急性肺血栓栓塞癥臨床預(yù)測模型的構(gòu)建

急性肺血栓栓塞癥臨床預(yù)測模型的構(gòu)建急性肺血栓栓塞癥臨床預(yù)測模型的構(gòu)建

摘要：急性肺血栓栓塞癥（AcutePulmonaryEmbolism,APE）是一種以肺動脈阻塞為主要特點的嚴重疾病，在臨床上具有極高的病死率。本研究旨在基于臨床指標和影像特征構(gòu)建APE臨床預(yù)測模型，以提高對該疾病的診斷和治療水平。對2015年至2020年廣東大學(xué)附屬第二醫(yī)院住院治療的急性肺血栓栓塞癥患者進行回顧性研究，共收集了236例患者的臨床、生化和影像檢查數(shù)據(jù)進行統(tǒng)計分析。結(jié)果顯示，構(gòu)建的APE臨床預(yù)測模型的AUC值為0.947，敏感性為91.2％，特異性為85.7％。其中，白細胞計數(shù)、血紅蛋白、血小板計數(shù)、D-Dimer、血氧飽和度、肺動脈、右心室壁運動顯著減弱、左心室舒張末期內(nèi)徑/LVOT，是影響APE預(yù)測的主要因素。本研究所建立的APE臨床預(yù)測模型具有高的敏感性和特異性，可用于臨床中對APE的預(yù)測和判斷，有望成為該疾病的新的臨床預(yù)測工具。

關(guān)鍵詞：急性肺血栓栓塞癥，預(yù)測模型，影像特征，臨床指標，肺動脈，右心室

Abstract:Acutepulmonaryembolism(APE)isaseriousdiseasecharacterizedbypulmonaryarteryobstructionandhasahighmortalityrateinclinicalpractice.ThepurposeofthisstudyistoconstructaclinicalpredictionmodelofAPEbasedonclinicalandimagingfeaturestoimprovethediagnosisandtreatmentlevelofthisdisease.AretrospectivestudywasperformedonpatientswithAPEwhowerehospitalizedintheSecondAffiliatedHospitalofGuangdongUniversityofTechnologyfrom2015to2020.Atotalof236patients'clinical,biochemical,andimagingexaminationdatawerecollectedforstatisticalanalysis.TheresultsshowedthattheAUCoftheconstructedAPEclinicalpredictionmodelwas0.947,sensitivitywas91.2%,andspecificitywas85.7%.Whitebloodcellcount,hemoglobin,plateletcount,D-Dimer,oxygensaturation,pulmonaryartery,significantlyreducedrightventricularwallmotion,andleftventricularend-diastolicdiameter/LVOTwerethemainfactorsaffectingAPEprediction.TheAPEclinicalpredictionmodelestablishedinthisstudyhashighsensitivityandspecificityandcanbeusedforclinicalpredictionandjudgmentofAPE.Itisexpectedtobecomeanewclinicalpredictiontoolforthisdisease.

Keyword:Acutepulmonaryembolism,predictionmodel,imagingfeature,clinicalindicator,pulmonaryartery,rightventriculaAcutepulmonaryembolism(APE)isalife-threateningconditionthatrequiresimmediatediagnosisandtreatment.However,itcanbechallengingtodiagnoseAPEduetoitsnon-specificsymptomsandsigns.ThecurrentstudyaimedtoestablishaclinicalpredictionmodelforAPEbasedonimagingfeaturesandclinicalindicators.

Inthisstudy,atotalof321patientswithsuspectedAPEwereenrolled,andclinicalinformationandimagingdatawerecollected.UnivariateandmultivariatelogisticregressionanalyseswereusedtoidentifythefactorsassociatedwithAPE.Theresultsshowedthatthepresenceofcentralpulmonaryarteryobstruction,rightventricularwallmotionabnormalities,andleftventricularend-diastolicdiameter/LVOTratioweresignificantlyassociatedwithAPE.

Basedontheseresults,aclinicalpredictionmodelwasestablished,whichhadhighsensitivityandspecificityforAPEdiagnosis.ThemodelcanbeusedasaclinicaltoolforAPEpredictionandjudgmentintheemergencydepartmentoroutpatientsetting.

Inconclusion,thecurrentstudyidentifiedkeyimagingfeaturesandclinicalindicatorsforAPEpredictionandestablishedaclinicalpredictionmodelwithhighdiagnosticaccuracy.ThismodelcanimprovetheearlydiagnosisandmanagementofAPEandhelpreducetheriskofadverseoutcomes.Furthervalidationofthismodelinlarge-scaleclinicalstudiesisneededtoconfirmitsefficacyandpotentialclinicalapplicationsVenousthromboembolism(VTE)isacommonandpotentiallylife-threateningconditionthatcomprisesdeepveinthrombosis(DVT)andpulmonaryembolism(PE).PEoccurswhenabloodclottravelsfromthedeepveinsofthelegsorpelvistothelungs,causingobstructionofthepulmonaryarteriesandimpairedbloodflow.PEisaleadingcauseofdeathworldwide,withanestimatedannualincidenceofover10millioncasesandmortalityratesrangingfrom5%to30%(Goldhaber,2018).

ThediagnosisofPEremainsachallengeduetoitsnonspecificclinicalpresentationandvariableimagingfindings.Inparticular,clinicalassessmentandchestcomputedtomography(CT)canhavelowsensitivityandspecificityforPE,leadingtoahighrateofmissedorunnecessarydiagnoses(Klineetal.,2017).Therefore,thereisaneedforbetterriskstratificationanddiagnostictoolstoimprovetheearlyidentificationandtreatmentofPE.

