Table 5 Results of comparison models on ACDC dataset(%)
From: GH-UNet: group-wise hybrid convolution-VIT for robust medical image segmentation
Type | Method | Average↑ | RV↑ | MYO↑ | LV↑ |
|---|---|---|---|---|---|
CNN | UNet | 89.41 | 87.77 | 85.88 | 94.67 |
| Â | UNet++ | 89.58 | 87.23 | 87.13 | 94.37 |
| Â | Att-UNet | 89.01 | 87.30 | 85.07 | 94.66 |
| Â | PSPNet | 88.75 | 85.99 | 86.39 | 93.87 |
| Â | DeepLabv3+ | 88.25 | 85.41 | 85.44 | 93.90 |
| Â | SFA | 89.24 | 86.42 | 87.83 | 93.49 |
| Â | PraNet | 90.16 | 87.21 | 88.73 | 94.54 |
| Â | ACSNet | 89.39 | 86.94 | 87.04 | 94.21 |
| Â | nnUNet | 91.61 | 90.24 | 89.24 | 95.36 |
| Â | Rolling-UNet | 90.12 | 88.00 | 89.00 | 93.36 |
Trans | nnFormer | 92.06 | 90.94 | 89.58 | 95.65 |
| Â | Swin-UNet | 90.00 | 88.55 | 85.62 | 95.83 |
| Â | MISSFormer | 84.53 | 81.07 | 81.21 | 91.29 |
| Â | UNETR | 88.61 | 85.29 | 86.52 | 94.02 |
RWKV | Zig-RiR | 92.48 | 91.35 | 90.33 | 95.77 |
Hybrid | LeViT-UNet | 90.32 | 89.55 | 87.64 | 93.76 |
| Â | BoTNet | 88.75 | 85.99 | 86.39 | 93.87 |
| Â | TransUNet | 89.71 | 88.86 | 84.54 | 95.73 |
| Â | CvT | 89.01 | 87.30 | 85.07 | 94.66 |
| Â | MixFormer | 91.01 | 89.02 | 88.46 | 95.55 |
| Â | FSCA-Net | 91.44 | 89.32 | 90.15 | 94.87 |
| Â | EMCAD | 92.12 | 90.65 | 89.68 | 96.02 |
| Â | H2Former | 92.40 | 91.31 | 90.12 | 95.76 |
| Â | GH-UNet | 92.61 | 91.45 | 90.51 | 95.86 |
P-values: <5e-2 (Dice) | |||||
