DGIST Scholar: Condition and Modality-guided Medical Image Generation using Deep Learning-based Generative Models

Department of Robotics and Mechatronics Engineering Theses Ph.D.

Cited time in webofscience

Cited time in scopus

Condition and Modality-guided Medical Image Generation using Deep Learning-based Generative Models

Title: Condition and Modality-guided Medical Image Generation using Deep Learning-based Generative Models

Alternative Title: 딥러닝 기반 생성 모델을 사용한 조건부 및 모달리티 가이드 의료영상 생성 기법

Author(s): Euijin Jung

DGIST Authors: Euijin Jung ; Sang Hyun Park ; Kyong Hwan Jin

Advisor: 박상현

Co-Advisor(s): Kyong Hwan Jin

Issued Date: 2023

Awarded Date: 2023-08-01

Type: Thesis

Description: Conditional Image Generation; Deep Convolutional Neural Networks; Generative Adversarial Networks; Diffusion Models; MRI; Alzheimer’s disease

Table Of Contents: I. INTRODUCTION 1
1 Background and Motivation 1
2 Main Contributions 1
3 Thesis Outline 1
II. Paired Image-to-Image Translation for Perivascular Spaces Enhancement 3
1 Introduction 3
1.1 Related Works 4
1.2 Contributions 5
2 Methodology 6
2.1 Densely Connected Deep Neural Network 6
2.2 Implementation Details 8
3 Experiments and Results 8
3.1 Data set 8
3.2 Evaluation Settings 11
3.3 Quantitative Results 11
3.4 Qualitative Results 12
3.5 Discussion for comparison networks 12
3.6 Discussion for network depth 13
4 Discussion 14
III. Multi-domain Image-to-Image Translation for Alzheimer's disease progression 18
1 Introduction 18
1.1 Related Works 20
2 Methodology 22
2.1 Objective function 24
3 Experimental settings 26
4 Results 28
4.1 Quantitative results 28
4.2 Qualitative results 31
4.3 Comparison of Subcortical Structures 33
4.4 Ablation study 33
4.5 Computational efficiency 35
5 Discussion 36
IV. Paired Image-to-Image Translation using Guided Diffusion Model 37
1 Introduction 37
2 Methodology 38
2.1 Training Stage 41
2.2 Inference Stage 41
3 Experiments 42
3.1 Implementation details 42
3.2 Quantitative Results 44
3.3 Qualitative Results 45
4 Discussion 45
V. Multi-domain Image-to-Image Translation using Guided Diffusion Model 51
1 Introduction 51
2 Methodology 51
2.1 Training Stage 53
2.2 Inference Stage 53
3 Experiments 53
3.1 Implementation details 53
3.2 Quantitative Results 55
3.3 Qualitative Results 55
4 Discussion 56
VI. Conclusion and Future Directions 58
1 Conclusion 58
2 Future Directions 58
VII. Acknowledgement 60
References 62