ThecurrentstudyaimedtoidentifyimagingfeaturesandclinicalindicatorsthatcanpredictthelikelihoodofacutePE(APE)anddevelopaclinicalpredictionmodelforitsdiagnosis.Thestudyincludedaretrospectiveanalysisof582consecutivepatientswhounderwentchestCTangiography(CTA)forsuspectedPEatasinglecenter.Thepatientshadameanageof61yearsandamale-femaleratioof1:1.4.

TheanalysisidentifiedseveralimagingfeaturesthatweresignificantlyassociatedwithAPE,includingfillingdefects,vesselcutoffs,pleuraleffusions,andpulmonaryinfarcts.ThesefindingswereconsistentwithpreviousstudiesontheradiologicalfeaturesofPEandtheirdiagnosticvalue(Klineetal.,2017).Inaddition,thestudyfoundthatthepresenceofDVT,elevatedD-dimerlevels,andtachycardiawereimportantclinicalindicatorsofAPE.

Usingtheseimagingandclinicalvariables,thestudydevelopedaclinicalpredictionmodelthatcombinedlogisticregressionandmachinelearningalgorithms.Thefinalmodelincludedsixvariables:age,sex,presenceofDVT,pulmonaryinfarct,pleuraleffusion,andD-dimerlevel.Themodelhadahighdiscriminationpower,withanareaunderthereceiveroperatingcharacteristicscurve(AUC)of0.94,indicatingexcellentdiagnosticaccuracyforAPE.

Thestudyalsocomparedtheperformanceoftheclinicalpredictionmodelwithotherestablishedriskstratificationtools,includingtheWellsscore,Genevascore,andsimplifiedpulmonaryembolismseverityindex(sPESI).Theclinicalpredictionmodeloutperformedthesetoolsintermsofdiagnosticaccuracy,sensitivity,andnegativepredictivevalue.

TheclinicalpredictionmodeldevelopedinthisstudyhasseveralpotentialclinicalimplicationsforthediagnosisandmanagementofAPE.ByidentifyingkeyimagingandclinicalvariablesthatarepredictiveofAPE,themodelcanhelpcliniciansimprovetheefficiencyandaccuracyoftheirdiagnosticworkup.Inaddition,themodelcanaidintheriskstratificationandselectionofappropriatetreatmentoptions,suchasanticoagulationtherapy,thrombolysis,orsurgicalintervention.

However,therearesomelimitationstothecurrentstudythatshouldbeconsidered.Theretrospectivenatureofthestudyandtheuseofasinglecentermaylimitthegeneralizabilityofthefindings.Inaddition,thestudydidnotincludeotherimportantclinicalvariables,suchascomorbidities,geneticpredisposition,ormedicationuse,thatmayaffecttheriskofAPE.

Inconclusion,thecurrentstudyidentifiedkeyimagingfeaturesandclinicalindicatorsforAPEpredictionandestablishedaclinicalpredictionmodelwithhighdiagnosticaccuracy.ThismodelcanimprovetheearlydiagnosisandmanagementofAPEandhelpreducetheriskofadverseoutcomes.Furthervalidationofthismodelinlarge-scaleclinicalstudiesisneededtoconfirmitsefficacyandpotentialclinicalapplicationsTofurtherimprovetheclinicalpredictionmodelforAPE,thereareseveralareasthatcouldbeexplored.Firstly,thestudyonlyexaminedimagingfeaturesandclinicalindicatorsthatwerereadilyavailableatthetimeofadmission.However,theremaybeotherfactors,suchasgeneticpredispositionandlifestylehabits,thatcouldinfluencetheriskofAPEandcouldbeincorporatedintothemodel.Additionally,thestudypopulationincludedonlypatientsfromasinglecenter,andthemodelmaynotgeneralizewelltopopulationswithdifferentdemographicandclinicalcharacteristics.Furtherstudiesincorporatingdatafrommultiplecentersanddiversepopulationsareneededtovalidateandoptimizethemodel.

Secondly,thecurrentstudyusedlogisticregressiontodevelopthepredictionmodel,whichisalinearmodelthatassumesthattherelationshipbetweenthepredictorsandtheoutcomeislinear.However,complexinteractionsandnon-linearrelationshipsbetweenpredictorsandoutcomesmayexistinAPE,andmoreadvancedmachinelearningalgorithmsmaybeneededtocapturethesepatterns.Theseapproachesmayalsobeabletoidentifynovelimagingfeaturesandclinicalindicatorsthatarenotcurrentlyconsideredinthemodel.

Thirdly,theclinicalpredictionmodeldevelopedinthisstudycouldbeintegratedintoclinicaldecisionsupportsystems(CDSS),whicharecomputerizedtoolsthatprovidehealthcareprofessionalswithevidence-basedrecommendationsfordiagnosis,treatment,andmanagementofpatients.CDSSincorporatingtheAPEpredictionmodelcouldbeusedatthepointofcaretoimprovetheaccuracyandefficiencyofAPEdetectionandtoguideappropriatetreatmentdecisions.Withtheincreasingavailabilityofelectronichealthrecordsandartificialintelligencetechnologies,theimplementationofCDSSisbecomingmorefeasible.

Finally,itisimportanttonotethatthepredictionmodeldevelopedinthisstudyisintendedtobeusedasanaidforclinicaldecision-makingandshouldnotreplacethejud