Abstract
In MAP EM (OSL) reconstruction with the Gibbs prior, the parameter 5 which appears in the prior is commonly treated as a fixed value. Because the quality of reconstructed images depends on this parameter, we have to select 5 very carefully, and because the statistics of an image vary locally, we should not choose a single δ value for each image. We propose a new decision rule to select an appropriate local δ. In our proposed method, δ is determined as the median of the differences between a value of the pixel of interest and those of neighboring pixels. This selection yields an appropriate prior depending on the regional statistics. The prior therefore preserves the edge property without amplifying statistical noise and it is not necessary to know the appropriate δ value to obtain high quality images. We performed computer simulations to determine the effectiveness of the proposed method. The results showed that the quality of reconstructed images obtained with the proposed method was superior to those obtained with the prior with a fixed δ.
Similar content being viewed by others
References
Dempster AP, Laird NM, Rubin DB. Maximum likelihood from incomplete data via the EM algorithm (with discussion).J Royal Statist Soc B 39: 1–38, 1977.
Lange K, Bahn M, Little R. A theoretical study of some maximum likelihood algorithms for emission and transmission tomography.IEEE Trans Med Imag MI-6: 106–114, 1987.
Geman S, McClure DE. Bayesian image analysis: An application to single photon emission tomography, Proceedings of the Statistical Computing Section, Washington, DC, Amer Statist Assoc, pp. 12–18, 1985.
Hebert TJ, Leahy R. A generalized EM algorithm for 3-D Bayesian reconstruction from Poisson data using Gibbs priors.IEEE Trans Med Imag MI-8: 194–202, 1989.
Green PJ. On the use of the EM algorithm for penalized likelihood estimation.J Royal Statist Soc B 52: 443–452, 1990.
Lange K. Convergence of EM image reconstruction algorithms with Gibbs smoothing.IEEE Trans Med Imag MI-9: 439–446, 1990.
Lalush DS, Tsui BMW. Simulation evaluation of Gibbs prior distributions for use in maximuma posteriori SPECT reconstructions.IEEE Trans Med Imag MI-11: 267–275, 1992.
Lalush DS, Tsui BMW. A generalized Gibbs prior for maximuma posteriori reconstruction in SPECT.Phy Med Biol 38: 729–741, 1992.
Geman S, Geman D. Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images.IEEE Trans Pattern Anal Mach Intell PAMI-6: 721–741, 1984.
Besag J. On the statistical analysis of dirty pictures (with discussion).J Royal Statist Soc B 48: 259–302, 1986.
Shepp LA, Vardi Y. Maximum likelihood reconstruction for emission tomography.IEEE Trans Med Imag MI-1: 113–122, 1982.
Lange K, Carson R. EM reconstruction algorithms for emission and transmission tomography.J Comput Assist Tomog 8: 306–316, 1984.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Ogawa, K., Hiruma, K. A new decision rule for parameter δ in MAP EM (OSL) reconstruction with the Gibbs prior. Ann Nucl Med 10, 299–305 (1996). https://doi.org/10.1007/BF03164736
Received:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF03164736