Medical Image Segmentation on ISIC 2018
94.2Dice ScorePU-UNet
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PU-UNet2026.06 | 94.2 | 89.3 | — | — | — | — | 88.3 | |
| DilatedSkinNet2026.06 | 94.2 | 89.1 | — | — | — | — | — | |
| SegMoTEinteraction_type=bounding-box, trainable_params=17M, decoder_status=unfrozen2026.02 | 93.02 | — | — | — | — | — | — | |
| MedSAM + AdSTNet2026.06 | 92.7 | 86.8 | — | — | — | — | — | |
| SSFormerVariant=L2022.03 | 92.42 | 86.75 | — | — | — | — | — | |
| SSFormerVariant=S2022.03 | 91.95 | 86.15 | — | — | — | — | — | |
| SSFormerTraining images used=1450/25942024.03 | 91.9 | 86.1 | — | — | — | — | — | |
| IPS-16PNumber of Trainable Parameters=1.3M, Training images used=1450/2594, Prompt points=162024.03 | 91.5 | 84.7 | — | — | — | — | — | |
| U-KABS2026.02 | 91.4 | 84.33 | — | — | 1.48 | — | — | |
| Rich-U-Net2026.03 | 91.16 | 83.97 | — | — | 1.7637 | — | — | |
| EMCAD2026.03 | 90.96 | 83.42 | — | — | 2.1384 | — | — | |
| Rolling-Unet2026.02 | 90.9 | 83.74 | — | — | 1.99 | — | — | |
| C2PModel Type=Universal, Backbone=ConvNeXt-B, Resolution=384x3842026.03 | 90.79 | — | — | — | — | — | — | |
| SpiderModel Type=Universal2026.03 | 90.67 | — | — | — | — | — | — | |
| CTO2026.02 | 90.6 | 84 | — | — | 3.78 | — | — | |
| TransFuse + ConvFormerBackbone=TransFuse, Architecture Modification=ConvFormer2023.09 | 90.56 | — | — | 21.3 | — | — | — | |
| RWKV-UNetModel Type=Specialized2026.03 | 90.55 | — | — | — | — | — | — | |
| LEAFCategory=Proposed2025.07 | 90.5 | 84.1 | — | — | — | — | — | |
| MEGANetModel Type=Specialized2026.03 | 90.45 | — | — | — | — | — | — | |
| SETR + ConvFormerBackbone=SETR, Architecture Modification=ConvFormer2023.09 | 90.41 | — | — | 21.68 | — | — | — | |
| Swin-UMambaModel Type=Specialized2026.03 | 90.4 | — | — | — | — | — | — | |
| FAT-Net + ConvFormerBackbone=FAT-Net, Architecture Modification=ConvFormer2023.09 | 90.36 | — | — | 21.73 | — | — | — | |
| UNeXt2026.03 | 90.3 | 82.61 | — | — | 3.2457 | — | — | |
| Patcher + ConvFormerBackbone=Patcher, Architecture Modification=ConvFormer2023.09 | 90.18 | — | — | 21.88 | — | — | — | |
| Patcher + LayerScaleBackbone=Patcher, Architecture Modification=LayerScale2023.09 | 90.16 | — | — | 21.78 | — | — | — | |
| SCM-UNET2026.02 | 90.14 | 81.88 | — | — | — | — | — | |
| FAT-Net + LayerScaleBackbone=FAT-Net, Architecture Modification=LayerScale2023.09 | 90.06 | — | — | 21.93 | — | — | — | |
| LGMSNetModel Type=Specialized2026.03 | 90.02 | — | — | — | — | — | — | |
| SR-ICLModel Type=Universal2026.03 | 89.99 | — | — | — | — | — | — | |
| AURA-NetModel Type=Specialized2026.03 | 89.92 | — | — | — | — | — | — | |
| Med-URWKV-TType=RWKV, Source=Ours, Params=14.33 M2025.06 | 89.85 | 82.91 | — | — | — | — | — | |
| FAT-Net + Re-attentionBackbone=FAT-Net, Architecture Modification=Re-attention2023.09 | 89.84 | — | — | 22.54 | — | — | — | |
| Med-URWKV†Type=RWKV, Source=Ours, Params=24.41 M2025.06 | 89.8 | 82.95 | — | — | — | — | — | |
| Hybrid TransUNet2026.06 | 89.8 | — | — | — | — | — | — | |
