Table 3.

Mean ± standard deviation of quantitative results for various segmentation methods. The best performance is indicated in bold font.

LIDC Set	DSC (%)	ASD (mm)	SEN (%)	PPV (%)
Level Set	60.63± 17.39	0.48± 0.25	64.38± 22.75	71.03± 24.35
Graph Cut	68.90± 16.03	0.48± 0.30	80.81± 15.25	65.09± 22.42
U-Net	79.50± 13.95	0.24± 0.23	86.81± 18.43	78.18± 16.13
3-D-Patch Branch	79.20± 11.88	0.21± 0.17	90.93± 14.72	72.91± 13.73
2-D-Patch Branch	80.47± 11.23	0.18± 0.15	91.36± 14.40	74.64± 13.16
CF-CNN-MP	80.39± 11.90	0.18± 0.15	91.33± 14.88	74.52± 13.54
CF-CNN	82.15± 10.76	0.17± 0.23	92.75± 12.83	75.84± 13.14
GDGH Set	DSC (%)	ASD (mm)	SEN (%)	PPV (%)
Level Set	66.02± 17.21	0.78± 0.65	60.83± 17.98	79.24± 21.38
Graph Cut	74.13± 13.32	0.83± 0.56	82.94± 13.66	69.24± 16.60
U-Net	75.26± 11.82	0.49 ± 0.48	76.65± 16.42	77.21± 11.57
3-D-Patch Branch	77.89± 10.64	0.40± 0.31	81.29± 15.60	76.95± 11.62
2-D-Patch Branch	78.98± 11.96	0.38± 0.39	81.42± 16.90	79.65± 12.20
CF-CNN-MP	78.61± 12.18	0.39± 0.38	80.93± 17.07	79.38± 12.03
CF-CNN	80.02± 11.09	0.35± 0.34	83.19± 15.22	79.30± 12.09

Note: 3-D-Patch Branch and 2-D-Patch Branch represent the 3-D and 2-D branches in CF-CNN model. CF-CNN-MP represents the CF-CNN model using traditional max pooling instead of central pooling.