| S.P. OLESEN, J. GREGOR AND M.G. THOMASON, Parallel PET reconstruction by EM iteration with reduced processor communications, Technical Report CS-94-256, Univ. of Tennessee, Knoxville, TN, USA, Dec. 1994. |
....parallel programming for dedicated architectures is highly platform dependent. Recently, the idea of distributed processing on a cluster of workstations seems to become more and more popular. Sharing the workload between 8 and 16 CM 5 processors and (in a separate experiment) the same number of SPARC5 Workstations (Olesen, et al., 1994), as well as among 8 SPARCstation 10 40 class CPUs (Fricke, 1996) has been reported to significantly reduce the reconstruction time for dynamic studies. A parallel implementation of the algorithm on a heterogeneous workstation cluster has been recently presented (Zaers, et al., 1997) where the ....
S.P. OLESEN, J. GREGOR AND M.G. THOMASON, Parallel PET reconstruction by EM iteration with reduced processor communications, Technical Report CS-94-256, Univ. of Tennessee, Knoxville, TN, USA, Dec. 1994.
....cluster) thereby making the network a bottleneck. Being able to reduce communication requirements is thus of utmost importance. One solution that we have investigated for the EM algorithm is to let the individual processors compute several local iterations before exchanging their latest results [3]. By delaying the communication, each processor ends up having to use out ofdate data. Convergence can still occur, however, provided that the multiplicative pixel update is constrained not to be too extreme at each iteration. We have also found FOA preprocessing to be extremely helpful [2] ....
S.P. Olesen, J. Gregor, M.G. Thomason, and G.T. Smith, "Parallel PET Reconstruction by EM Iteration with Reduced Processor Communications," Imaging Systems & Technology, 7: 215--223, 1996.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC