Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

X-ray Imaging at High Brilliance Sources

Buakor, Khachiwan LU (2020) FYSM60 20201
Solid State Physics
Synchrotron Radiation Research
Department of Physics
Abstract
Coherent lensless X-ray imaging techniques provide very high-resolution images not limited by optical components. These techniques record diffraction patterns that can be used for retrieving the phase information of the X-rays passing through a sample. To perform high-resolution imaging with coherent imaging techniques, highly coherent X-rays from high brilliance sources, where high-intensity X-rays at a certain wavelength and direction can be produced, are required. Currently, the highest brilliance sources are diffraction-limited storage rings (DLSRs) and X-ray free-electron lasers (XFELs). Examples of those are MAX IV and the European XFEL (EuXFEL), respectively. The ultrashort pulses provided by XFELs allow the exploration of molecular... (More)
Coherent lensless X-ray imaging techniques provide very high-resolution images not limited by optical components. These techniques record diffraction patterns that can be used for retrieving the phase information of the X-rays passing through a sample. To perform high-resolution imaging with coherent imaging techniques, highly coherent X-rays from high brilliance sources, where high-intensity X-rays at a certain wavelength and direction can be produced, are required. Currently, the highest brilliance sources are diffraction-limited storage rings (DLSRs) and X-ray free-electron lasers (XFELs). Examples of those are MAX IV and the European XFEL (EuXFEL), respectively. The ultrashort pulses provided by XFELs allow the exploration of molecular and chemical dynamics. Therefore, the combination of such pulses with coherent techniques allows the recording of molecular movies at atomic resolution.

The aim of this study is to establish a framework to process imaging data coming from high brilliance X-ray sources, specifically the EuXFEL and MAX IV. The here presented data were collected using novel phase-contrast approaches, which can record images using single pulses of the EuXFEL and MAX IV. The framework of image processing in this works includes flat-field correction (FFC) and phase reconstruction.

FFC is a technique used for decreasing fixed-pattern noise, which is a crucial problem when imaging using single pulses of DLSRs and XFELs. Several methods used for FFC are presented in this study. A computational method based on the principal component analysis (PCA) for FFC is developed and compared with state-of-the-art algorithms. A machine learning approach for FFC is also successfully applied and the accuracy was at the same level as that of the method based on the PCA. The FFC done by machine learning has the potential to be used for online correction of images acquired with ultrafast techniques as the computational time is below milliseconds and several orders of magnitude faster than the other FFC approaches.

The second part of the developed framework is phase reconstruction. In this process, analytic contrast-transfer function (CTF), which is a method used for reconstruction of a weak object, and the iterative method of CTF based on the alternating direction method of multipliers (ADMM-CTF), were implemented on imaging data. The data were recorded at a single distance. The results suggested that the ADMM-CTF algorithm together with the FFC could be a good framework to perform the single-distance reconstruction. The process can be further improved by including new compressed sensing approaches to ADMM-CTF as the current version is limited to piece-wise sparse samples.

This study exploited the unique capabilities of high-brilliance sources and performed the first reconstruction done by ADMM-CTF together with the FFC. The results from the EuXFEL showed that sample movies can be recorded at least 100 times faster than state-of-the-art techniques. The nano-scale resolution images were achieved at the NanoMAX beamline at the MAX IV laboratory. This research could be one step forward towards the development of optimal data processing infrastructures for novel imaging methods based on coherent imaging. (Less)
Please use this url to cite or link to this publication:
author
Buakor, Khachiwan LU
supervisor
organization
course
FYSM60 20201
year
type
H2 - Master's Degree (Two Years)
subject
language
English
id
9023804
date added to LUP
2020-07-02 10:40:29
date last changed
2020-07-02 10:40:53
@misc{9023804,
  abstract     = {{Coherent lensless X-ray imaging techniques provide very high-resolution images not limited by optical components. These techniques record diffraction patterns that can be used for retrieving the phase information of the X-rays passing through a sample. To perform high-resolution imaging with coherent imaging techniques, highly coherent X-rays from high brilliance sources, where high-intensity X-rays at a certain wavelength and direction can be produced, are required. Currently, the highest brilliance sources are diffraction-limited storage rings (DLSRs) and X-ray free-electron lasers (XFELs). Examples of those are MAX IV and the European XFEL (EuXFEL), respectively. The ultrashort pulses provided by XFELs allow the exploration of molecular and chemical dynamics. Therefore, the combination of such pulses with coherent techniques allows the recording of molecular movies at atomic resolution.

The aim of this study is to establish a framework to process imaging data coming from high brilliance X-ray sources, specifically the EuXFEL and MAX IV. The here presented data were collected using novel phase-contrast approaches, which can record images using single pulses of the EuXFEL and MAX IV. The framework of image processing in this works includes flat-field correction (FFC) and phase reconstruction. 

FFC is a technique used for decreasing fixed-pattern noise, which is a crucial problem when imaging using single pulses of DLSRs and XFELs. Several methods used for FFC are presented in this study. A computational method based on the principal component analysis (PCA) for FFC is developed and compared with state-of-the-art algorithms. A machine learning approach for FFC is also successfully applied and the accuracy was at the same level as that of the method based on the PCA. The FFC done by machine learning has the potential to be used for online correction of images acquired with ultrafast techniques as the computational time is below milliseconds and several orders of magnitude faster than the other FFC approaches.

The second part of the developed framework is phase reconstruction. In this process, analytic contrast-transfer function (CTF), which is a method used for reconstruction of a weak object, and the iterative method of CTF based on the alternating direction method of multipliers (ADMM-CTF), were implemented on imaging data. The data were recorded at a single distance. The results suggested that the ADMM-CTF algorithm together with the FFC could be a good framework to perform the single-distance reconstruction. The process can be further improved by including new compressed sensing approaches to ADMM-CTF as the current version is limited to piece-wise sparse samples. 

This study exploited the unique capabilities of high-brilliance sources and performed the first reconstruction done by ADMM-CTF together with the FFC. The results from the EuXFEL showed that sample movies can be recorded at least 100 times faster than state-of-the-art techniques. The nano-scale resolution images were achieved at the NanoMAX beamline at the MAX IV laboratory. This research could be one step forward towards the development of optimal data processing infrastructures for novel imaging methods based on coherent imaging.}},
  author       = {{Buakor, Khachiwan}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{X-ray Imaging at High Brilliance Sources}},
  year         = {{2020}},
}