Cardiac Segmentation on ACDC (test)
93.68Avg DiceAdaptive t-vMF Dice loss
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Adaptive t-vMF Dice lossNetwork=U-Net, lambda=322022.07 | 93.68 | 95.56 | 86.73 | 92.61 | — | — | — | — | — | |
| t-vMF Dice lossNetwork=U-Net, k=1282022.07 | 93.62 | 95.69 | 86.62 | 92.35 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossNetwork=U-Net, lambda=1282022.07 | 93.5 | 95.7 | 86.45 | 92.07 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossNetwork=TransUNet, lambda=1282022.07 | 93.43 | 95.62 | 85.8 | 92.5 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossNetwork=TransUNet, lambda=322022.07 | 93.29 | 95.47 | 85.66 | 92.22 | — | — | — | — | — | |
| CardiacNASMode=Evolutionary, Params=3.58M, GFLOPs=14.56, GPU Days=0.182026.05 | 93.22 | 97.57 | 89.94 | 92.15 | 4.73 | 4.69 | 5.38 | 4.12 | — | |
| Generalized Dice lossNetwork=TransUNet2022.07 | 93.01 | 94.89 | 85.21 | 92.15 | — | — | — | — | — | |
| BFL-log(Dice) lossNetwork=TransUNet2022.07 | 92.94 | 95.08 | 84.91 | 91.96 | — | — | — | — | — | |
| Noise-robust Dice lossNetwork=TransUNet2022.07 | 92.84 | 94.96 | 84.67 | 91.95 | — | — | — | — | — | |
| Focal Dice lossNetwork=U-Net2022.07 | 92.82 | 95.27 | 85.68 | 90.56 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossNetwork=U-Net, lambda=22022.07 | 92.79 | 95.15 | 85.6 | 90.62 | — | — | — | — | — | |
| Adaptive t-vMF Dice lossNetwork=TransUNet, lambda=22022.07 | 92.79 | 94.98 | 84.7 | 91.69 | — | — | — | — | — | |
| t-vMF Dice lossNetwork=U-Net, k=22022.07 | 92.75 | 95.07 | 85.78 | 90.38 | — | — | — | — | — | |
| Focal Tversky lossNetwork=TransUNet2022.07 | 92.74 | 95.02 | 84.48 | 91.71 | — | — | — | — | — | |
| Dice lossNetwork=TransUNet2022.07 | 92.72 | 94.8 | 84.58 | 91.72 | — | — | — | — | — | |
| t-vMF Dice lossNetwork=TransUNet, k=22022.07 | 92.72 | 94.92 | 84.55 | 91.63 | — | — | — | — | — | |
| AgileFormer-BModel Dimension=2D, Deep Supervision=true2024.03 | 92.55 | 96.19 | 90.4 | 91.05 | — | — | — | — | — | |
| BFL-log(Dice) lossNetwork=U-Net2022.07 | 92.53 | 94.93 | 84.94 | 90.47 | — | — | — | — | — | |
| Focal Dice lossNetwork=TransUNet2022.07 | 92.46 | 94.64 | 83.91 | 91.49 | — | — | — | — | — | |
| Focal Tversky lossNetwork=U-Net2022.07 | 92.44 | 95.31 | 84.57 | 90.14 | — | — | — | — | — | |
| Generalized Dice lossNetwork=U-Net2022.07 | 92.42 | 94.84 | 84.7 | 90.37 | — | — | — | — | — | |
| Noise-robust Dice lossNetwork=U-Net2022.07 | 92.4 | 94.88 | 84.64 | 90.29 | — | — | — | — | — | |
| Parallel MERITArchitecture variant=Parallel, Decoder=CASCADE (Additive), Loss Strategy=MUTATION, Number of runs=52023.03 | 92.32 | 96.08 | 90 | 90.87 | — | — | — | — | — | |
| RWKV-UNet2025.01 | 92.29 | 96.83 | 88.78 | 91.26 | — | — | — | — | — | |
| Dice lossNetwork=U-Net2022.07 | 92.21 | 94.76 | 84.37 | 89.92 | — | — | — | — | — | |
| t-vMF Dice lossNetwork=TransUNet, k=322022.07 | 92.21 | 94.76 | 84.37 | 89.92 | — | — | — | — | — | |
| nnAvgdescription=Average of predictions from nnUNet and nnFormer2021.09 | 92.15 | 95.68 | 89.75 | 91.03 | 1.1 | 1.19 | 1.04 | 1.06 | — | |
| PVT-EMCAD-B22025.01 | 92.12 | 96.02 | 89.68 | 90.65 | — | — | — | — | — | |
