrolofoxygenconcentration.Sinceafuzzylogicsystemisguidedbygeneralparameterswewillnotneedtomathematicallymodelourdata.Fuzzylogicisespeciallynicebecausecomplexprocessescanbecontrolledbyrelativelyfewlogicrules,thisallowsforfastercomputationinreal-timeapplications.(8)WeactuallydiddesignasimplefuzzylogiccontrollerusingtheLabVIEWsoftware,howeveruponimplementation,itwasdecidedthattheprogramwastoorigid.PartofourdesigncriteriawastohaveanadjustableoptimalFiO2level,andourfuzzylogiccontrolleronlyallowedasetoptimallevel.Booleanlogicissimplyaseriesoftrueandfalsestatementsthatguidethecontrollertoadesignatedoutput.Whileitmightbemoresusceptibletosystemnoisethanfuzzylogic,ourBooleancontrolprogramallowedforadjustableoptimalFiO2levels,alongwithfittingtherestofourdesignconstraints.AnotheradvantageofBooleancontrolisthattheprogramitselfiseasilymodifiedandadjusted.Forexample,iffutureresearchindicatesthattheoptimallevelsofFiO2foraninfantofacertainsizeisXamount,withYrange,withZmaximumrateofchange,theBooleanprogramcaneasilybemodifiedtofitthesenewparameters.Forthesereasons,wedecidedtogoaheadandusetheBooleancontrolsystemasourfinaldesignproduct.Earlyonintheideationprocess,weselectedLabVIEWasourmeanstoreadandanalyzetheRGMdata.Thereweremanyreasonsbehindthisselection.Firstandforemost,LabVIEWhasVirtualInstrumentsbuiltintoitthatallowinputfromaserialport.TheRS-232portontheRGMwasouronlydigitaloutputonthedeviceandtheobviousgatewaytogetourdatafromitintothecomputer.TheDAQwizardinLabVIEWprovedinvaluabletoourlink-up,asitisadiagnostictoolthatallowsqueryingofserialandparallelports.AnotherreasonfortheselectionofLabVIEWwastheeaseofprogramming.Asagraphicprogramminglanguage,LabVIEWmakesiteasytoimplementprograms.ThoughourgroupisexperiencedinprogrammingwithC++,LabVIEWprovedtobemuchsimplertoworkwith.Also,duetothemodularnatureofLabVIEW,variousportionsofourprogramcouldbecompletedandtestedseparately.LabVIEWproveditselftobethebestprogramtowriteoursoftware.ItwasimportantthatsettingsontheRGMmatchedthesettingsinourLabVIEWprogramtoinsuresimpleandpromptdatatransfer.TherearemanyoptionswithintheRGMssetupmenurelevanttocomputerinput.TheRS-232connectionisanindustrystandardthatallowsforeasyintegrationbetweenperipheraldevicesandacomputerorprinter.The232connectorinthecaseofourOhmeda5250RGMappearstobeprimarilydesignedforprinterusage.Itsoutputisfixedinoneformat:58linesofdataperpage,dividedinto19columns.TheoptionsbuiltintotheRGMsoftwarepertainedtotherateofoutputanditscommunicationprotocol.Interestingly,LabVIEWsdefaultserialI/OVIisconfiguredatthesamesetofserialprotocolsasoneofthepresetsontheRGM.Bydefault,LabVIEWsserialI/OVIreadsat9600bauds,noparity,withonestopbit.ThesesettingscorrespondtothosebuiltintotheRGMsconnectivityFigure1:RGMOutputPortFigure6:RS232OutputFigure7:RS232ConnectorCablesettings.TheotherLabVIEWsettingrelevanttoRS-232connectivityisthePrintPeriodmenuoption.TheRGMhas5possibleoutputperiods:nooutput,10seconds,1minute,5minutes,and60minutes.Atthemostfrequentsetting,theRGMsavesvitalinformationaboutthesubjectevery10seconds.AttachingourRS-232cabletoourRGM,connectingittoacomputerserialportlabeledCOM1,andqueryingtheportwithLabVIEWsDAQwizardallowedustostudythebasicoutputofitsRS-232port.Onceprintedortransferredtothecomputer,storedvaluesareremovedfromtheRGMsbuffer.TheRGMbuffersthesevitalstatisticsinternally,savingupto24hoursofdatadependingontheprintperiodsetting.Withthebasicoperatingparametersofourconnectionintact,ournextstepwastoestablishareliableandusefulconnectionbetweentheRGMandLabVIEW