Table 6 Results of comparison models on Synapse dataset
From: GH-UNet: group-wise hybrid convolution-VIT for robust medical image segmentation
Type | Method | Dice(%) ↑ | HD95 ↓ |
|---|---|---|---|
CNN | UNet | 67.89 | 26.60 |
| Â | UNet++ | 68.50 | 42.39 |
| Â | Att-UNet | 67.40 | 35.73 |
| Â | PSPnet | 67.74 | 30.28 |
| Â | DeepLabv3+ | 66.53 | 29.58 |
| Â | SFA | 67.43 | 26.94 |
| Â | PraNet | 68.79 | 21.49 |
| Â | ACSNet | 68.04 | 24.52 |
| Â | nnUNet | 69.67 | 23.58 |
| Â | Rolling-UNet | 68.21 | 23.67 |
Trans | nnFormer | 69.76 | 20.55 |
| Â | Swin-UNet | 67.74 | 20.28 |
| Â | MISSFormer | 66.50 | 24.36 |
RWKV | Zig-RiR | 74.39 | 13.10 |
Hybrid | BoTNet | 65.98 | 36.92 |
| Â | TransUNet | 68.04 | 27.21 |
| Â | CvT | 67.11 | 21.77 |
| Â | MixFormer | - | - |
| Â | FSCA-Net | 71.74 | 15.01 |
| Â | EMCAD | 72.55 | 14.30 |
| Â | H2Former | 70.52 | 15.03 |
| Â | GH-UNet | 77.68 | 12.46 |
P-values: <5e-2 (Dice) | |||
