Table 3 Comparative benchmark of SOTA 3D segmentation networks on CirrMRI600+ → T2W.
From: Large Scale MRI Collection and Segmentation of Cirrhotic Liver
Method | mIoU | Dice | HD95 | Precision | Recall | ASDD |
|---|---|---|---|---|---|---|
VNet18 | 68.98 | 70.01 | 35.67 | 69.98 | 70.56 | 7.18 |
Attention UNet19 | 68.72 | 79.18 | 37.87 | 79.99 | 83.21 | 7.53 |
SynergyVNet3D34 | 75.17 | 77.56 | 28.19 | 83.78 | 85.42 | 5.79 |
TransBTS35 | 62.80 | 74.88 | 43.73 | 76.69 | 79.75 | 8.18 |
UXNet3D36 | 72.11 | 82.16 | 32.01 | 84.83 | 83.68 | 6.00 |
TransUNet3D24 | 77.89 | 79.09 | 34.11 | 78.11 | 79.97 | 6.69 |
LinTransUnet23 | 80.08 | 82.11 | 26.01 | 84.21 | 86.17 | 5.98 |
SwinUNeTr21 | 79.89 | 81.21 | 32.78 | 80.05 | 81.10 | 6.19 |
nnUNet3D20 | 82.11 | 84.76 | 27.73 | 85.78 | 86.66 | 4.76 |
nnFormer3D22 | 83.42 | 86.47 | 25.92 | 87.67 | 88.02 | 4.04 |
nnSynergyNet3D34 | 83.01 | 86.51 | 24.19 | 85.66 | 87.01 | 3.96 |
