Titre :	Classification and Segmentation of COVID-19 CXR and Chest CT Imag;es Using Deep Learning Algorithms
Type de document :	texte imprimé
Auteurs :	Abdesselam Ferdi, Auteur ; Said Benierbah, Directeur de thèse
Editeur :	CONSTANTINE [ALGERIE] : Université Frères Mentouri Constantine
Année de publication :	2020
Importance :	77 f.
Format :	30cm.
Note générale :	Une copie electronique PDF disponible au BUC.
Langues :	Anglais (eng)
Catégories :	Lettres et Langues Etrangères:Langue Anglaise
Tags :	COVID-19  Convolutional Network  Chest X-Ray  Chest Computed Tomography  Classification  Deep Learning  Fully Convolutional Network  Medical image processing  Semantic Segmentation  Transfer Learning.
Index. décimale :	420 Langue anglaise
Résumé :	In recent years, deep learning algorithms have a tremendous growth in their popularity and use in many fields, such as medical imaging. In this work, we will use these algorithms to help in the diagnosis of the COVID 19. Although RT-PCR testing is considered the gold standard for COVID-19 screening, diagnosis of this disease by processing chest x-ray and chest CT images is also widely used nowadays. The work carried out in this thesis has two parts. The first part is devoted to classification and the second part deals with segmentation. Four deep learning algorithms based on convolutional networks has been used in the task of COVID-19 chest xray images classification. Three of them are pre-trained networks, namely, ResNet-50, Inception-v3, and Inception-ResNet-v2, and the fourth one is our proposed deep learning network. In the segmentation task, three algorithms based on fully convolutional networks has been used to segment the chest CT images of COVID-19 into three regions of lung infection (ground-glass opacity, consolidation, and pleural effusion). Two algorithms has been taken from literature, namely, U-Net and SegNet, and the third one is our proposed deep learning model. The classification results show that our proposed model COVID-Net has yielded the highest classification performance as the ResNet-50 does with a validation accuracy of 99%. In the case of segmentation, the results show that the SegNet model have yielded the highest dice score of 0.69
Diplome :	Master 2
Permalink :	https://bu.umc.edu.dz/master/index.php?lvl=notice_display&id=13551

Classification and Segmentation of COVID-19 CXR and Chest CT Imag;es Using Deep Learning Algorithms [texte imprimé] / Abdesselam Ferdi, Auteur ; Said Benierbah, Directeur de thèse . - CONSTANTINE [ALGERIE] : Université Frères Mentouri Constantine, 2020 . - 77 f. ; 30cm.
Une copie electronique PDF disponible au BUC.
Langues : Anglais (eng)

Catégories :	Lettres et Langues Etrangères:Langue Anglaise
Tags :	COVID-19  Convolutional Network  Chest X-Ray  Chest Computed Tomography  Classification  Deep Learning  Fully Convolutional Network  Medical image processing  Semantic Segmentation  Transfer Learning.
Index. décimale :	420 Langue anglaise
Résumé :	In recent years, deep learning algorithms have a tremendous growth in their popularity and use in many fields, such as medical imaging. In this work, we will use these algorithms to help in the diagnosis of the COVID 19. Although RT-PCR testing is considered the gold standard for COVID-19 screening, diagnosis of this disease by processing chest x-ray and chest CT images is also widely used nowadays. The work carried out in this thesis has two parts. The first part is devoted to classification and the second part deals with segmentation. Four deep learning algorithms based on convolutional networks has been used in the task of COVID-19 chest xray images classification. Three of them are pre-trained networks, namely, ResNet-50, Inception-v3, and Inception-ResNet-v2, and the fourth one is our proposed deep learning network. In the segmentation task, three algorithms based on fully convolutional networks has been used to segment the chest CT images of COVID-19 into three regions of lung infection (ground-glass opacity, consolidation, and pleural effusion). Two algorithms has been taken from literature, namely, U-Net and SegNet, and the third one is our proposed deep learning model. The classification results show that our proposed model COVID-Net has yielded the highest classification performance as the ResNet-50 does with a validation accuracy of 99%. In the case of segmentation, the results show that the SegNet model have yielded the highest dice score of 0.69
Diplome :	Master 2
Permalink :	https://bu.umc.edu.dz/master/index.php?lvl=notice_display&id=13551