Accelerated MAP reconstructions using an accelerated factor

Yu Jung Tsai*, Ing Tsung Hsiao

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

In emission tomography, statistical reconstruction has been proved outperforming FBP (filtered backprojection) [1]. ML-EM (maximum likelihood expectation-maximization) [2] algorithm is one of the most widely used statistical reconstruction methods. Although with good image quality, the ML-EM algorithm is challenged by its slow convergence speed and ill-condition problem. To solve for the problems, based on the ideas from the accelerated EM (AEM) [3] and maximum a posteriori EM (MAP-EM) [4] algorithms, we derived two approaches (MAP-AEM1 and MAP-AEM2) in this study to accelerate MAP reconstruction using a bigger step size instead of ordered subsets. The results were compared with ordered-subset methods of both BSREM and MAPCOSEM. From the preliminary results, our methods showed the speedup ability for the traditional MAP-EM.

Original languageEnglish
Title of host publicationIEEE Nuclear Science Symposuim and Medical Imaging Conference, NSS/MIC 2010
Pages3278-3281
Number of pages4
DOIs
StatePublished - 2010
Externally publishedYes
Event2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010 - Knoxville, TN, United States
Duration: 30 10 201006 11 2010

Publication series

NameIEEE Nuclear Science Symposium Conference Record
ISSN (Print)1095-7863

Conference

Conference2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010
Country/TerritoryUnited States
CityKnoxville, TN
Period30/10/1006/11/10

Fingerprint

Dive into the research topics of 'Accelerated MAP reconstructions using an accelerated factor'. Together they form a unique fingerprint.

Cite this