An Efficient Parallel Implementation Of The Em Algorithm For Pet Image Reconstruction Utilizing Transputers

1990 IEEE Nuclear Science Symposium Conference Record Pub Date : 1990-10-22 DOI:10.1109/NSSMIC.1990.693599

F. Rosenberger, D. Politte, G. Johns, C. E. Molnar

{"title":"An Efficient Parallel Implementation Of The Em Algorithm For Pet Image Reconstruction Utilizing Transputers","authors":"F. Rosenberger, D. Politte, G. Johns, C. E. Molnar","doi":"10.1109/NSSMIC.1990.693599","DOIUrl":null,"url":null,"abstract":"Absnacr - PET image reconstruction using maximumlikelihood estimates obtained via expectation-maximization (EM) has previously been considered infeasible because of exorbitant computation requirements, despite the greatly improved images available compared to reconstruction with filtered backprojection. An efficient parallel EM algorithm implementation, including intemal corrections, is described for PE\" VI and timing measurements on SPARCstation 330 and DECstation 3100 workstations and T800 Transputers are given. A Transputer implementation of the algorithm using 19 INMOS T800 Transputers and a Sun4260 host has reduced the execution time for a 7-slice study, with 200 iterations per slice, to about 30 minutes. Analysis of a proposed configuration using 73 T800 Transputers indicates that a 7slice study could be reconstructed in about 8 minutes. Transmission-packet size and a ring-plus-linear processor interconnection pattem are chosen to minimize communication time. Processor utilization efficiency for the 73-processor system would be about 0.78.","PeriodicalId":234642,"journal":{"name":"1990 IEEE Nuclear Science Symposium Conference Record","volume":"80 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1990-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"7","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"1990 IEEE Nuclear Science Symposium Conference Record","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/NSSMIC.1990.693599","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 7

Abstract

Absnacr - PET image reconstruction using maximumlikelihood estimates obtained via expectation-maximization (EM) has previously been considered infeasible because of exorbitant computation requirements, despite the greatly improved images available compared to reconstruction with filtered backprojection. An efficient parallel EM algorithm implementation, including intemal corrections, is described for PE" VI and timing measurements on SPARCstation 330 and DECstation 3100 workstations and T800 Transputers are given. A Transputer implementation of the algorithm using 19 INMOS T800 Transputers and a Sun4260 host has reduced the execution time for a 7-slice study, with 200 iterations per slice, to about 30 minutes. Analysis of a proposed configuration using 73 T800 Transputers indicates that a 7slice study could be reconstructed in about 8 minutes. Transmission-packet size and a ring-plus-linear processor interconnection pattem are chosen to minimize communication time. Processor utilization efficiency for the 73-processor system would be about 0.78.

查看原文本刊更多论文

Em算法在Pet图像重建中的高效并行实现

利用期望最大化(EM)获得的最大似然估计进行PET图像重建之前被认为是不可行的，因为计算需求过高，尽管与滤波后的反向投影重建相比，图像质量大大提高。描述了一种有效的并行电磁算法实现，包括内部校正，并给出了SPARCstation 330和DECstation 3100工作站和T800变频器上的定时测量。该算法的Transputer实现使用19台INMOS T800 Transputer和一台Sun4260主机，将每片200次迭代的7片研究的执行时间减少到30分钟左右。使用73台T800 Transputers对提议的配置进行分析表明，7层研究可以在大约8分钟内重建。传输数据包大小和环形+线性处理器互连模式的选择，以尽量减少通信时间。73处理器系统的处理器利用率约为0.78。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

1990 IEEE Nuclear Science Symposium Conference Record

自引率

0.00%

发文量