Deep learning algorithms applied to dermatological pathology monitoring
Delarbre, François
Promotor(s) : Geurts, Pierre
Date of defense : 9-Sep-2019/10-Sep-2019 • Permalink : http://hdl.handle.net/2268.2/8221
Details
Title : | Deep learning algorithms applied to dermatological pathology monitoring |
Author : | Delarbre, François |
Date of defense : | 9-Sep-2019/10-Sep-2019 |
Advisor(s) : | Geurts, Pierre |
Committee's member(s) : | Marée, Raphaël
Boigelot, Bernard Bours, Jérémie |
Language : | English |
Discipline(s) : | Ingénierie, informatique & technologie > Sciences informatiques |
Target public : | Professionals of domain Other |
Institution(s) : | Université de Liège, Liège, Belgique |
Degree: | Master : ingénieur civil en science des données, à finalité spécialisée |
Faculty: | Master thesis of the Faculté des Sciences appliquées |
Abstract
[en] Nowadays, deep learning techniques are applied in a wide range of problems. With the number of patients and so the numerous data, medical sector is a domain that will benefit from this technology.
In this thesis, we investigate the use of such techniques for dermatology and dermatoscopic applications. More precisely, we apply deep-learning techniques on skin lesion images to monitor nevi and to detect the skin lesions.
To tackle this problem, we first review the literature on deep-learning based solution for semantic segmentation and image retrieval.
We identify the key elements that make a semantic segmentation algorithm performant.
Then, we develop the state of the art algorithms and compare them.
Lastly, we experiment a nevus retrieval algorithm based on our semantic segmentation network.
To train this algorithm, we implement a siamese network and a loss function suitable to this network. we then evaluate the performances of our model.
Based on this analysis, we identify DeepLab as the best model for our use case. We also point out that the image retrieval based on that nevus segmentation algorithm is not efficient.
File(s)
Document(s)
Description: -
Size: 6.66 MB
Format: Adobe PDF
Annexe(s)
Cite this master thesis
The University of Liège does not guarantee the scientific quality of these students' works or the accuracy of all the information they contain.