Techniques for implementing the EM and simulated-annealing image reconstruction algorithms on a large-grain parallel computer for faster execution per iteration are developed for emission tomography applications. The speedups obtained by implementing the algorithms on up to 54 processors connected in a ring topology are found to be nearly linear. Reconstruction involves finding an estimate of the emission distribution that minimizes an energy function that contains a data-agreement term and a noise-control term. The EM algorithm minimizes the complete-incomplete form of the data-agreement term, which is easily partitioned for parallel computation. The simulated-annealing algorithm is a Monte Carlo method in which any form of data-agreement and noise-control term can be minimized. In the reconstruction of a thyroid phantom, it is demonstrated that the complete-incomplete data-agreement term can be used to facilitate the parallel implementation of simulated annealing while still guaranteeing convergence.
Identifer | oai:union.ndltd.org:arizona.edu/oai:arizona.openrepository.com:10150/277300 |
Date | January 1990 |
Creators | Magee, Kathleen Ann, 1959- |
Contributors | Barrett, Harrision H. |
Publisher | The University of Arizona. |
Source Sets | University of Arizona |
Language | en_US |
Detected Language | English |
Type | text, Thesis-Reproduction (electronic) |
Rights | Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author. |
Page generated in 0.0905 seconds