Parallelization of a Fully 3D CT Iterative Reconstruction Algorithm

Abstract

Iterative image reconstruction algorithms for computed tomography are able to incorporate highly accurate physical models for the measured data. While such algorithms provide a high degree of accuracy, their large computational cost currently makes them infeasible for clinical practice. Clinical scanners instead use a linear model that provides fast reconstruction times but limited accuracy in some situations. Using a variety of approaches, the iterative image reconstruction algorithms can be modified to run faster and more accurately. Although the algorithms based on linear models can also be made faster, they cannot provide the same level of accuracy. We describe a parallelized implementation of an alternating minimization algorithm for fully three-dimensional image reconstruction. Various performance results are shown for the reconstruction of simulation data, a phantom scan, and a large clinical scan.

Noise‐resolution tradeoffs in x‐ray CT imaging: A comparison of penalized alternating minimization and filtered backprojection algorithms

“…The literature has shown that the simplified ASIR algorithm has the capability to reconstruct image volumes acceptable to current trained observers in a clinically relevant timeframe of 65 s compared to 50 s for FBP. 39 For the AM algorithm, Keesing et al 40,41 has demonstrated the feasibility of speeding up the computation time by parallelizing the projection operations for a fully 3D helical geometry.…”

A hardware projector/backprojector pair for 3D PET reconstruction

“…There are several implementations of reconstruction systems to speed up its computation : PC clusters 1, 2 , DSP 3 , multi-processor machines [4][5][6] , Cell 7, 8 , GPU 9 or FPGA [10][11][12] . All these implementations have to face the memory bottleneck due to the limited bandwidth of the main memory.…”

Simplified Statistical Image Reconstruction Algorithm for Polyenergetic X-Ray CT

“…The Recent work on algorithm acceleration using parallel computers can be found in [42] and [41]. In [42], the authors reported a speedup of 30 in reconstructing an image of size 512 × 512 × 400 for a C-arm CT scanner.…”

“…In [42], the authors reported a speedup of 30 in reconstructing an image of size 512 × 512 × 400 for a C-arm CT scanner. The authors used two strategies for parallelism : In [41], the authors describe the implementation of the AM-OS algorithm for MSCT scanner data on a parallel-computer with 64 processors.…”

“…7.1(a)) are reconstructed well only with OS algorithms that use a low number of subsets, e.g., 41. While the region of interest (ROI, Fig.…”