Fig. 1.

NCAT Body-Motion Correction of noiseless datasets. (a) Mid-ventricular short-axis slice from simulation which did include motion with no motion correction included in 18 iterations of MLEM. This slice illustrates the significant degradation uncorrected motion causes. (b) Short-axis slice from projections simulated with the phantom solely in the reference (first) motion-state after 18 iterations of MLEM reconstruction. This slice serves as the standard for comparison of the following motion corrections against. (c) Short-axis slice from simulation which did include motion reconstructed with 18 iterations of MC-MLEM motion correction. Notice the similarity to the case reconstructed in the absence of motion. (d) Short-axis slice from simulation which did include motion reconstructed by 9 iterations of MGEM-1. Note the similarity to 18 iterations of MC-MLEM as expected with use of 2 subsets. (e) Short-axis slice from simulation which did include motion reconstructed by 18 iterations of MGEM-1. Note the further recovery of resolution due to the inclusion of modeling spatial-resolution in reconstruction. (f) Short-axis slice from simulation which did include motion reconstructed by 18 iterations of MGEM-2. Notice the considerable smoothing compared to 18 iterations of MGEM-1. (g) Short-axis slice from simulation which did include motion but motion was an integer multiple of the voxel dimension such that interpolation was not required reconstructed by 18 iterations of MGEM-2. Notice that in this case without interpolation error performance of MGEM-2 is close to that of MGEM-1.