Open AccessTomographic reconstruction with a generative adversarial network
Author(s) -
Yang Xiaogang,
Kahnt Maik,
Brückner Dennis,
Schropp Andreas,
Fam Yakub,
Becher Johannes,
Grunwaldt Jan-Dierk,
Sheppard Thomas L.,
Schroer Christian G.
Publication year - 2020
Publication title -
journal of synchrotron radiation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.172
H-Index - 99
ISSN - 1600-5775
DOI - 10.1107/s1600577520000831
Subject(s) - iterative reconstruction , computer science , tomographic reconstruction , missing data , inverse problem , wedge (geometry) , algorithm , tomography , radon transform , artificial intelligence , inverse , mathematics , physics , optics , machine learning , mathematical analysis , geometry
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.