Skip to main content
. 2021 May 20;11:618496. doi: 10.3389/fonc.2021.618496

Table 3.

Automatic segmentation results based on different backbones.

Backbone Dice (%) JI (%) Pre (%) Recall (%) ASD (voxel) 95HD (voxel)
ResNet (40) 61.4 ± 3.2 44.3 ± 2.9 63.2 ± 3.0 59.5 ± 4.8 17.8 ± 10.2 33.0 ± 11.3
ResNeXt (20) 62.1 ± 2.8 45.1 ± 3.1 65.6 ± 3.5 61.0 ± 4.3 15.5 ± 9.5 30.6 ± 8.7
SE-ResNeXt (Ours) 62.6 ± 3.4 45.5 ± 3.6 67.1 ± 6.1 61.7 ± 5.4 14.2 ± 8.9 28.2 ± 9.9

Mean ± SD, with best results highlighted in bold.