| Titre : | Deep Learning technique and parallel optimization algorithm for intelligent pattern recognition |
| Auteurs : | Hakima Rym RAHAL, Auteur ; Sihem Slatnia, Auteur |
| Type de document : | Thése doctorat |
| Année de publication : | 2025 |
| Format : | 1 vol. (84 p.) |
| Langues: | Anglais |
| Mots-clés: | Blockchain,MedicalBigdata,DeepLearning(DL)methods,Pattern Recognition. |
| Résumé : |
Misdiagnosis posesasignificantchallengewithinthehealthcaresector,carryingpotentially severeconsequencesforpatients,includingdelayedorinappropriatetreatment,unnecessaryme- dical procedures,emotionaldistress,financialburdens,andlegalrepercussions.Toaddressthis issue, weproposetheutilizationofdeeplearningalgorithmstoenhancetheprecisionofmedi- cal diagnoses.However,thedevelopmentofaccuratedeeplearningmodelsformedicalpurposes necessitates substantialquantitiesoftop-qualitydata,aresourcethatcanbechallengingfor individual healthcareentitiestoacquire.Consequently,thereisaneedtoaggregatedatafrom varioussourcestocreateadiversedatasetsuitableforeffectivemodeltraining.Nevertheless, the sharingofmedicaldataacrossdifferenthealthcaresectorsisfraughtwithsecurityconcerns due tothesensitivenatureoftheinformationandstringentprivacyregulations.Totacklethese complex challenges,weadvocatefortheadoptionofBlockchaintechnology,whichoffersase- cure, decentralized,andprivacy-centricapproachtosharinglocallytraineddeeplearningmodels, therebyobviatingtheneedtoexchangerawdata.Ourproposedtechnique,knownasmodelen- sembling,combinesthestrengthsofmultiplelocaldeeplearningmodelsbyaggregatingtheir weightstoconstructaunifiedglobalmodel.Thisglobalmodelenablesaccuratediagnosisof intricatemedicalconditionsacrossvariouslocationswhilesafeguardingpatientprivacyanddata integrity.Ourresearchservesasatestamenttotheefficacyofthisapproach,achievinghigh accuracy ratesinthediagnosisofthreediseases(accuracyof97.44%fortheBreastCancer, 97.14 %fortheDiabetes,and98.51%fortheLungCancer)thatsurpassthoseofindividual localmodels.Furthermore,wehavesuccessfullydevelopedamulti-diagnosisapplicationasan outcome ofthisinnovativemethodology. |
| Sommaire : |
Acknowledgementsi Abstract ii Résumé iii Contents v List ofFiguresix List ofTables xi 1 Introduction1 2 StateoftheArt4 2.1 Introduction......................................4 2.2 PreviousWorks....................................4 2.2.1 APredictiveToolforOvarianCancerbasedonBlockchaintechnology..4 2.2.2 FrameworkforDiagnosingCOVID-19usingX-rayImages........4 2.2.3 ABlockchain-PoweredExpertSystemforHealthcareEmergencies....5 2.2.4 ABlockchain-basedDeepLearningPlatformforMyopia.........5 2.2.5 ABlockchain-basedSmartphonePlatformforDiagnosingMalaria....6 2.2.6 Utilizingblockchainforsecuredatasharingwithinhealthcaresystems..6 2.2.7 APersonalHealthRecord(PHR)ApplicationbuiltonBlockchainTech- nology .....................................7 2.2.8 AdeepLearningModelforDetectingandEvaluatingSportsInjuries..7 2.2.9 ACNNmodelbasedonblockchainforassessinglungcancer’sfoodquality and well-beingaspects............................7 2.2.10 Leveragingablockchain-powereddigitalpathologysystemtoenhancedi- agnostic processes...............................8 2.2.11 HeartDiseasePredictionAlgorithm.....................8 2.2.12 Blockchain-BasedSelf-DefinedAccessControlforDataSecurity.....8 2.3 TheMainFocusDiseasesinOurImplementation.................9 2.3.1 BreastCancer.................................9 2.3.2 LungCancer.................................9 2.3.3 Diabetes....................................10 2.4 Conclusion.......................................10 v CONTENTS vi 3 MisdiagnosisandDeepLearning12 3.1 Introduction......................................12 3.2 MisdiagnosisinHealthcare..............................12 3.2.1 StatisticsintheEuropeanUnion......................12 3.2.2 StatisticsintheU.S.A............................13 3.2.3 DeepLearningEffectiveness.........................13 3.3 DeepLearningBasics.................................13 3.3.1 DeepNeuralNetworkArchitectures.....................14 3.3.1.1 ArtificialNeuralNetworks(ANNs)................14 3.3.1.2 RecurrentNeuralNetworks(UnidirectionalRNN)........15 3.3.1.3 RecurrentNeuralNetworks(BidirectionalRNN)........15 3.3.1.4 ConvolutionalNeuralNetworks(CNNs).............16 3.3.1.5 Autoencoders(AEs)........................16 3.3.1.6 StackedAutoencoders.......................16 3.3.1.7 VariationalAutoencoders(VAEs).................17 3.3.2 OverfittingandUnderfitting.........................17 3.3.2.1 Overfitting.............................17 3.3.2.2 