| Patcher + Re-attentionBackbone=Patcher, Architecture Modification=Re-attention2023.09 | 89.73 | — | — | 21.86 | — | — | — | |
| FAT-NetBackbone=FAT-Net, Architecture Modification=None2023.09 | 89.72 | — | — | 22.63 | — | — | — | |
| UNeXt2026.02 | 89.7 | 81.7 | — | — | 1.66 | — | — | |
| Deco-Mamba-V0Architecture=Deco-Mamba-V02026.03 | 89.67 | — | — | — | — | — | — | |
| DoubleU-NetBackbone=VGG-192020.06 | 89.62 | — | — | — | — | — | — | |
| Med-URWKV-SType=RWKV, Source=Ours, Params=53.07 M2025.06 | 89.53 | 82.66 | — | — | — | — | — | |
| MedT2026.02 | 89.49 | 81.48 | — | — | 1.89 | — | — | |
| TransUNet + ConvFormerBackbone=TransUNet, Architecture Modification=ConvFormer2023.09 | 89.4 | — | — | 23.19 | — | — | — | |
| PVT-EMCAD-B2Architecture=PVT-EMCAD-B22026.03 | 89.4 | — | — | — | — | — | — | |
| TransUNetType=ViT, Source=Arxiv’21, Params=92.23 M2025.06 | 89.38 | 82.56 | — | — | — | — | — | |
| Deco-Mamba-V1Architecture=Deco-Mamba-V12026.03 | 89.36 | — | — | — | — | — | — | |
| UTANetModel Type=Specialized2026.03 | 89.32 | — | — | — | — | — | — | |
| Med2DNumber of Trainable Parameters=184.5M, Training images used=5838/7935, Prompt points=162024.03 | 89.3 | 82.3 | — | — | — | — | — | |
| TransFuseBackbone=TransFuse, Architecture Modification=None2023.09 | 89.28 | — | — | 23.08 | — | — | — | |
| Cascaded-MERITArchitecture=Cascaded-MERIT2026.03 | 89.23 | — | — | — | — | — | — | |
| FAT-Net + RefinerBackbone=FAT-Net, Architecture Modification=Refiner2023.09 | 89.2 | — | — | 23.35 | — | — | — | |
| VM-UNetType=Mamba, Source=Arxiv’24, Params=34.62 M2025.06 | 89.2 | 82.24 | — | — | — | — | — | |
| RWKV-UNetType=RWKV, Source=Arxiv’25, Params=17.40 M2025.06 | 89.18 | 82.17 | — | — | — | — | — | |
| PatcherBackbone=Patcher, Architecture Modification=None2023.09 | 89.11 | — | — | 22.16 | — | — | — | |
| WTCM-UNet2026.02 | 89.05 | 80.26 | — | — | — | — | — | |
| MALUNet2026.02 | 89.04 | 80.25 | — | — | 2.73 | — | — | |
| Swin-UNetArchitecture=Swin-UNet2026.03 | 89.04 | — | — | — | — | — | — | |
| SETRBackbone=SETR, Architecture Modification=None2023.09 | 89.03 | — | — | 22.33 | — | — | — | |
| TransFuse + LayerScaleBackbone=TransFuse, Architecture Modification=LayerScale2023.09 | 89 | — | — | 23.96 | — | — | — | |
| MGFuseSegType=CNN, Source=BIBM’23, Params=22.27 M2025.06 | 88.97 | 82 | — | — | — | — | — | |
| Swin-UMambaType=Mamba, Source=MICCAI’24, Params=55.06 M2025.06 | 88.97 | 81.92 | — | — | — | — | — | |
| IMISinteraction_type=bounding-box2026.02 | 88.93 | — | — | — | — | — | — | |
| Trans-U-Net2026.03 | 88.91 | 80.51 | — | — | 4.5603 | — | — | |
| IPS-5PNumber of Trainable Parameters=1.3M, Training images used=1450/2594, Prompt points=52024.03 | 88.9 | 80.8 | — | — | — | — | — | |
| Zig-RiRType=RWKV, Source=TMI’25, Params=24.58 M2025.06 | 88.79 | 82.11 | — | — | — | — | — | |
| TransUNetBackbone=TransUNet, Architecture Modification=None2023.09 | 88.75 | — | — | 25.11 | — | — | — | |
| TransUNet + LayerScaleBackbone=TransUNet, Architecture Modification=LayerScale2023.09 | 88.75 | — | — | 23.32 | — | — | — | |
