Titre : | Une approche basée deep-learning pour la segmentation d'images médicales |
Auteurs : | MOHAMED MESSAOUD KISRANE, Auteur ; Samir Tigane, Directeur de thèse |
Type de document : | Monographie imprimée |
Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2021 |
Format : | 1 vol. (59 p.) / ill. / 29 cm |
Langues: | Anglais |
Résumé : | The COVID-19 pandemic is an infectious disease that affects many people all over the world, And has caused thousands of deaths since thebeginning of it’s spread.Reverse transcription polymerase chain reaction(RT-PCR)it is one of the standard traditional di-agnostic methods for detecting COVID-19 in infected persons. However,recent reports have stated that the sensitivity of RT-PCR may not be high enough to detect COVID-19,and it takes a long time to detect those infected with COVID-19.One of the most successful automated methods that achieve state-of-the-arttechnology is deep learning,which is a convolutional neural network(CNNs).Diagnosing COVID-19 based on deep learning(CNNs) applied to lung computed to-mography (CT)scans captured from old COVID-19 patients could be a feasible solutions for detecting and labeling infected tissues on CT lung images of new patients.In this work,webuilt one of the most popular CNNs-based architectures that is U-Net and applied it to CT lung scans,for automatic detection and segmentation of Covid-19 affected regions of the lung,then evaluated this architecture’s performance using different metrics. |
Sommaire : |
General Introduction1
1 IntroductiontoCTSegmentationMethods2 1.1 Introduction..................................... 3 1.2 Computed-Tomography(CT)........................... 3 1.2.1 CTConcepts................................. 3 1.2.2 CTPrinciple................................. 4 1.3 ImageSegmentation................................ 4 1.3.1 ConceptsofImageSegmentation..................... 5 1.3.2 ConceptsofImageSegmentation..................... 5 1.3.3 ImageSegmentationAlgorithms..................... 6 1.4 DeepLearningApproach............................. 10 1.4.1 NeuralNetworks.............................. 10 1.4.2 WhatIsDeepLearning?.......................... 11 1.5 TrainingProcedure................................. 12 1.5.1 SupervisedLearning............................ 12 1.5.2 UnsupervisedLearning.......................... 12 1.5.3 Semi-SupervisedLearning......................... 13 1.6 DeepLearningCategories............................. 13 1.6.1 ConvolutionalNeuralNetworks(CNNs)................ 13 1.6.2 Pre-trainedUnsupervisedNetworks................... 18 1.6.3 RecurrentNeuralNetworks........................ 21 1.7 MedicalImageSegmentationBasedonDeepLearning............ 22 1.7.1 U-NetArchitecture............................. 22 1.7.2 U-Netlayers................................. 23 1.7.3 U-NetAdvantagesandDisadvantages................. 25 2 CNNforCovid-19SegmentationandImplementation27 2.1 Introduction..................................... 28 2.2 GeneralArchitecture................................ 28 2.3 DetailedArchitecture................................ 29 2.3.1 DataPre-process.............................. 29 2.3.2 TrainU-NetModel............................. 30 2.3.3 ModelCompilationprocess........................ 31 2.4 ImplantationDetails................................ 35 2.4.1 DataSource................................. 35 2.4.2 DataSource................................. 35 2.5 Results........................................ 38 2.5.1 ModelsScoresandLosses......................... 39 2.5.2 SegmentationResults........................... 40 2.6 Discussion...................................... 42 2.7 conclusion...................................... 42 3 GeneralConclusion44 Appendices 45 3.1 DownloadandUploadCovid-19Dataset.................... 46 3.2 createandconfiguregooglecolab......................... 47 3.3 Generaldescription................................. 53 Bibliography 54 |
Type de document : | Mémoire master |
Disponibilité (1)
Cote | Support | Localisation | Statut |
---|---|---|---|
MINF/665 | Mémoire master | bibliothèque sciences exactes | Consultable |