Open AccessMinimizing Symptom-based Diagnostic Errors Using Weighted Input Variables and Fuzzy Logic Rules in Clinical Decision Support Systems
Publication year - 2021
Publication title -
international journal of advanced trends in computer science and engineering
Language(s) - Uncategorized
Resource type - Journals
ISSN - 2278-3091
DOI - 10.30534/ijatcse/2021/121032021
Subject(s) - fuzzy logic , computer science , clinical decision support system , artificial intelligence , data mining , decision support system , machine learning
ThisstudycoversanapproachtowardsminimizingDiseasediagnosticerrorsusingweightedinputvariablesandFuzzyLogicruleswithmultiphasediagnosticengine.Theweightswereappliedbecausedifferentsymptomsmayhavedifferentdegreesofimportanceindifferentdiseases.Thisistoensurethatrecommendationsfordiseaseconfirmationbasedonsymptomsreturngoodpercentageoftruepositiveandtruenegatives.Thestudycreatesanenhanced,accurateandprecisesystemformedicaldiagnosisevenwhenonlythesymptomsareconsidered.Inordertoevaluatethemodel,fourcategoriesofdiagnoseswerecarriedoutwithoutusingthemodelatthefirstinstanceandusingthemodelatthesecondinstancewith50patientsdoneat4differentdiagnosticinstances.Thetruepositive(TP)andtheFalsenegativestatisticswereobtainedfromwherethefalsepositiverate(TPR)orsensitivityandfalsepositiverate(FPR)werederived.ThegraphofTPRvsFPRwasplottedfromwherethequalityofdiagnosescouldbegottenfromtheReceiverOperatingCharacteristics(ROC)space.Theresultshowsthatsensitivity,whichistheabilityofatesttocorrectlyidentifythosewiththediseaseorTruePositiveRate,andspecificity,whichistheabilityofthetesttocorrectlyidentifythosewithoutthediseasealsocalledTrueNegativeRateTNRstoodat87%and86%respectivelyusingthedevelopedmodelandthesameparameteryielded72%and56%respectivelywithoutusingthemodel.Theresultalsoshowsthatthefalsepositiverate(FPR)whichindicatesthedegreeoffalsealarmis19%usingthenewmodelwhileitis44%withoutusingthemodel.Thisresultshowsthatthelikelihoodofmakingwrongclinicaldiagnosticdecisionsismuchlowerwiththisapproach