MARC details
| 000 -LEADER |
|---|
| fixed length control field |
08060nam a22002657a 4500 |
| 008 - FIXED-LENGTH DATA ELEMENTS--GENERAL INFORMATION |
|---|
| fixed length control field |
201210b2025 a|||f bm|| 00| 0 eng d |
| 024 7# - Author Identifier |
|---|
| Standard number or code |
0009-0005-8406-6000 |
| Source of number or code |
ORCID |
| 040 ## - CATALOGING SOURCE |
|---|
| Original cataloging agency |
EG-CaNU |
| Transcribing agency |
EG-CaNU |
| 041 0# - Language Code |
|---|
| Language code of text |
eng |
| Language code of abstract |
eng |
| -- |
ara |
| 082 ## - DEWEY DECIMAL CLASSIFICATION NUMBER |
|---|
| Classification number |
610 |
| 100 0# - MAIN ENTRY--PERSONAL NAME |
|---|
| Personal name |
Ghadir Moemen Helmy Ali |
| 245 1# - TITLE STATEMENT |
|---|
| Title |
Balancing Performance and Explainability in Deep Learning for Lung Disease Detection Using Unimodal and Multimodal Approaches |
| Statement of responsibility, etc. |
/Ghadir Moemen Helmy Ali |
| 260 ## - PUBLICATION, DISTRIBUTION, ETC. |
|---|
| Date of publication, distribution, etc. |
2025 |
| 300 ## - PHYSICAL DESCRIPTION |
|---|
| Extent |
214 p. |
| Other physical details |
ill. |
| Dimensions |
21 cm. |
| 500 ## - GENERAL NOTE |
|---|
| Materials specified |
Supervisor: <br/>Dr. Mustafa Elattar<br/> |
| 502 ## - Dissertation Note |
|---|
| Dissertation type |
Thesis (M.A.)—Nile University, Egypt, 2025 . |
| 504 ## - Bibliography |
|---|
| Bibliography |
"Includes bibliographical references" |
| 505 0# - Contents |
|---|
| Formatted contents note |
Contents:<br/>Contents<br/>Page<br/>Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii<br/>Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix<br/>List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv<br/>List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii<br/>List of Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix<br/>Chapters:<br/>1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1<br/>1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1<br/>1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . 2<br/>1.3 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . 3<br/>1.4 Summary of Contributions . . . . . . . . . . . . . . . . . . . . 5<br/>1.5 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5<br/>2. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7<br/>2.1 Inherent Deep Learning Trade-offs . . . . . . . . . . . . . . . 7<br/>2.2 Explainable AI and Clinical Applicability . . . . . . . . . . . 9<br/>2.3 Lung Diseases Clinical Background . . . . . . . . . . . . . . . 15<br/>2.3.1 Non-Small Cell Lung Carcinoma (NSCLC) . . . . . . . 16<br/>2.3.2 Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . 18<br/>3. Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21<br/>3.1 Explainable AI and Clinical Reasoning . . . . . . . . . . . . . 22<br/>3.1.1 Explainable AI Approaches . . . . . . . . . . . . . . . 25<br/>3.1.2 Evaluation of Explainable AI Methods . . . . . . . . . 34<br/>3.1.3 Limitations and Future Directions . . . . . . . . . . . 42<br/>xi<br/>3.2 Attention Mechanisms in Deep Learning . . . . . . . . . . . . 44<br/>3.2.1 Main Approaches . . . . . . . . . . . . . . . . . . . . . 49<br/>3.2.2 Attention for Explainable AI . . . . . . . . . . . . . . 61<br/>3.3 Pneumonia Classification from Chest X-Ray . . . . . . . . . . 66<br/>3.4 Non-Small Cell Lung Carcinoma . . . . . . . . . . . . . . . . 70<br/>3.4.1 Data Types . . . . . . . . . . . . . . . . . . . . . . . . 73<br/>3.4.2 AI Contribution . . . . . . . . . . . . . . . . . . . . . 83<br/>3.5 Literature Survey Summary . . . . . . . . . . . . . . . . . . . 102<br/>4. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105<br/>4.1 Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106<br/>4.1.1 Dataset and Preprocessing . . . . . . . . . . . . . . . . 107<br/>4.1.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 109<br/>4.2 Non-Small Cell Lung Carcinoma . . . . . . . . . . . . . . . . 113<br/>4.2.1 Dataset Description and Inclusion Criteria . . . . . . . 114<br/>4.2.2 Preprocessing and Featurization . . . . . . . . . . . . . 116<br/>4.2.3 Models . . . . . . . . . . . . . . . . . . . . . . . . . . 121<br/>5. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127<br/>5.1 Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127<br/>5.2 Non-Small Cell Lung Carcinoma . . . . . . . . . . . . . . . . 136<br/>5.2.1 Molecular Features & Model . . . . . . . . . . . . . . . 136<br/>5.2.2 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . 136<br/>6. