|
Titre :
|
Deep learning for segmentation and pathology analysis in medical imaging
|
|
Auteurs :
|
Yamina Azzi, Auteur ;
Mohand Tahar Kechadi, Directeur de thèse
|
|
Type de document :
|
document électronique
|
|
Editeur :
|
Sétif : Université ferhat Abbas faculté des Sciences département des Mathématique, 2024
|
|
ISBN/ISSN/EAN :
|
E-TH/2356
|
|
Format :
|
1vol.(125 f.) / ill.en coul
|
|
Note générale :
|
Bibliogr.
|
|
Langues:
|
Anglais
|
|
Catégories :
|
Thèses (en français - en anglais) > Document électronique
|
|
Mots-clés:
|
Deep learning
;
Class imbalance
;
Medical imaging
|
|
Résumé :
|
Since the advent of deep learning, there has been a technological revolution, endowing computers with remarkable abilities in learning, reasoning, thinking, and decision making. Deep learning has played a pivotal role in facilitating feature extraction, particularly in the realm of computer vision, where it significantly contributes to learning and recognizing image and video content.This impact extends to the field of medical image research,as demonstrated in this thesis.The focus is on the role of deep learning in the precise automatic segmentation of medical images, specifically in segmenting brain glioma tumors and their associated sub-tumoral tissue from multimodal MRI scans. The thesis presents various contributions aimed at achieving accurate segmentation, including a comparative study between machine learning-based methods and deep learning architectures. Additionally, it sheds light on the substantial influence of class imbalance on image segmentation evaluation metrics.Lastly, the thesis proposes a transfer learning-based method to construct a model capable of classifying three types of brain tumors: Meningioma, glioma, and pituitary tumors.
|
|
En ligne :
|
http://dspace.univ-setif.dz:8888/jspui/bitstream/123456789/4418/1/these.pdf
|
Exemplaires (1)
|
| E-TH/2356 | Thèse | Bibliothèque centrale | Disponible |
Accueil
