|
| Titre : |
Generelized Anatomical Landmark Detection In Cephalometric: A Comparative Study |
| Type de document : |
document électronique |
| Auteurs : |
Sonia Nedjar, Auteur ; Amira Guermache ; Hacene Azizi, Directeur de thèse |
| Editeur : |
Setif:UFA |
| Année de publication : |
2024 |
| Importance : |
1 vol (86 f.) |
| Format : |
29 cm |
| Langues : |
Anglais (eng) |
| Catégories : |
Thèses & Mémoires:Physique
|
| Mots-clés : |
Anatomical landmarks
Cephalometric
Landmark detection
Automatic identification
U-net
GU2net |
| Index. décimale : |
530 - Physique |
| Résumé : |
The detection of anatomical landmarks on lateral cephalograms is crucial for
precise cephalometric analysis. These landmarks, representing specific structures,
are essential for orthodontic diagnosis, treatment planning, and outcome evaluation.
However, manual marking of these landmarks is a tedious, time-consuming, subjective,
and laborious task. Given these limitations, it is imperative to develop automatic and
accurate methods for landmark detection. The objective of our work is to compare
the results obtained with different architectures, to assist physicians in registration
and segmentation practices. Our approach is based on the U-net architecture,
including U-net2D and GU2net, trained for automatic identification of landmarks. To
conduct this study, we used databases containing radiographic images. Gu2Net model
demonstrated better results over Unet2d in terms of MRE values for all datasets,
for the cephalometric dataset the Gu2Net scored 1.16±1.02 mm, Unet2d scored
18.60±32.02 mm, similar to the previous dataset, hand and chest datasets scored
respectfully 0.81±4.00 mm and 11.81±48.03 mm with GU2Net, and for the Unet2d
2.04±12.33 mm and 56.85±135.63 mm. |
| Note de contenu : |
Sommaire
Aknowledgments i
Abstract ii
R´esum´e iii
Contents iv
List of Figures vii
List of Tables xi
List of Abbreviations xii
Introduction 1
1 LANDMARKS IN CEPHALOMETRIC 4
1.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Landmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Definition Of Landmarks . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Types Of Landmarks . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Cephalometrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Imaging Modalities . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Landmark Identification In Cephalometric . . . . . . . . . . . . . . . . 11
1.5 Application Of Landmark Identification In Cephalometrics . . . . . . . 12
1.5.1 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.2 Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.3 Clinical Application Of Landmark Detection In Radiotherapy . 13
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 AUTOMATIC CEPHALOMETRIC LANDMARK DETECTION 16
2.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Automated Landmark Detection Using AI . . . . . . . . . . . . . . . . 16
2.2.1 Software-Based Landmark Identification . . . . . . . . . . . . . 17
2.2.2 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Automated Landmark Representation . . . . . . . . . . . . . . . . . . 23
2.3.1 Coordinate Regression-Based Methods . . . . . . . . . . . . . . 24
2.3.2 Heatmap Regression-Based Methods . . . . . . . . . . . . . . . 24
2.3.3 Landmark Detection With Graphs . . . . . . . . . . . . . . . . 25
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3 MATERIALS & METHODS 27
3.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Dataset & Implementation Details . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Data-set Description . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2 Data-set Pre-processing . . . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Implementation Details . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Software & Hardware Implementation . . . . . . . . . . . . . . . . . . . 32
3.3.1 Local GPU Description . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.2 Uploading Required Libraries On Google Colab . . . . . . . . . 34
3.3.3 Importing Images To Google Colab . . . . . . . . . . . . . . . . 34
3.3.4 Installing Python Packages With Specific Versions . . . . . . . . 34
3.4 Model Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.1 Unet2d Architecture . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.2 GU2Net Architecture . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.3 Loss Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.4 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.5 Activation function . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.6 Evaluations Metrics . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.5 Train , Validation & Test . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.1 Overfitting & Underfitting . . . . . . . . . . . . . . . . . . . . . 50
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 EXPERIMENTS, RESULTS & DISCUSSION 52
4.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2 Protocol Of Training Models . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.1 Qualitative Comparison . . . . . . . . . . . . . . . . . . . . . . 53
4.3.2 Quantitative Comparison . . . . . . . . . . . . . . . . . . . . . . 66
4.4 Training & Test Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.5 Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
CONCLUSION & PERCPECTIVES 74 |
| Côte titre : |
MAPH/0627 |
Generelized Anatomical Landmark Detection In Cephalometric: A Comparative Study [document électronique] / Sonia Nedjar, Auteur ; Amira Guermache ; Hacene Azizi, Directeur de thèse . - [S.l.] : Setif:UFA, 2024 . - 1 vol (86 f.) ; 29 cm. Langues : Anglais ( eng)
| Catégories : |
Thèses & Mémoires:Physique
|
| Mots-clés : |
Anatomical landmarks
Cephalometric
Landmark detection
Automatic identification
U-net
GU2net |
| Index. décimale : |
530 - Physique |
| Résumé : |
The detection of anatomical landmarks on lateral cephalograms is crucial for
precise cephalometric analysis. These landmarks, representing specific structures,
are essential for orthodontic diagnosis, treatment planning, and outcome evaluation.