| H-VmnetType=Mamba, Source=Neurocomputing’25, Params=6.44 M2025.06 | 88.75 | 81.61 | — | — | — | — | — | |
| TransUNetCategory=CNN/Transformer-based2025.07 | 88.7 | 79.7 | — | — | — | — | — | |
| PVT-EMCAD-b1Type=CNN, Source=CVPR’24, Params=15.41 M2025.06 | 88.59 | 81.59 | — | — | — | — | — | |
| Med2DNumber of Trainable Parameters=184.5M, Training images used=5838/7935, Prompt points=52024.03 | 88.4 | 81.3 | — | — | — | — | — | |
| UNeXtType=CNN, Source=MICCAI’22, Params=1.47 M2025.06 | 88.37 | 81.31 | — | — | — | — | — | |
| TransUNet + Re-attentionBackbone=TransUNet, Architecture Modification=Re-attention2023.09 | 88.35 | — | — | 23.15 | — | — | — | |
| SAM-Med2Dinteraction_type=bounding-box2026.02 | 88.32 | — | — | — | — | — | — | |
| VM-UNetArchitecture=VM-UNet2026.03 | 88.32 | — | — | — | — | — | — | |
| MISSFormerType=ViT, Source=TMI’23, Params=35.45 M2025.06 | 88.3 | 81.09 | — | — | — | — | — | |
| TransFuse + Re-attentionBackbone=TransFuse, Architecture Modification=Re-attention2023.09 | 88.28 | — | — | 24.56 | — | — | — | |
| MSRF-Net2022.03 | 88.24 | 83.73 | — | — | — | — | — | |
| TycheModel Type=Universal2026.03 | 88.24 | — | — | — | — | — | — | |
| UL-VM-UNet2026.03 | 88.2 | 78.9 | — | — | 5.4321 | — | — | |
| SDSegCategory=Diffusion-based2025.07 | 88.1 | 79.7 | — | — | — | — | — | |
| MAAU2026.06 | 88.1 | 80.9 | — | — | — | — | — | |
| SETR + Re-attentionBackbone=SETR, Architecture Modification=Re-attention2023.09 | 88 | — | — | 24.57 | — | — | — | |
| TransUNetTraining images used=1450/25942024.03 | 88 | 80.9 | — | — | — | — | — | |
| Multipath Fusion Model2026.06 | 88 | 80 | — | — | — | — | — | |
| PAG-TransYnetArchitecture=PAG-TransYnet2026.03 | 87.99 | — | — | — | — | — | — | |
| SETR + LayerScaleBackbone=SETR, Architecture Modification=LayerScale2023.09 | 87.98 | — | — | 23.94 | — | — | — | |
| Att-UNet2026.02 | 87.91 | 78.43 | — | — | 1.69 | — | — | |
| TransUNet + RefinerBackbone=TransUNet, Architecture Modification=Refiner2023.09 | 87.9 | — | — | 25.31 | — | — | — | |
| U-Net++2026.02 | 87.83 | 78.31 | — | — | 1.56 | — | — | |
| UCTransNetType=ViT, Source=AAAI’22, Params=66.24 M2025.06 | 87.8 | 80.57 | — | — | — | — | — | |
| UNetModel Type=Specialized2026.03 | 87.76 | — | — | — | — | — | — | |
| Deeplabv3+2022.03 | 87.72 | 81.28 | — | — | — | — | — | |
| SemiGDALabeled Ratio=30%2026.04 | 87.62 | 80.51 | — | — | 4.53 | — | — | |
| U-NetCategory=CNN/Transformer-based2025.07 | 87.6 | 77.9 | — | — | — | — | — | |
| U-Net2026.02 | 87.55 | 77.86 | — | — | 1.79 | — | — | |
| nnUNetV2Model Type=Specialized2026.03 | 87.54 | — | — | — | — | — | — | |
| PraNet2026.03 | 87.54 | 78.74 | — | — | 7.1189 | — | — | |
| PraNetTraining images used=1450/25942024.03 | 87.5 | 78.7 | — | — | — | — | — | |
| SAM2interaction_type=bounding-box2026.02 | 87.46 | — | — | — | — | — | — | |
| IPS-3P|16PNumber of Trainable Parameters=1.3M, Training images used=1450/2594, Training prompt points=3, Testing prompt points=162024.03 | 87.4 | 78.5 | — | — | — | — | — |