Table 1.

Registration accuracy of USENet, GLUE and ReUSENet. The mean TRE is calculated in pixels for all cases (8 landmark pairs per cases) from the in vivo testing dataset.

Case	Initial	USENet	GLUE	ReUSENet
1	6.98	5.02	4.05	3.29
2	6.48	4.08	3.12	2.57
3	10.00	4.21	3.05	1.05
4	8.62	4.15	4.42	2.48
5	3.22	4.82	2.91	3.44
6	7.65	3.09	3.07	2.18
7	3.92	3.50	3.33	4.03
8	9.91	5.05	4.91	2.61
9	9.17	4.62	3.64	3.92
10	7.50	3.39	3.24	2.80
11	8.80	3.89	2.83	1.77
12	9.90	4.88	3.61	3.55
13	8.18	2.58	2.32	2.35
14	6.83	4.64	4.68	3.62
15	8.00	2.62	4.24	1.20
16	10.45	9.16	9.00	7.45
17	4.05	2.59	2.28	2.32
18	5.58	2.15	2.66	1.85
19	5.92	2.49	3.57	2.37
20	18.06	10.40	5.12	2.94
Mean	7.96	4.37	3.80	2.87
Stddev	3.09	2.03	1.44	1.31