资源描述:
《Artificial Neural Networks Improve the Accuracy of Cancer Survival Prediction》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、857ArtificialNeuralNetworksImprovetheAccuracyofCancerSurvivalPredictionHarryB.Burke,M.D.,Ph.D.1BACKGROUND.TheTNMstagingsystemoriginatedasaresponsetotheneedforPhilipH.Goodman,M.D.,M.S.2anaccurate,consistent,universalcanceroutcomepredictionsystem.SincetheDavidB.Rosen,Ph.
2、D.1TNMstagingsystemwasintroducedinthe1950s,newprognosticfactorshaveDonaldE.Henson,M.D.3beenidenti®edandnewmethodsforintegratingprognosticfactorshavebeen4developed.ThisstudycomparesthepredictionaccuracyoftheTNMstagingsystemJohnN.Weinstein,M.D.,Ph.D.FrankE.Harrell,Jr.,5
3、withthatofarti®cialneuralnetworkstatisticalmodels.Ph.D.6METHODS.For5-yearsurvivalofpatientswithbreastorcolorectalcarcinoma,JeffreyR.Marks,Ph.D.7theauthorscomparedtheTNMstagingsystem'spredictiveaccuracywiththatofDavidP.Winchester,M.D.8arti®cialneuralnetworks(ANN).Thear
4、eaunderthereceiveroperatingcharacteris-DavidG.Bostwick,M.D.ticcurve,asappliedtoanindependentvalidationdataset,wasthemeasureofaccuracy.1BioinformaticsandHealthServicesResearch,RESULTS.FortheAmericanCollegeofSurgeons'PatientCareEvaluation(PCE)DepartmentofMedicine,NewYor
5、kMedicalCol-dataset,usingonlytheTNMvariables(tumorsize,numberofpositiveregionallege,Valhalla,NewYork.lymphnodes,anddistantmetastasis),thearti®cialneuralnetwork'spredictions2DepartmentofMedicine,UniversityofNevadaofthe5-yearsurvivalofpatientswithbreastcarcinomaweresign
6、i®cantlymoreSchoolofMedicine,Reno,Nevada.accuratethanthoseoftheTNMstagingsystem(TNM,0.720;ANN,0.770;P3DivisionofCancerPreventionandControl,Na-0.001).FortheNationalCancerInstitute'sSurveillance,Epidemiology,andEndtionalCancerInstitute,Bethesda,Maryland.Resultsbreastcar
7、cinomadataset,usingonlytheTNMvariables,thearti®cial4neuralnetwork'spredictionsof10-yearsurvivalweresigni®cantlymoreaccurateDivisionofCancerTreatment,NationalCancerInstitute,Bethesda,Maryland.thanthoseoftheTNMstagingsystem(TNM,0.692;ANN,0.730;P0.01).ForthePCEcolorectal
8、dataset,usingonlytheTNMvariables,thearti®cialneural5DepartmentofHealthEvaluationSciences,network'spredictionsofthe5-yearsurv