Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Advancing X-ray Multi-Projection Imaging : Towards Four-Dimensional Reconstruction and Analysis of Rapid Flows

Yao, Zisheng LU orcid (2026)
Abstract
Fourth-generation synchrotron light sources deliver flux densities up to two orders of magnitude higher than previous generations, enabling X-ray imaging at unprecedented spatiotemporal resolution. This is attractive for non-destructive 4D (3D + time) studies of fast dynamics under in-situ or operando conditions. State-of-the-art 4D X-ray imaging at large-scale facilities relies on tomography and can achieve 1000 volume reconstructions per second at micrometer resolution. However, the requirement of rapid sample rotation induces large centrifugal forces, limiting studies of rotation-sensitive dynamics. To address this limitation, X-ray Multi-Projection Imaging (XMPI) has emerged as an alternative rotation-free approach. By splitting the... (More)
Fourth-generation synchrotron light sources deliver flux densities up to two orders of magnitude higher than previous generations, enabling X-ray imaging at unprecedented spatiotemporal resolution. This is attractive for non-destructive 4D (3D + time) studies of fast dynamics under in-situ or operando conditions. State-of-the-art 4D X-ray imaging at large-scale facilities relies on tomography and can achieve 1000 volume reconstructions per second at micrometer resolution. However, the requirement of rapid sample rotation induces large centrifugal forces, limiting studies of rotation-sensitive dynamics. To address this limitation, X-ray Multi-Projection Imaging (XMPI) has emerged as an alternative rotation-free approach. By splitting the primary beam into several angularly
separated beams, XMPI enables simultaneous multi-angle acquisitions, enabling ultrafast imaging but making 4D reconstruction and analysis challenging in the sparse-view regime.

In this thesis, we first introduce the basics of X-ray imaging. Second, we present the implementation of the XMPI setup, enabled by the unique capabilities of fourth-generation synchrotron light sources. Third, we present the development of 3D particle tracking velocimetry with XMPI, and demonstrate its capability to quantitatively retrieve flow properties by tracking microscale tracers via triangulation. Lastly, we focus on a more ambitious goal: to develop full-field deep-learning-based 4D reconstruction methods for XMPI, including droplet collisions and flows through porous media, two representative fluid-dynamics examples for validation. This thesis demonstrates that XMPI, together with advanced 4D reconstruction and analysis approaches, becomes an enabling tool for fluid dynamics studies, providing sub-10 micrometer and kHz-to-MHz spatiotemporal resolution. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Associate Professor Boone, Matthieu, Department of Physics and Astronomy, Ghent University, Belgium.
organization
publishing date
type
Thesis
publication status
published
subject
keywords
X-ray imaging, four-dimensional, deep learning, sparse-view reconstruction, ultrafast dynamics
pages
248 pages
publisher
Lund University
defense location
Rydberg Hall, Department of Physics, Lund University Join via zoom: https://lu-se.zoom.us/j/69303679382?pwd=PHNQVVirUGTzTIFYIsOpynQrpQm52O.1
defense date
2026-05-22 09:00:00
ISBN
978-91-8104-771-4
978-91-8104-772-1
language
English
LU publication?
yes
id
4e03d3e8-805d-4ac8-a420-981480b7276b
date added to LUP
2026-04-14 14:36:24
date last changed
2026-04-29 03:24:07
@phdthesis{4e03d3e8-805d-4ac8-a420-981480b7276b,
  abstract     = {{Fourth-generation synchrotron light sources deliver flux densities up to two orders of magnitude higher than previous generations, enabling X-ray imaging at unprecedented spatiotemporal resolution. This is attractive for non-destructive 4D (3D + time) studies of fast dynamics under in-situ or operando conditions. State-of-the-art 4D X-ray imaging at large-scale facilities relies on tomography and can achieve 1000 volume reconstructions per second at micrometer resolution. However, the requirement of rapid sample rotation induces large centrifugal forces, limiting studies of rotation-sensitive dynamics. To address this limitation, X-ray Multi-Projection Imaging (XMPI) has emerged as an alternative rotation-free approach. By splitting the primary beam into several angularly<br/>separated beams, XMPI enables simultaneous multi-angle acquisitions, enabling ultrafast imaging but making 4D reconstruction and analysis challenging in the sparse-view regime.<br/><br/>In this thesis, we first introduce the basics of X-ray imaging. Second, we present the implementation of the XMPI setup, enabled by the unique capabilities of fourth-generation synchrotron light sources. Third, we present the development of 3D particle tracking velocimetry with XMPI, and demonstrate its capability to quantitatively retrieve flow properties by tracking microscale tracers via triangulation. Lastly, we focus on a more ambitious goal: to develop full-field deep-learning-based 4D reconstruction methods for XMPI, including droplet collisions and flows through porous media, two representative fluid-dynamics examples for validation. This thesis demonstrates that XMPI, together with advanced 4D reconstruction and analysis approaches, becomes an enabling tool for fluid dynamics studies, providing sub-10 micrometer and kHz-to-MHz spatiotemporal resolution.}},
  author       = {{Yao, Zisheng}},
  isbn         = {{978-91-8104-771-4}},
  keywords     = {{X-ray imaging; four-dimensional; deep learning; sparse-view reconstruction; ultrafast dynamics}},
  language     = {{eng}},
  publisher    = {{Lund University}},
  school       = {{Lund University}},
  title        = {{Advancing X-ray Multi-Projection Imaging : Towards Four-Dimensional Reconstruction and Analysis of Rapid Flows}},
  url          = {{https://lup.lub.lu.se/search/files/247446415/Thesis_Zisheng_Yao_LUCRIS.pdf}},
  year         = {{2026}},
}