.AbasicVIcontainedinLabVIEWisitsSerialStringReadinstrument.Thisprogramisdesignedtoreadastringinfromacommunicationsportofthecomputer.Bydefault,COM1wasused,soweselectedthisasourcommunicationsport.ThetrickysettinginthisVIisitssamplingparameters.Anycommunicationprogramneedsalimittoitsinput.ThelimitsoftheSerialStringReadVIaretimeorbytesize.Basically,LabVIEWcaneitherreaddatainforacertainamountoftimebeforestoppingtoreadit;orreadindatauntilthatdataexceedsapre-setnumberofbytes.OurRGMwasconfiguredtooutputdataevery10seconds.Thisseemedlikethemostobviouslimittoobserve,asitcorrelateswiththerateofdataoutputfromtheRGM.Testingthisimplementationrevealedverypromisingresults.OnelineofvitalstatisticswasreadintoLabVIEWeverytenseconds.Theformatofeachlinewasidentical,withrepeatablespacingandformatbetweenthevariousbitsofrespiratorydata.Thedataisarrangedinto16columns.Manyofthesevaluesrelatetopressures,pulserate,andvariousdataonoxygenandC02levels.Theimportantdataisincolumn14:SaO2%.Thisvaluerelayspercentageofsaturatedbloodoxygen.OfallthedataoutputbytheRGM,thisistheonecriticalpieceofinformationinmonitoringtrendsinthebloodoxygenlevelsofthesubject.AnalyzingtheoutputstringfromtheDAQwizard,wedeterminedthattheSaO2%alwaysappearedonthe60thspaceofanygivenline.TheRGMdataisoutputinaveryrepeatableformat,asitisdesignedtobeoutputinevenrowandeasilyprintedout.Whatwassousefulaboutthisformatisthatitmakesiteasytoscanthestringforthisspecificnumber,usingitsplacementasthesearchprotocol.AfterimplementingLabVIEWsserialstringreadVIwithatimeoutof10seconds,wewereabletoreliablygetonelineofRGMdatareadinevery10seconds.Theproblemwithserialcommunicationisthatdataisalwaysreadinasastring.Astringisdifferentthananinteger,asitmaycontainspaces,dashes,orcharacters.Clearly,ournextstepwastoconverttherelevantvaluesinthestringintoaninteger.Aspreviouslystated,theSaO2%appearspredictablyuponthe60thspaceofanyinputstring.UsingaLabVIEWIntegerStringSearchVI,wewereabletoconvertthenumberstartingonthe60thlineoftheinputstringintoaninteger.ThetwoinputsintothisVIarethestringtobesearchedandthespacenumberfromlefttorighttosearch.AsymbolorletterinthisinputspacewillresultintheVIreturningazerovalue.UsingthisVI,wewereabletoreadinthecurrentSaO2%valueevery10seconds,andconvertitintoanintegerthatLabVIEWcanfurtheranalyze.AftertheconnectionbetweenthecomputerandLabVIEWhadbeenestablished,thenextstepwastobeginanalysisofdata.Thefirststepwastoplacetheentireprograminaswitchedwhileloop.Thisensuresthattheprogramcontinuestosampledataevery10seconds,nomatterwhat.Inourapproachtothedata,wedidanumberofthingstoreducevariabilityandprovidePrematureInfant1OxygenLevels020406080100120050010001500200025003000time(seconds)%oxygenlevelsSpO2%FiO2%PrematureInfant2OxygenLevels0204060801001200200040006000800010000time(s)%OxygenlevelsSpO2FiO2NICUstaffwithusefulinformation.BycreatingaseriesofBooleanlogicstatements,wewerequicklyabletosortthedataintofiveclasses,basedontheseverityofthebabysrespiratorystate.Anotherfeatureourgroupdecidedtoimplementintotheprogramwasavariableideallevel.Basedonourresearchconcerningprematurebabies,wefoundthatinfantshavevaryingidealoxygenlevels.WhereonebabymightbecontenthavingaSa02levelof90%,anothermaybecomfortablearound95%andgointogreatdistressshoulditsbloodoxygenlevelreach90%.Withthisinmind,wedecidedtomakeavariableideallevelcontrolsothatitcouldbeconfiguredtotheneedsofanychild.TherestofthissectionoftheprogramcomputesthedifferencebetweentheSa02valuereadinfromtheRGMandtheidealSa02level,andthenindicatesiftheoxygenlevelinthesubjectneedstoberaisedorloweredtobebroughtdowntotheideallevel.RESULTSOuranalysisofthefiveprematureinfantdatasetsgatheredbythepulseoximetershowedthatidealFiO2levelsvarygreatlyfromoneinfanttothenext,andthereisnoclearcutequationtorelateFiO2levelswithSaO2levels.Comparingfigure8withfigure9,wecanseethatInfant1’sFiO2levelsaresetataround45%whileInfant2’sFiO2levelsareFigure8:PrematureInfant#1Figure9:PrematureInfant#2around35%average.TherearetwoverydifferentFiO2levels,buthaveprimarilythesameSaO2patterns.ThisisthemainreasoningbehindtheuseofanadjustableIdealFiO2levelinourcontrolprogram.SincetheSaO2levelsaresodifficulttomap,itwouldbeveryhardtomakeacomputersimulationtotestourcontrolprogram.Inadditiontothis,asterilecontrolledoxygenchamber,suchasanincubator,withalivesubjectwouldberequiredtoproducedatathattestsifourdesignactuallydoesreducevariationinSaO2levels.Becauseofthis,wewilljustberelyingonDr.Penn’sresearchdiscoveries,whichshowthatadjustingFiO2levelswillhelpreduceSaO2variation.Theendresultofourprogrammingisafrontpanelthatisremarkablyeasytounderstand.ItprovidesataglanceinformationaboutthebabysSa02level,thedifferencebetweenitandthedesiredlevel,aswellascategorizationinoneoffivelevelsofdistress.Figure10:LabVIEWFrontPanelScreenshotTherearenotmanysafetyissuesthatconcernourdesign.SinceourprojectonlysuggestschangesinFi02levels,humanjudgmentisstillutilizedincasethemachinemakesanerroneousmeasurement.Atthesametime,weplacealarmbarriersincaseSaO2levelsbegintogetoutofhand.Evenifoursystemfails,thepulseoximeterhasitsowninternalalarmsystemsthatwillsoundiftheinfantsoxygenlevelsgetoutofhand.Withanaverageof448,500prematureinfantsborneveryyear,weexpectourproducttobeusedin100%ofcasesinvolvingnewbornsandsupplementaloxygen.Therearethefollowingcostsinusingourproduct:arateof$30/hour/personforactualdesigningandconsultingperprogram,PulseOximeter:$1000-1500,ComputerwithLabVIEW:$500-1000RS-232cable:$10,TotalCostperunit:$1540-2540.Thebulkofthesecostslieinthecomputerandoximeter.Seeingasthistechnologywillideallyberefinedintoacontroller-microprocessorwhichwillnotneedaPCtorunitispossiblethatthisdevicewillbeverycheaptocreate.Itispossiblethatbypiggybackingthisdevicewithapulseoximetercompanywewillbeabletoselltheunitsasapackagedeal,whichwillbefarsuperiortocompetitorssingleoximeterunit.Plustheseunitsshouldlastashelflifeof5years;wewillquotetheshelflifeaslessthanactualforliabilityreasons.Hospitalswillbeencouragedtobuythepackagedunitsbecauseofthepotentialmillionsofdollarsinretinopathymalpracticelawsuitsourdevicecouldprevent.CONCLUSIONOurmajordesigncriteriawastodevelopanoxygenmonitoringsystem,whichcouldtakeindatafromapulseoximeterandmakearecommendedFiO2changeinordertohelpdecreaseSaO2variability.WealsowantthissystemtobeabletomakeitscalculationsatanyFiO2inputlevel.Ourprogramdoesindeedreachthesedesigncriteria.Someoftheminorcriteriawedidnotreachwasbeingabletotestourprogramaswellascomparecontrollogictechniquesonlivesubjects.ThefinalversionofthisprogramreadsinRGMdata,comparesittoanidealvalue,andcalculatesthedifferencebetweenthem.Thisprovidesataglancedatathatmedicalprofessionalscanusetoquicklydeterminetheamountbywhichoxygenflowshouldbealtered.Oursystem,asitcurrentlyexists,ispiggy-backedontoapulseoximeter.Thenextstepwewouldliketodoisuselivesubjects,suchasrats,andcollecttheirsaturatedbloodoxygenlevels.ThisdatacanthenbestatisticallyanalyzedtoseeifthereisasignificantimprovementinSaO2variation.Wearealsoveryinterestedtoseeifafuzzylogicsystemcouldbedevisedwhichismoreflexiblewithregar