256
F. Shi et al.
Table 10.2 Mean unsigned positioning errors for all labeled B-scans of PED data between our
segmentation results and reference standards, compared with mean unsigned positioning differences
between manual tracings from two observers and mean unsigned positioning errors of the Iowa
reference algorithm [14] (Mean ± SD, in µm, 3.5 µm 1 pixel)
Surface #
Algo. versus Ref.
Obs. 1 versus Obs. 2
[14] versus Ref.
1
3.91 ± 0.65
5.28 ± 0.79
4.07 ± 0.67
2
7.79 ± 2.47
7.22 ± 1.33
9.82 ± 2.55
4
10.30 ± 2.57
8.09 ± 2.08
12.78 ± 7.36
5
9.14 ± 3.83
9.48 ± 4.59
12.36 ± 8.81
6
9.14 ± 2.87
8.66 ± 1.61
11.42 ± 8.25
7
7.19 ± 3.57
7.18 ± 1.73
8.36 ± 4.26
10
8.99 ± 3.42
7.37 ± 2.39
10.16 ± 5.87
11
6.53 ± 2.28
9.09 ± 1.36
9.49 ± 5.38
Overall
7.87 ± 3.38
7.80 ± 2.54
9.81 ± 6.42
Table 10.3 p-values of the proposed algorithm versus reference standards and the Iowa reference
algorithm [14] for all labeled B-scans of PED data
Surface #
p value Algo. versus Ref.
p value Algo. versus [14]
1
0.001
0.2888
2
0.2951
0.001
4
0.0009
0.0846
5
0.6493
0.0742
6
0.2914
0.1574
7
0.9896
0.2048
10
0.0398
0.3036
11
0.001
0.0030
Overall
0.7357
0.001
Here numbers in bold indicate statistically significantly better performance
Table 10.4 Mean unsigned positioning errors for B-scans with PED between our segmentation
results and reference standards, compared with mean unsigned positioning differences between
manual tracings from two observers and mean unsigned positioning errors of the Iowa reference
algorithm [14] (Mean ± SD, in µm, 3.5 µm 1 pixel)
Surface #
Algo. versus Ref.
Obs. 1 versus Obs. 2
[14] versus Ref
1
4.21 ± 0.89
5.52 ± 1.53
5.17 ± 1.56
2
8.65 ± 3.76
7.07 ± 1.62
10.90 ± 3.36
4
11.48 ± 3.09
8.90 ± 3.02
15.70 ± 6.90
5
11.81 ± 4.30
11.87 ± 6.20
16.08 ± 8.45
6
13.00 ± 4.54
10.25 ± 2.79
16.68 ± 9.55
7
8.70 ± 2.99
10.54 ± 3.66
14.17 ± 5.55
10
9.70 ± 2.95
8.82 ± 3.99
17.99 ± 7.88
11
8.59 ± 6.52
9.41 ± 2.41
18.55 ± 7.60
Overall
9.52 ± 4.61
9.05 ± 3.86
14.41 ± 7.87