Medical Image Segmentation on GlaS (three fixed-seed random data splits)
88.72IoUALL U-KAN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ALL U-KANArchitecture=fully KA-based2026.02 | 88.72 | 94 | |
| U-KANArchitecture=KAN-incorporated2026.02 | 87.64 | 93.37 | |
| U-Net++2026.02 | 87.07 | 92.96 | |
| U-MambaArchitecture=Mamba-based2026.02 | 87.01 | 93.02 | |
| Att-Unet2026.02 | 86.84 | 92.89 | |
| U-Net2026.02 | 86.66 | 92.79 | |
| Rolling-UNet2026.02 | 86.42 | 92.63 | |
| U-NeXtArchitecture=MLP-based2026.02 | 84.51 | 91.55 |