Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Tomographic reconstruction with a generative adversarial network

Yang, Xiaogang ; Kahnt, Maik LU orcid ; Bruckner, Dennis ; Schropp, Andreas ; Fam, Yakub ; Becher, Johannes ; Grunwaldt, Jan Dierk ; Sheppard, Thomas L. and Schroer, Christian G. (2020) In Journal of Synchrotron Radiation 27. p.486-493
Abstract

This paper presents a deep learning algorithm for tomographic reconstruction (GANrec). The algorithm uses a generative adversarial network (GAN) to solve the inverse of the Radon transform directly. It works for independent sinograms without additional training steps. The GAN has been developed to fit the input sinogram with the model sinogram generated from the predicted reconstruction. Good quality reconstructions can be obtained during the minimization of the fitting errors. The reconstruction is a self-Training procedure based on the physics model, instead of on training data. The algorithm showed significant improvements in the reconstruction accuracy, especially for missing-wedge tomography acquired at less than 180° rotational... (More)

This paper presents a deep learning algorithm for tomographic reconstruction (GANrec). The algorithm uses a generative adversarial network (GAN) to solve the inverse of the Radon transform directly. It works for independent sinograms without additional training steps. The GAN has been developed to fit the input sinogram with the model sinogram generated from the predicted reconstruction. Good quality reconstructions can be obtained during the minimization of the fitting errors. The reconstruction is a self-Training procedure based on the physics model, instead of on training data. The algorithm showed significant improvements in the reconstruction accuracy, especially for missing-wedge tomography acquired at less than 180° rotational range. It was also validated by reconstructing a missing-wedge X-ray ptychographic tomography (PXCT) data set of a macroporous zeolite particle, for which only 51 projections over 70° could be collected. The GANrec recovered the 3D pore structure with reasonable quality for further analysis. This reconstruction concept can work universally for most of the ill-posed inverse problems if the forward model is well defined, such as phase retrieval of in-line phase-contrast imaging.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
generative adversarial network (GAN), missing-wedge tomography, ptychography, reconstruction algorithms
in
Journal of Synchrotron Radiation
volume
27
pages
8 pages
publisher
International Union of Crystallography
external identifiers
  • pmid:32153289
  • scopus:85081676719
ISSN
0909-0495
DOI
10.1107/S1600577520000831
language
English
LU publication?
yes
id
ec9ece37-2856-4761-a8d6-ff20c4183488
date added to LUP
2020-03-30 13:41:05
date last changed
2024-04-17 06:34:39
@article{ec9ece37-2856-4761-a8d6-ff20c4183488,
  abstract     = {{<p>This paper presents a deep learning algorithm for tomographic reconstruction (GANrec). The algorithm uses a generative adversarial network (GAN) to solve the inverse of the Radon transform directly. It works for independent sinograms without additional training steps. The GAN has been developed to fit the input sinogram with the model sinogram generated from the predicted reconstruction. Good quality reconstructions can be obtained during the minimization of the fitting errors. The reconstruction is a self-Training procedure based on the physics model, instead of on training data. The algorithm showed significant improvements in the reconstruction accuracy, especially for missing-wedge tomography acquired at less than 180° rotational range. It was also validated by reconstructing a missing-wedge X-ray ptychographic tomography (PXCT) data set of a macroporous zeolite particle, for which only 51 projections over 70° could be collected. The GANrec recovered the 3D pore structure with reasonable quality for further analysis. This reconstruction concept can work universally for most of the ill-posed inverse problems if the forward model is well defined, such as phase retrieval of in-line phase-contrast imaging.</p>}},
  author       = {{Yang, Xiaogang and Kahnt, Maik and Bruckner, Dennis and Schropp, Andreas and Fam, Yakub and Becher, Johannes and Grunwaldt, Jan Dierk and Sheppard, Thomas L. and Schroer, Christian G.}},
  issn         = {{0909-0495}},
  keywords     = {{generative adversarial network (GAN); missing-wedge tomography; ptychography; reconstruction algorithms}},
  language     = {{eng}},
  month        = {{03}},
  pages        = {{486--493}},
  publisher    = {{International Union of Crystallography}},
  series       = {{Journal of Synchrotron Radiation}},
  title        = {{Tomographic reconstruction with a generative adversarial network}},
  url          = {{http://dx.doi.org/10.1107/S1600577520000831}},
  doi          = {{10.1107/S1600577520000831}},
  volume       = {{27}},
  year         = {{2020}},
}