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143<br/>6.1 Pneumonia . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143<br/>6.2 Non-Small Cell Carcinoma . . . . . . . . . . . . . . . . . . . . 145<br/>7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147<br/>7.1 Key Findings and Contributions . . . . . . . . . . . . . . . . 147<br/>7.1.1 Pneumonia Classification with Attention Mechanisms . 147<br/>7.1.2 Non-Small Cell Carcinoma Metastasis Detection using<br/>Radiogenomics . . . . . . . . . . . . . . . . . . . . . . 148<br/>7.1.3 Explainability and Clinical Usability . . . . . . . . . . 148<br/>7.2 Implications for AI in Healthcare . . . . . . . . . . . . . . . . 149<br/>7.3 Limitations and Future Directions . . . . . . . . . . . . . . . 149<br/>7.4 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . 150<br/>8. Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151<br/>8.1 Multi-Modal Imaging Expansion . . . . . . . . . . . . . . . . 151<br/>8.2 Advanced Multi-Information Fusion Strategies . . . . . . . . . 152<br/>8.3 Dataset Development and Clinical Validation . . . . . . . . . 152<br/>xii<br/>8.4 Toward Human-Centric Explainable AI . . . . . . . . . . . . . 153<br/>Appendices:<br/>A. Publications and Competitions . . . . . . . . . . . . . . . . . . . . 155<br/>Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157<br/>Arabic Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 |
| 520 3# - Abstract |
|---|
| Abstract |
Abstract:<br/>The adoption of deep learning in healthcare applications has highlighted a<br/>critical trade-off between model performance and explainability, particularly in<br/>clinical decision-making contexts. This thesis explores methods to address this<br/>trade-off through two key studies focusing on lung disease detection: pneumonia<br/>classification using chest X-rays and metastatic classification for non-small-cell<br/>lung cancer using CT imaging and genomics.<br/>In our work with pneumonia, attention mechanisms were integrated into<br/>DL models, serving as a balance between post-hoc and inherently explainable<br/>approaches. The use of attention improved classification performance while<br/>enhancing model focus on clinically relevant regions, as demonstrated through<br/>Grad-CAM visualizations. Attention proved to be an effective explainability<br/>choice when chest X-ray was the sole data source.<br/>For the non-small-cell lung cancer study, we adopted a multimodal approach<br/>combining CT imaging and RNA-sequencing data. Ablation studies revealed<br/>a novel clinical insight: surrounding tissue regions were more predictive of<br/>metastasis than tumor regions, aligning with Grad-CAM visualizations that<br/>emphasized tissue edges. This finding correlates with medical knowledge where<br/>tissue involvement signals metastasis progression. Genomic features alone<br/>exhibited limited predictive power but improved model performance when fused<br/>with imaging data. In contrast, handcrafted features, such as shape and size<br/>metrics, showed poor predictive capability. Given access to multimodal data,<br/>vii<br/>this study opens the door for exploring more clinically meaningful explainability<br/>techniques to further align AI model outputs with clinical reasoning.<br/>These results underscore the importance of context-specific explainability<br/>approaches in healthcare AI. While attention mechanisms are effective for<br/>unimodal data like chest X-ray, complex tasks with rich multimodal inputs,<br/>such as non-small-cell lung cancer classification, can benefit from advanced<br/>explainability methods. This thesis contributes to developing clinically explainable AI systems that strike a balance between performance and interpretability,<br/>ensuring their utility in real-world medical applications.<br/>Keywords<br/>Deep Learning, Convolutional Neural Networks, Attention Mechanisms, Explainable Artificial Intelligence, Lung Diseases, Medical Imagining, Bioinformatics,<br/>Radiogenomics, Clinical Applicability |
| 546 ## - Language Note |
|---|
| Language Note |
Text in English, abstracts in English and Arabic |
| 650 #4 - Subject |
|---|
| Subject |
InformaticsIFM |
| 655 #7 - Index Term-Genre/Form |
|---|
| Source of term |
NULIB |
| focus term |
Dissertation, Academic |
| 690 ## - Subject |
|---|
| School |
InformaticsIFM |
| 942 ## - ADDED ENTRY ELEMENTS (KOHA) |
|---|
| Source of classification or shelving scheme |
Dewey Decimal Classification |
| Koha item type |
Thesis |
| 655 #7 - Index Term-Genre/Form |
|---|
| -- |
187 |