Evaluation of a Fully 3-D BPF Method for Small Animal Pet Images on Mimd Architectures

Abstract

Positron Emission Tomography (PET) images can be reconstructed using Fourier transform methods. This paper describes the performance of a fully 3-D Backprojection-Then-Filter (BPF) algorithm on the Cray T3E machine and on a cluster of workstations. PET reconstruction of small animals is a class of problems characterized by poor counting statistics. The low-count nature of these studies necessitates 3-D reconstruction in order to improve the sensitivity of the PET system: by including axially oblique Lines Of Response (LORs), the sensitivity of the system can be significantly improved by the 3-D acquisition and reconstruction. The BPF method is widely used in clinical studies because of its speed and easy implementation. Moreover, the BPF method is suitable for on-time 3-D reconstruction as it does not need any sinogram or rearranged data. In order to investigate the possibility of on-line processing, we reconstruct a phantom using the data stored in the list-mode format by the data acquisition system. We show how the intrinsically parallel nature of the BPF method makes it suitable for on-line reconstruction on a MIMD system such as the Cray T3E. Lastly, we analyze the performance of this algorithm on a cluster of workstations.

GPU Based Modified HYPR Technique

“…The third approach is for parallelization of the computation (Guerrini et al, 1989;Chen et al, 1990;McCarty et al, 1991;Atkins et al, 1991). The most time-consuming module in the reconstruction process is backprojection in FBP, and both forward projection and backprojec-tion in IR (Bevilacqua et al, 1999;Backfrieder et al, 2001;Shattuck et al, 2002). A variety of parallelization schemes have been proposed to allocate the workload of backprojection and forward-projection among parallel computing units.…”

Review of Parallel Computing Techniques for Computed Tomography Image Reconstruction

“…They are built based on the multiple instructions multiple data (MIMD) architecture. As the computing technology advances dramatically most recent parallel implementations become this type [48][49][50][51][52]. The PC cluster comprises several workstation nodes.…”