Research Article

ADMM-EM Method for -Norm Regularized Weighted Least Squares PET Reconstruction

Table 3

Comparison of algorithmic performance similar to Table 1. We still use   to represent the best value, and all experiments are still executed twenty times to obtain an average value. Note that ADMM-EM-68-11 uses the same inner iteration count as ADMM-CG.

ADMM- MAE Fun. () LC (%) HC (%) Var. (%) Time Comp.

PL-292-20 23.31 8.31 96.11 64.37 13.25 74.17 5840
CG-71-11 14.73 7.70 97.85 81.21 10.78 16.17 781
EM-68-11 13.217.67 98.42 84.5210.61 9.94 748
EM-256-1 14.45 7.68 98.62 84.33 11.00 6.95256