|
| Titre : |
”Application of Deep Learning for the Automated Detection and Classification of Medical Pathologies from High-Resolution Medical Images” |
| Type de document : |
document électronique |
| Auteurs : |
Mohamed Abdessamed Krache, Auteur ; Chouaib Maiza, Auteur ; Khababa,Abdellah, Directeur de thèse |
| Editeur : |
Sétif:UFA1 |
| Année de publication : |
2024 |
| Importance : |
1 vol (72 f .) |
| Format : |
29 cm |
| Langues : |
Français (fre) |
| Catégories : |
Thèses & Mémoires:Informatique
|
| Mots-clés : |
Deep learning
Machine learning
Artificial intelligence
Diabetic retinopathy (DR)
DenseNet121 architecture
MobileNet architecture |
| Index. décimale : |
004 Informatique |
| Résumé : |
Diabetic Retinopathy (DR) is an eye condition that affects individuals with diabetes, causing damage
to the retina and potentially leading to blindness. Traditionally, DR screening is performed manually by
ophthalmologists, which is a time-consuming process. This project aims to analyze various stages of DR
using Deep Learning (DL), a subset of artificial intelligence (AI) and machine learning that utilizes artificial
neural networks for high accuracy in medical image analysis.
We trained a models on a large dataset consisting of approximately 3662 training images to automatically
detect the stages of DR. These stages are classified using high-resolution fundus images. The dataset
used is available on Kaggle (APTOS). DR stages are categorized into five levels: 0, 1, 2, 3, and 4. In this
study, patient fundus eye images serve as the input parameters. Trained models, including DenseNet121,
MobileNet, and Xception architectures, extract features from the fundus images, and an activation function
produces the output.
The DenseNet121 architecture achieved an accuracy of 0.95 with a cohen kappa score of 0.968. MobileNet
attained an accuracy of 0.93 with a cohen kappa score of 0.949, while Xception achieved an accuracy of
0.82 in DR detection. Consequently, we propose an application that facilitates early diagnosis of diabetic
retinopathy, assisting doctors in making quick and accurate diagnoses. Deep learning algorithms, specifically
neural networks for image classification, were employed to distinguish between healthy and affected eyes and
to identify the type of injury. |
| Note de contenu : |
Sommaire
1 Image and Medical Imaging 14
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2 Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3 Digital image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.2 Digital image characteristics . . . . . . . . . . . . . . . . . . . . . . 16
1.3.3 Types of digital image . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4 Differences between Raster and Vector graphics . . . . . . . . . . . . . . . . 21
1.5 Medical Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6 Types of Medical Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.1 X-rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.2 Radiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.3 Mammography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.4 Magnetic Resonance Imaging (MRI) . . . . . . . . . . . . . . . . . . 24
1.6.5 Tomography (Scanner) . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.6 Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.7 Scintigraphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Diabetic Retinopathy 27
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2 Definition of the eye . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Diabetic retinopathy (DR) . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 The acquisition systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.1 Fluorescence angiography . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 Retinograph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 The context of DR in informatics . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6 Artificial Intelligence (AI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.7 Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Model Architecture of Diabetic Retinopathy Analysis . . . . . . . . . . . . . 34
2.8.1 Convolutional Neural Network(CNN) . . . . . . . . . . . . . . . . . . 34
2.8.2 Neural Networks with Multilayer Perceptron . . . . . . . . . . . . . . 34
2.8.3 Deep Neural Network (DNN) . . . . . . . . . . . . . . . . . . . . . . 35
2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 Discussion and Results 37
3.1 Intoduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Summary of Research Papers on Diabetic Retinopathy Classification 38
3.3 Experience and results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.3 Experimental environment . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.4 Classification Performance Measures . . . . . . . . . . . . . . . . . . 47
3.3.5 Evaluation Metrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.6 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3.7 DenseNet 121 Architecture . . . . . . . . . . . . . . . . . . . . . . . 49
3.3.8 MobileNet Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.9 Xception Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.10 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Presentation of the Application 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 Graphic User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.1 The first interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 The second interface . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3 No DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.4 Mild . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.5 Moderate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.6 Severe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.7 Proliferative DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 |
| Côte titre : |
MAI/0944 |
”Application of Deep Learning for the Automated Detection and Classification of Medical Pathologies from High-Resolution Medical Images” [document électronique] / Mohamed Abdessamed Krache, Auteur ; Chouaib Maiza, Auteur ; Khababa,Abdellah, Directeur de thèse . - [S.l.] : Sétif:UFA1, 2024 . - 1 vol (72 f .) ; 29 cm. Langues : Français ( fre)
| Catégories : |
Thèses & Mémoires:Informatique
|
| Mots-clés : |
Deep learning
Machine learning
Artificial intelligence
Diabetic retinopathy (DR)
DenseNet121 architecture
MobileNet architecture |
| Index. décimale : |
004 Informatique |
| Résumé : |
Diabetic Retinopathy (DR) is an eye condition that affects individuals with diabetes, causing damage
to the retina and potentially leading to blindness. Traditionally, DR screening is performed manually by
ophthalmologists, which is a time-consuming process. This project aims to analyze various stages of DR
using Deep Learning (DL), a subset of artificial intelligence (AI) and machine learning that utilizes artificial
neural networks for high accuracy in medical image analysis.
