Volume 27, Issue 5
Uncertainty Quantification of Density Reconstruction Using MCMC Method in High-Energy X-ray Radiography

Xinge Li, Haibo Xu, Na Zheng, Qinggang Jia, Tongxiang Gu & Suhua Wei

Commun. Comput. Phys., 27 (2020), pp. 1485-1504.

Published online: 2020-03

Preview Full PDF 83 753
Export citation
  • Abstract

High-energy X-ray radiography is a measuring technique for quantitativemeasurement and diagnosis of the object and its internal structure. Tomographic re-construction determines the geometric and physical properties of the object according to the energy distribution on the imaging plane. Considering the noise and blur in the process of radiographing, we construct a general reconstruction model for the axisym-metric single image photographic system. This inverse problem is then cast within a statistical framework in order to compute volumetric object densities from X-ray ra-diographs and to quantify uncertainties in the reconstruction. A hierarchical Bayesian model is developed with a likelihood based on a Gaussian noise model and with pri-ors placed on the unknown nonnegative density profile, the precision matrix, and two scale parameters. This results in a joint posterior distribution, which can be readily sampled using the Markov chain Monte Carlo (MCMC) method. To study the role of hyperparameters and their sensitivity analysis, a wide variety of tests were conducted which led to a number of definitive conclusions. Results of the density reconstructions and pointwise uncertainty estimates are presented for simulated signals with various physical factors in the imaging process included.

  • Keywords

65K99, 65Z05, 62F15, 62P35

  • AMS Subject Headings

Inverse problem, density reconstruction, uncertainty quantification, Bayesian infer- ence, MCMC method.

  • Copyright

COPYRIGHT: © Global Science Press

  • Email address

lixinge@lsec.cc.ac.cn (Xinge Li)

xu haibo@iapcm.ac.cn ( Haibo Xu)

zheng na@iapcm.ac.cn ( Na Zheng)

QGJIA XJTU@126.com (Qinggang Jia)

txgu@iapcm.ac.cn ( Tongxiang Gu)

wei suhua@iapcm.ac.cn (Suhua Wei)

  • BibTex
  • RIS
  • TXT
@Article{CiCP-27-1485, author = {Li , Xinge and Xu , Haibo and Zheng , Na and Jia , Qinggang and Gu , Tongxiang and Wei , Suhua }, title = {Uncertainty Quantification of Density Reconstruction Using MCMC Method in High-Energy X-ray Radiography}, journal = {Communications in Computational Physics}, year = {2020}, volume = {27}, number = {5}, pages = {1485--1504}, abstract = {

High-energy X-ray radiography is a measuring technique for quantitativemeasurement and diagnosis of the object and its internal structure. Tomographic re-construction determines the geometric and physical properties of the object according to the energy distribution on the imaging plane. Considering the noise and blur in the process of radiographing, we construct a general reconstruction model for the axisym-metric single image photographic system. This inverse problem is then cast within a statistical framework in order to compute volumetric object densities from X-ray ra-diographs and to quantify uncertainties in the reconstruction. A hierarchical Bayesian model is developed with a likelihood based on a Gaussian noise model and with pri-ors placed on the unknown nonnegative density profile, the precision matrix, and two scale parameters. This results in a joint posterior distribution, which can be readily sampled using the Markov chain Monte Carlo (MCMC) method. To study the role of hyperparameters and their sensitivity analysis, a wide variety of tests were conducted which led to a number of definitive conclusions. Results of the density reconstructions and pointwise uncertainty estimates are presented for simulated signals with various physical factors in the imaging process included.

}, issn = {1991-7120}, doi = {https://doi.org/10.4208/cicp.OA-2019-0060}, url = {http://global-sci.org/intro/article_detail/cicp/15766.html} }
TY - JOUR T1 - Uncertainty Quantification of Density Reconstruction Using MCMC Method in High-Energy X-ray Radiography AU - Li , Xinge AU - Xu , Haibo AU - Zheng , Na AU - Jia , Qinggang AU - Gu , Tongxiang AU - Wei , Suhua JO - Communications in Computational Physics VL - 5 SP - 1485 EP - 1504 PY - 2020 DA - 2020/03 SN - 27 DO - http://dor.org/10.4208/cicp.OA-2019-0060 UR - https://global-sci.org/intro/cicp/15766.html KW - 65K99, 65Z05, 62F15, 62P35 AB -

High-energy X-ray radiography is a measuring technique for quantitativemeasurement and diagnosis of the object and its internal structure. Tomographic re-construction determines the geometric and physical properties of the object according to the energy distribution on the imaging plane. Considering the noise and blur in the process of radiographing, we construct a general reconstruction model for the axisym-metric single image photographic system. This inverse problem is then cast within a statistical framework in order to compute volumetric object densities from X-ray ra-diographs and to quantify uncertainties in the reconstruction. A hierarchical Bayesian model is developed with a likelihood based on a Gaussian noise model and with pri-ors placed on the unknown nonnegative density profile, the precision matrix, and two scale parameters. This results in a joint posterior distribution, which can be readily sampled using the Markov chain Monte Carlo (MCMC) method. To study the role of hyperparameters and their sensitivity analysis, a wide variety of tests were conducted which led to a number of definitive conclusions. Results of the density reconstructions and pointwise uncertainty estimates are presented for simulated signals with various physical factors in the imaging process included.

Xinge Li, Haibo Xu, Na Zheng, Qinggang Jia, Tongxiang Gu & Suhua Wei. (2020). Uncertainty Quantification of Density Reconstruction Using MCMC Method in High-Energy X-ray Radiography. Communications in Computational Physics. 27 (5). 1485-1504. doi:10.4208/cicp.OA-2019-0060
Copy to clipboard
The citation has been copied to your clipboard