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

Table 1

Algorithmic performance including mean absolute error (MAE), cost function (Fun.), contrast (LC and HC denote that of low and high ROI, resp.), variability (Var.), running time (Time, units of seconds), and computational complexity (Comp.), in which we use [ ] to represent the best value. All experiments are executed twenty times to obtain an average value.

ADMM-	MAE	Fun. ()	LC (%)	HC (%)	Var. (%)	Time	Comp.
PL-50-10	82.40	44.55	71.74	33.85	12.56	6.64	500
CG-50-10	16.00	7.70	97.48	78.06	11.26	10.79	500
EM-50-10	13.73	7.67	98.33	83.44	10.91	6.71	500
EM-50-1	25.69	8.24	98.54	74.93	13.66	1.37	50
PL-100-20	42.88	12.21	89.98	48.17	15.62	25.35	2000
CG-100-20	14.26	7.66	98.01	83.18	10.73	33.56	2000
EM-100-20	13.12	7.66	98.15	84.09	10.69	25.27	2000
EM-100-1	18.98	7.84	98.61	81.17	12.27	2.73	100
PL-200-40	20.47	7.99	96.74	68.93	12.65	99.47	8000
CG-200-40	14.22	7.65	98.24	83.82	10.73	115.17	8000
EM-200-40	13.03	7.64	98.32	84.35	10.68	97.62	8000
EM-200-1	15.30	7.69	98.63	83.77	11.34	5.42	200
PL-300-80	14.43	7.68	97.98	81.03	11.13	295.93	24000
CG-300-80	14.20	7.65	98.24	84.30	10.69	318.08	24000
EM-300-80	12.96	7.64	98.31	84.78	10.64	288.76	24000
EM-300-1	14.23	7.68	98.54	84.66	11.02	8.14	300
PL-400-120	13.23	7.65	98.26	83.85	10.87	590.77	48000
CG-400-120	14.23	7.64	98.19	83.84	10.82	617.80	48000
EM-400-120	13.05	7.64	98.26	84.28	10.78	615.27	48000
EM-400-1	13.15	7.66	98.42	84.30	10.88	10.84	400