We trained a models on a large dataset consisting of approximately 3662 training images to automatically
detect the stages of DR. These stages are classified using high-resolution fundus images. The dataset
used is available on Kaggle (APTOS). DR stages are categorized into five levels: 0, 1, 2, 3, and 4. In this
study, patient fundus eye images serve as the input parameters. Trained models, including DenseNet121,
MobileNet, and Xception architectures, extract features from the fundus images, and an activation function
produces the output.
The DenseNet121 architecture achieved an accuracy of 0.95 with a cohen kappa score of 0.968. MobileNet
attained an accuracy of 0.93 with a cohen kappa score of 0.949, while Xception achieved an accuracy of
0.82 in DR detection. Consequently, we propose an application that facilitates early diagnosis of diabetic
retinopathy, assisting doctors in making quick and accurate diagnoses. Deep learning algorithms, specifically
neural networks for image classification, were employed to distinguish between healthy and affected eyes and
to identify the type of injury. |
| Note de contenu : |
Sommaire
1 Image and Medical Imaging 14
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2 Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3 Digital image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.2 Digital image characteristics . . . . . . . . . . . . . . . . . . . . . . 16
1.3.3 Types of digital image . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4 Differences between Raster and Vector graphics . . . . . . . . . . . . . . . . 21
1.5 Medical Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6 Types of Medical Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.1 X-rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.2 Radiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.3 Mammography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.4 Magnetic Resonance Imaging (MRI) . . . . . . . . . . . . . . . . . . 24
1.6.5 Tomography (Scanner) . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.6 Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.7 Scintigraphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Diabetic Retinopathy 27
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2 Definition of the eye . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Diabetic retinopathy (DR) . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 The acquisition systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.1 Fluorescence angiography . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 Retinograph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 The context of DR in informatics . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6 Artificial Intelligence (AI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.7 Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Model Architecture of Diabetic Retinopathy Analysis . . . . . . . . . . . . . 34
2.8.1 Convolutional Neural Network(CNN) . . . . . . . . . . . . . . . . . . 34
2.8.2 Neural Networks with Multilayer Perceptron . . . . . . . . . . . . . . 34
2.8.3 Deep Neural Network (DNN) . . . . . . . . . . . . . . . . . . . . . . 35
2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 Discussion and Results 37
3.1 Intoduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Summary of Research Papers on Diabetic Retinopathy Classification 38
3.3 Experience and results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.3 Experimental environment . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.4 Classification Performance Measures . . . . . . . . . . . . . . . . . . 47
3.3.5 Evaluation Metrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.6 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3.7 DenseNet 121 Architecture . . . . . . . . . . . . . . . . . . . . . . . 49
3.3.8 MobileNet Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.9 Xception Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.10 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Presentation of the Application 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 Graphic User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.1 The first interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 The second interface . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3 No DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.4 Mild . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.5 Moderate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.6 Severe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.7 Proliferative DR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 |
| Côte titre : |
MAI/0944 |
|
Ajouter le résultat dans votre panier Affiner la recherche”Application of Deep Learning for the Automated Detection and Classification of Medical Pathologies from High-Resolution Medical Images” / Mohamed Abdessamed Krache
