Table 1.

Metal artifact reduction performances of tomosynthesis reconstruction algorithms. (streak artifact area).

ROI_Location_1
Variable	Difference	Standard Error	p	95% CI *
				Lower Limit	Upper Limit
CGpM-MAR vs. Original-FBP	−0.0664	0.01717	0.002	−0.1118	−0.0211
CGpM-MAR vs. DT-MAR	−0.0190	0.01717	0.686	−0.0644	0.0263
CGpM-MAR vs. CNNMAR	−0.0441	0.01717	0.060	−0.0895	0.0013
Original FBP vs. DT-MAR	0.0474	0.01717	0.037	0.0021	0.0928
Original FBP vs. CNNMAR	0.0224	0.01717	0.565	−0.0230	0.0677
Original FBP vs. CGpM-MAR	0.0664	0.01717	0.002	0.0211	0.1118
DT-MAR vs. Original-FBP	−0.0474	0.01717	0.037	−0.0928	−0.0021
DT-MAR vs. CNNMAR	0.1389	0.0601	0.468	−0.0704	0.0203
DT-MAR vs. CGpM-MAR	0.8692	0.0537	0.686	−0.0263	0.0644
CNNMAR vs. Original-FBP	−0.0224	0.01717	0.565	−0.0677	0.0230
CNNMAR vs. DT-MAR	0.0251	0.01717	0.468	−0.0203	0.0704
CNNMAR vs. CGpM-MAR	0.0441	0.01717	0.060	−0.0013	0.0895
Source of variation	df *	sums of squares	mean square	F	p
Algorithm	3	0.045	0.015	5.709	0.002
Dose	2	0.010	0.005	1.827	0.170
Algorithm × Dose	6	0.017	0.003	1.086	0.381
Error	60	0.159	0.003	-	-

* CI: confidence interval; dependent variable: artificial index value. Tukey–Kramer test; p < 0.05 indicates a significant difference (without metal artifact reduction processing). * df: degree of freedom; dependent variable: artificial index value. Tukey–Kramer test; p < 0.05 indicates a significant difference (without metal artifact reduction processing).