Underfitting.............................17 3.3.3 RegularizationsinDeepLearning......................19 3.3.3.1 L1Regularization(LassoRegularization).............19 3.3.3.2 L2Regularization..........................19 3.3.3.3 Dropout...............................20 3.3.3.4 BatchNormalization........................20 3.3.3.5 EarlyStopping...........................20 3.3.4 CombinationofDeepLearningModels...................20 3.3.4.1 TransferLearning..........................21 3.3.4.2 ArchitecturalCombinations....................21 3.4 Conclusion.......................................22 4 SecurityandPrivacyinHealthcare24 4.1 Introduction......................................24 4.2 SecurityandPrivacyChallengesinHealthcare...................24 4.2.1 DataBreaches.................................24 4.2.2 IoTVulnerabilities..............................25 4.2.3 LackofStandardization...........................25 4.2.4 Third-PartyVulnerabilities..........................25 4.2.5 ComplianceChallenges............................26 4.2.6 PatientConsentandDataAccess......................26 4.2.7 DataEncryption...............................26 4.2.8 EmergingTechnologies............................27 4.3 SolutionsforSecurityandPrivacyinHealthcare..................27 4.3.1 AccessControlandAuthentication.....................27 4.3.1.1 Role-BasedAccessControl(RBAC)...............27 4.3.1.2 Multi-FactorAuthentication(MFA)...............28 4.3.2 DataEncryption...............................30 4.3.3 SecureCommunication............................30 4.3.3.1 VirtualPrivateNetworks(VPNs).................30 CONTENTS vii 4.3.3.2 SecureSocketLayer(SSL)/TransportLayerSecurity(TLS)..31 4.3.4 DigitalSignatures...............................32 4.3.5 SecureCloudStorage.............................34 4.3.6 BlockchainTechnology............................35 4.3.6.1 BlockchainArchitecture......................35 4.3.6.2 BlockchainConcepts........................37 4.3.6.3 TypesofBlockchain........................38 4.4 FileSharingmethodsThroughBlockchain.....................39 4.4.1 DecentralizedFileSharingPlatforms....................39 4.4.2 SmartContractsforFileSharing......................39 4.4.3 Blockchain-basedContentPlatforms....................39 4.5 Blockchainandhealthcare..............................39 4.6 Conclusion.......................................40 5 Theproposedmethodandtheimplementation41 5.1 Introduction......................................41 5.2 Proposedmethod...................................41 5.3 Implementation....................................42 5.3.1 Datasets....................................43 5.3.1.1 BreastCancerWisconsin(Diagnostic)DataSet.........43 5.3.1.2 LungCancerPredictionDataset.................43 5.3.1.3 DiabetesUCIDataset.......................44 5.3.2 PreparationoftheDatasets.........................44 5.3.3 DeepLearningModels............................46 5.3.3.1 BinaryClassificationANN(BreastCancerandDiabetes)....47 5.3.3.2 CategoricalClassificationANN(LungCancercase).......49 5.3.4 ModelEnsemblingmethod..........................50 5.3.5 EthereumBlockchain.............................51 5.3.6 IPFS......................................52 5.4 ComplexityCalculation................................55 5.4.1 Binaryclassification.............................55 5.4.1.1 ANN1/ANN2............................55 5.4.1.2 ANN3................................56 5.4.2 categoricalclassification...........................56 5.4.2.1 ANN1/ANN2............................56 5.4.2.2 ANN3................................56 5.5 EnvironmentandLibraries..............................57 5.5.1 GoogleColab.................................57 5.5.2 GoogleColabHardwareConfigurations...................58 5.5.3 ProgrammingLanguage...........................58 5.5.3.1 Libraries...............................58 5.6 Conclusion.......................................60 6 ResultsandDiscussion61 6.1 Introduction......................................61 6.1.1 BreastCancer.................................61 6.1.2 LungCancer.................................64 CONTENTS viii 6.1.3 Diabetes....................................66 6.2 ApplicationOverview.................................69 6.3 Conclusion.......................................70 7 AchievementsandConclusions72 Bibliography 75 List OfPublications84 |
| En ligne : | http://thesis.univ-biskra.dz/id/eprint/6826 |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| TINF/201 | Théses de doctorat | bibliothèque sciences exactes | Consultable |
