Advancing X-ray imaging with deep learning : Physics-inspired reconstruction approaches
(2024)- Abstract
- The development of high-brilliance X-ray sources, such as the fourth-generation diffraction-limited storage rings and X-ray free-electron lasers, have opened up new possibilities for X-ray imaging and pushed the temporal resolutions of imaging techniques to unprecedented levels. Capturing fast dynamics in two-dimensional (2D), three-dimensional (3D), and even four-dimensional (4D, 3D + time) beyond microsecond temporal resolution has become possible. To fully exploit the unique capabilities of these facilities, challenges such as the data problem must be addressed. Automated tools are needed to handle the large amount of data acquired from each experiment.
As a data-driven approach, deep learning has undergone rapid development over... (More) - The development of high-brilliance X-ray sources, such as the fourth-generation diffraction-limited storage rings and X-ray free-electron lasers, have opened up new possibilities for X-ray imaging and pushed the temporal resolutions of imaging techniques to unprecedented levels. Capturing fast dynamics in two-dimensional (2D), three-dimensional (3D), and even four-dimensional (4D, 3D + time) beyond microsecond temporal resolution has become possible. To fully exploit the unique capabilities of these facilities, challenges such as the data problem must be addressed. Automated tools are needed to handle the large amount of data acquired from each experiment.
As a data-driven approach, deep learning has undergone rapid development over the past decade and offers a promising solution to this problem. However, state-of-the-art deep learning methods applied to X-ray imaging ignore the physics of X-ray propagation and interaction with matter and require paired training datasets. In this thesis, we show that combining the physical principles of X-ray imaging with deep learning greatly improves the performance and robustness of the approaches, and it is possible to construct reliable unsupervised approaches, where no paired datasets are needed.
Firstly, we present a theoretical background on X-ray imaging and various imaging methods. Secondly, we provide an overview of deep learning, including training strategies and common frameworks for addressing imaging tasks. Lastly, we introduce novel algorithms developed during this thesis:
1. FFCGAN, a supervised approach for shot-to-shot flat-field correction at X-ray free-electron lasers.
2. PhaseGAN, a phase-retrieval approach for unpaired datasets.
3. ONIX, a self-supervised approach for 3D reconstruction from sparse views.
4. 4D-ONIX, a self-supervised approach for reconstructing 3D movies from sparse projections.
These approaches offer high-quality image reconstructions for X-ray imaging techniques, enabling further exploration and understanding of the structure and dynamic properties of various samples. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/400002b7-a201-42d7-ae4c-662bbe7b5d68
- author
- Zhang, Yuhe LU
- supervisor
- opponent
-
- Professor Jacobsen, Chris, Northwestern University.
- organization
- publishing date
- 2024-05-20
- type
- Thesis
- publication status
- published
- subject
- keywords
- X-ray microscopy, X-ray imaging, Deep learning, Artificial intellgience, Phase contrast, Tomography, X-ray, X-ray multi-projection imaging, 4D imaging, 3D imaging
- pages
- 76 pages
- publisher
- Department of Process and Life Science Engineering, Lund University.
- defense location
- Rydberg Lecture Hall, Department of Physics
- defense date
- 2024-06-14 09:15:00
- ISBN
- 978-91-8039-983-8
- 978-91-8039-982-1
- language
- English
- LU publication?
- yes
- id
- 400002b7-a201-42d7-ae4c-662bbe7b5d68
- date added to LUP
- 2024-05-20 14:20:32
- date last changed
- 2024-05-24 08:48:17
@phdthesis{400002b7-a201-42d7-ae4c-662bbe7b5d68,
abstract = {{The development of high-brilliance X-ray sources, such as the fourth-generation diffraction-limited storage rings and X-ray free-electron lasers, have opened up new possibilities for X-ray imaging and pushed the temporal resolutions of imaging techniques to unprecedented levels. Capturing fast dynamics in two-dimensional (2D), three-dimensional (3D), and even four-dimensional (4D, 3D + time) beyond microsecond temporal resolution has become possible. To fully exploit the unique capabilities of these facilities, challenges such as the data problem must be addressed. Automated tools are needed to handle the large amount of data acquired from each experiment.<br/>As a data-driven approach, deep learning has undergone rapid development over the past decade and offers a promising solution to this problem. However, state-of-the-art deep learning methods applied to X-ray imaging ignore the physics of X-ray propagation and interaction with matter and require paired training datasets. In this thesis, we show that combining the physical principles of X-ray imaging with deep learning greatly improves the performance and robustness of the approaches, and it is possible to construct reliable unsupervised approaches, where no paired datasets are needed.<br/><br/>Firstly, we present a theoretical background on X-ray imaging and various imaging methods. Secondly, we provide an overview of deep learning, including training strategies and common frameworks for addressing imaging tasks. Lastly, we introduce novel algorithms developed during this thesis:<br/><br/>1. FFCGAN, a supervised approach for shot-to-shot flat-field correction at X-ray free-electron lasers.<br/>2. PhaseGAN, a phase-retrieval approach for unpaired datasets.<br/>3. ONIX, a self-supervised approach for 3D reconstruction from sparse views.<br/>4. 4D-ONIX, a self-supervised approach for reconstructing 3D movies from sparse projections.<br/><br/>These approaches offer high-quality image reconstructions for X-ray imaging techniques, enabling further exploration and understanding of the structure and dynamic properties of various samples.}},
author = {{Zhang, Yuhe}},
isbn = {{978-91-8039-983-8}},
keywords = {{X-ray microscopy; X-ray imaging; Deep learning; Artificial intellgience; Phase contrast; Tomography; X-ray; X-ray multi-projection imaging; 4D imaging; 3D imaging}},
language = {{eng}},
month = {{05}},
publisher = {{Department of Process and Life Science Engineering, Lund University.}},
school = {{Lund University}},
title = {{Advancing X-ray imaging with deep learning : Physics-inspired reconstruction approaches}},
url = {{https://lup.lub.lu.se/search/files/183801628/Yuhe_s_Thesis_intro_.pdf}},
year = {{2024}},
}