Table 2.

Comparisons of Different Methods (unit: %). The highest DSC is in bold.

Method	DSC (mean ± SD)	Precision (mean ± SD)	Recall (mean ± SD)
Input 1 + 3D U-Net	66.05±11.52	72.33±15.51	60.78±12.26
Input 2 + 3D U-Net	67.86±9.42	74.11±13.57	62.58±10.01
Our input + 3D U-Net	69.62±9.27	73.88±14.54	65.82±8.53
Input 1 + our network	65.68 ±13.85	66.64±19.70	64.74±9.47
Input 2 + our network	70.00±10.31	68.79±14.13	71.25±8.95
Proposed method	70.73±10.19	72.68±14.51	68.89±10.00
Grow-cut-based method12	57.01 ±17.39	68.99±21.31	53.01±15.97
Component tree-based method13	69.20±10.26	70.29±15.21	66.43±12.05
GoogLeNet14	63.67±12.25	78.25±20.24	53.66 ±9.61
Input 1 + ResUNet15	63.93±13.86	66.24±18.30	61.78±11.47
Our input + ResUNet15	67.12±11.50	68.72±16.32	65.60±7.29
Input 1 + DUNet18	60.64±14.18	71.62±22.06	52.58±13.69
Our input + DUNet18	65.23±14.57	59.19±17.72	72.64±7.59
Input 1 + FCN19	64.47±14.31	63.87±19.72	65.08±8.51
Our input + FCN19	64.49±14.03	63.15±18.62	65.89±8.52
Input 1 + U-Net++20	66.29±12.42	76.53±18.39	58.47±10.90
Our input + U-Net++20	70.67±12.45	68.35±16.70	73.16 ±8.35