However, manual marking of these landmarks is a tedious, time-consuming, subjective,
and laborious task. Given these limitations, it is imperative to develop automatic and
accurate methods for landmark detection. The objective of our work is to compare
the results obtained with different architectures, to assist physicians in registration
and segmentation practices. Our approach is based on the U-net architecture,
including U-net2D and GU2net, trained for automatic identification of landmarks. To
conduct this study, we used databases containing radiographic images. Gu2Net model
demonstrated better results over Unet2d in terms of MRE values for all datasets,
for the cephalometric dataset the Gu2Net scored 1.16±1.02 mm, Unet2d scored
18.60±32.02 mm, similar to the previous dataset, hand and chest datasets scored
respectfully 0.81±4.00 mm and 11.81±48.03 mm with GU2Net, and for the Unet2d
2.04±12.33 mm and 56.85±135.63 mm. |
| Note de contenu : |
Sommaire
Aknowledgments i
Abstract ii
R´esum´e iii
Contents iv
List of Figures vii
List of Tables xi
List of Abbreviations xii
Introduction 1
1 LANDMARKS IN CEPHALOMETRIC 4
1.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Landmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Definition Of Landmarks . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Types Of Landmarks . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Cephalometrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Imaging Modalities . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Landmark Identification In Cephalometric . . . . . . . . . . . . . . . . 11
1.5 Application Of Landmark Identification In Cephalometrics . . . . . . . 12
1.5.1 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.2 Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.3 Clinical Application Of Landmark Detection In Radiotherapy . 13
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 AUTOMATIC CEPHALOMETRIC LANDMARK DETECTION 16
2.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Automated Landmark Detection Using AI . . . . . . . . . . . . . . . . 16
2.2.1 Software-Based Landmark Identification . . . . . . . . . . . . . 17
2.2.2 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 Deep Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Automated Landmark Representation . . . . . . . . . . . . . . . . . . 23
2.3.1 Coordinate Regression-Based Methods . . . . . . . . . . . . . . 24
2.3.2 Heatmap Regression-Based Methods . . . . . . . . . . . . . . . 24
2.3.3 Landmark Detection With Graphs . . . . . . . . . . . . . . . . 25
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3 MATERIALS & METHODS 27
3.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Dataset & Implementation Details . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Data-set Description . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2 Data-set Pre-processing . . . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Implementation Details . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Software & Hardware Implementation . . . . . . . . . . . . . . . . . . . 32
3.3.1 Local GPU Description . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.2 Uploading Required Libraries On Google Colab . . . . . . . . . 34
3.3.3 Importing Images To Google Colab . . . . . . . . . . . . . . . . 34
3.3.4 Installing Python Packages With Specific Versions . . . . . . . . 34
3.4 Model Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.1 Unet2d Architecture . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4.2 GU2Net Architecture . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.3 Loss Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.4 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.5 Activation function . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.6 Evaluations Metrics . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.5 Train , Validation & Test . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.1 Overfitting & Underfitting . . . . . . . . . . . . . . . . . . . . . 50
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 EXPERIMENTS, RESULTS & DISCUSSION 52
4.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2 Protocol Of Training Models . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.1 Qualitative Comparison . . . . . . . . . . . . . . . . . . . . . . 53
4.3.2 Quantitative Comparison . . . . . . . . . . . . . . . . . . . . . . 66
4.4 Training & Test Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.5 Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
CONCLUSION & PERCPECTIVES 74 |
| Côte titre : |
MAPH/0627 |
|
Ajouter le résultat dans votre panier Affiner la recherche
