| Titre : | AI-Powered Diagnosis Tool for Advanced Spatio-Temporal Segmentation of Cardiac MRI Images |
| Auteurs : | Nouha Benzine, Auteur ; Tebra Abbassi, Auteur ; Asma Bendahmane, Auteur |
| Type de document : | Mémoire magistere |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2025 |
| Format : | 1 vol. (77 p.) / ill.couv.ill.encoul / 30cm |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Résumé : |
Accurate quantification of the Right Ventricle (RV) in cardiac magnetic resonance imaging (MRI) is clinically crucial, as it plays an important role in diagnosing various cardiac diseases. Conventionally, this quantification relies on manual segmentations, which involve outlining the endocardial contours across multiple MRI slices. However, this process is time-consuming and labor-intensive, often taking over thirty minutes per patient. Thus, moving toward automated approaches is significantly suitable. Despite the substantial progress in imaging technologies and Deep Learning (DL) methods, (RV) segmentation using MRI slices remains a challenging task. Although several studies have explored automated segmentation techniques, much of the existing methods achieves satisfactory results primarily on central slices of short-axis MRI scans. However,basal and apical slices remain difficult to be segmented accurately due to their ambiguousanatomical boundaries. These slices often suffer from poor contrast and structural discontinuity, which significantly hinders the generalization capability of existing models.The current study aims to tackle the above-mentioned issues by investigating a hybridspatially guided approach based on DL techniques, with a specific focus on enhancing performance over basal and apical slice levels. Specifically, a Long Short-Term Memory (LSTM)-based mechanism is incorporated to capture spatial relationships between adjacent slices. This progressive strategy is initialized by an optimized DL-based model that combines Vision-Transformers (ViT) encoder into the U-Net down-sampling path. Accordingly, the obtained results exhibit a significant improvement among basal and apical slices.Furthermore, to ensure clinical applicability and maintain human oversight, a personalized platform is elaborated using the Django framework. The platform is designed to automatically generate RV edges while providing clinicians with the ability to visually review and manually adjust the segmentation results as required. keywords:Right Ventricle, basal slices, apical slices, Magnetic Resonance Imagin. |
| Sommaire : |
List of Figures X
List of Tables XII Abbreviations XIII General introduction 1 1 Medical Background 4 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2 Right Ventricular Anatomy . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.1 Anatomy and Positions . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.2.2 Shape and Structure . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Right Ventricle Related Pathologies . . . . . . . . . . . . . . . . . . . . . . 8 1.3.1 Right Ventricular Hypertrophy (RVH) . . . . . . . . . . . . . . . . 8 1.3.2 Right Ventricular Failure (RVF) . . . . . . . . . . . . . . . . . . . . 9 1.3.3 Arrhythmogenic Right Ventricular Cardiomyopathy (ARVC) . . . . 10 1.3.4 Pulmonary Hypertension (PH) and Its Impact on the Right Ventricle 10 1.3.5 Tetralogy of Fallot (TOF) . . . . . . . . . . . . . . . . . . . . . . . 11 1.4 Modalities for cardiac Imging . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.4.1 Chest X-ray (CXR) . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.4.2 Echocardiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.4.3 Myocardial Perfusion Imaging (MPI) . . . . . . . . . . . . . . . . . 13 1.4.4 Computed Tomography (CT) . . . . . . . . . . . . . . . . . . . . . 13 1.4.5 Angiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.4.6 Magnetic Resonance Imaging (MRI) . . . . . . . . . . . . . . . . . 13 1.5 Aquisition Short Axis Transition . . . . . . . . . . . . . . . . . . . . . . . . 14 1.6 Clinical practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.7 Challenging Slices in Clinical Quantitative Assessment . . . . . . . . . . . 18 1.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2 Bibliographic Study 21 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2 Overview on Image Segmentation Methods . . . . . . . . . . . . . . . . . . 21 2.2.1 Conventional Methods . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.2.2 Machine Learning Methods . . . . . . . . . . . . . . . . . . . . . . 22 2.2.3 Other Advanced and Hybrid Methods . . . . . . . . . . . . . . . . . 23 2.3 Right Ventricle Segmentation Review Methods . . . . . . . . . . . . . . . . 24 2.3.1 Conventional Methods . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3.2 Machine Learning Methods . . . . . . . . . . . . . . . . . . . . . . 25 2.3.3 Hybrid Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3 Initial Model Selection Through Comparative Analysis 34 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.2 Dataset Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.3.1 U-Net for Image Segmentation . . . . . . . . . . . . . . . . . . . . . 36 3.3.2 Vision Transformer for Image Segmentation . . . . . . . . . . . . . 37 3.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.4.1 Training phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.4.2 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.5 Evaluation of Model Performance Across Spatial Slices . . . . . . . . . . . 45 3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4 Progressive Spatial-Temporal Segmentation Approach 48 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 4.2 Methods and Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 4.2.1 LSTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.2.2 Dataset Challenges and Handling . . . . . . . . . . . . . . . . . . . 51 4.3 Proposed Progressive RV Segmentation Method . . . . . . . . . . . . . . . 52 4.4 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4.5 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.6 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 5 Platform Design and Implementation 59 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.2 General Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.3 Platform Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 5.3.1 Principal Actors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 5.3.2 Functional requirements . . . . . . . . . . . . . . . . . . . . . . . . 61 5.4 Unified modeling language(UML) . . . . . . . . . . . . . . . . . . . . . . . 63 5.4.1 Use Case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.4.2 Class Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 5.4.3 Sequence diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 5.5 Integrating the Pre-Trained Model to the . . . . . . . . . . . . . . . . . . . 68 5.6 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 5.6.1 Hardware and Software Requirements: . . . . . . . . . . . . . . . . 69 5.6.2 Development Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 5.7 Exhibition Of The Developed Functionalities (Graphical User Interface (GUI)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 5.8 Navigation diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 5.8.1 Home page and login . . . . . . . . . . . . . . . . . . . . . . . . . . 78 5.8.2 Doctor panel pages . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 5.8.3 Secretary panel pages . . . . . . . . . . . . . . . . . . . . . . . . . . 79 5.8.4 Center Admin panel pages . . . . . . . . . . . . . . . . . . . . . . . 79 5.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 General conclusion 81 |
| Type de document : | Mémoire master |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/935 | Mémoire master | bibliothèque sciences exactes | Consultable |
