Skip to main content
. 2022 Jul 25;13:100127. doi: 10.1016/j.jpi.2022.100127

Table 3.

Mean MCC and HD scores using U-Net for lymph node segmentation with single versus all scanners training strategy. P-values are based on MCC scores between train and test scanners when the U-Net is applied.

Train scanner Test scanner MCC HD (μm) P-value
Single scanner A A 0.80 ± 0.22 141.95 ± 82.36 2.72E − 25
B 0.76 ± 0.23 85.03 ± 41.89
C 0.81 ± 0.20 150.91 ± 79.45
D 0.38 ± 0.40 248.79 ± 132.49
B A 0.78 ± 0.20 149.63 ± 83.08 3.74E − 29
B 0.81 ± 0.20 87.52 ± 46.23
C 0.79 ± 0.20 159.13 ± 78.71
D 0.22 ± 0.27 284.86 ± 122.38
C A 0.78 ± 0.20 154.20 ± 91.41 1.02E − 44
B 0.79 ± 0.20 90.81 ± 47.93
C 0.81 ± 0.19 156.70 ± 79.84
D 0.18 ± 0.22 292.53 ± 119.52
D A 0.69 ± 0.29 170.25 ± 100.26 2.52E − 16
B 0.84 ± 0.18 74.77 ± 41.86
C 0.87 ± 0.15 132.36 ± 71.71
D 0.83 ± 0.19 149.67 ± 83.95
All scanners A 0.82 ± 0.21 133.74 ± 76.78 1.03E − 19
B 0.85 ± 0.21 80.00 ± 42.21
C 0.82 ± 0.20 146.01 ± 71.24
D 0.76 ± 0.25 162.67 ± 82.14