| AgileFormer-TModel Dimension=3D, Deep Supervision=true2024.03 | 92.07 | 95.81 | 89.8 | 90.59 | — | — | — | — | — | |
| nnFormer2021.09 | 92.06 | 95.65 | 89.58 | 90.94 | 1.12 | 1.23 | 1.04 | 1.09 | — | |
| nnFormer2021.09 | 92.06 | 95.65 | 89.58 | 90.94 | — | — | — | — | — | |
| nnFormer2021.09 | 92.06 | 95.65 | 89.58 | 90.94 | — | — | — | — | — | |
| Attention-MambaParam (M)=14.05, FLOPs (G)=5.04, Training Size=192 × 1922024.02 | 92.03 | 95.99 | 89.91 | 90.19 | 5.52 | — | — | — | 85.68 | |
| nnUNetAnnotations=mask2025.08 | 92 | 94.3 | 90.1 | 91.5 | — | — | — | — | — | |
| RWKV-UNet-T2025.01 | 91.9 | 96.85 | 88.21 | 90.63 | — | — | — | — | — | |
| Cascaded MERITArchitecture variant=Cascaded, Decoder=CASCADE (Additive), Loss Strategy=MUTATION, Number of runs=52023.03 | 91.85 | 95.8 | 89.53 | 90.23 | — | — | — | — | — | |
| DualAttenUNetParam (M)=19.57, FLOPs (G)=16.23, Training Size=192 × 1922024.02 | 91.85 | 95.94 | 89.69 | 89.92 | 8.37 | — | — | — | 85.37 | |
| nnFormerModel Dimension=3D2024.03 | 91.84 | 95.17 | 89.44 | 90.91 | — | — | — | — | — | |
| UNETR++Model Dimension=3D2024.03 | 91.83 | 95 | 90.61 | 89.89 | — | — | — | — | — | |
| MERITModel Dimension=2D2024.03 | 91.81 | 95.85 | 89.12 | 90.44 | — | — | — | — | — | |
| PHTransModel Dimension=3D2024.03 | 91.79 | 95.76 | 89.48 | 90.13 | — | — | — | — | — | |
| AgileFormer-TModel Dimension=2D, Deep Supervision=true2024.03 | 91.76 | 95.77 | 89.71 | 89.8 | — | — | — | — | — | |
| MulitTransParam (M)=39.38, FLOPs (G)=13.43, Training Size=192 × 1922024.02 | 91.76 | 95.99 | 89.57 | 89.71 | 5.57 | — | — | — | 85.3 | |
| UNetPlusPlusParam (M)=9.16, FLOPs (G)=19.62, Training Size=192 × 1922024.02 | 91.74 | 95.91 | 89.78 | 89.52 | 8.85 | — | — | — | 85.27 | |
| Swin UMambaParam (M)=59.89, FLOPs (G)=24.65, Training Size=192 × 1922024.02 | 91.65 | 95.86 | 89.36 | 89.74 | 4.4 | — | — | — | 85.07 | |
| TransCASCADE2025.01 | 91.63 | 95.5 | 90.25 | 89.14 | — | — | — | — | — | |
| TransCASCADEArchitecture=TransCASCADE2023.03 | 91.63 | 95.5 | 90.25 | 89.14 | — | — | — | — | — | |
| Trans-CASCADEModel Dimension=2D2024.03 | 91.63 | 95.5 | 90.25 | 89.14 | — | — | — | — | — | |
| nnUNet2021.09 | 91.61 | 95.36 | 89.24 | 90.24 | 1.15 | 1.31 | 1.06 | 1.09 | — | |
| nnUNetModel Dimension=3D2024.03 | 91.61 | 95.36 | 89.28 | 90.24 | — | — | — | — | — | |
| H2FormerParam (M)=33.68, FLOPs (G)=18.12, Training Size=192 × 1922024.02 | 91.55 | 95.96 | 89.19 | 89.52 | 7.94 | — | — | — | 84.92 | |
| SelfReg-SwinUnet2025.01 | 91.49 | 95.7 | 89.27 | 89.49 | — | — | — | — | — | |
| SelfReg-UNet2025.12 | 91.49 | 95.7 | 89.27 | 89.49 | 1.28 | — | — | — | — | |
| VM-UNet2025.01 | 91.47 | 95.44 | 89.04 | 89.93 | — | — | — | — | — | |
| PVT-CASCADEArchitecture=PVT-CASCADE2023.03 | 91.46 | 95.5 | 89.97 | 88.9 | — | — | — | — | — | |
| PVT-CASCADDEModel Dimension=2D2024.03 | 91.46 | 95.5 | 89.97 | 88 | — | — | — | — | — | |
| AttenUNetParam (M)=34.88, FLOPs (G)=37.45, Training Size=192 × 1922024.02 | 91.36 | 95.89 | 89.69 | 88.5 | 5.45 | — | — | — | 84.73 | |
| VM UNet V2Param (M)=22.77, FLOPs (G)=2.48, Training Size=192 × 1922024.02 | 91.28 | 95.34 | 88.78 | 89.71 | 3.47 | — | — | — | 84.37 | |
| HiFormerParam (M)=37.46, FLOPs (G)=17.51, Training Size=224 × 2242024.02 | 91.24 | 95.7 | 89.04 | 89 | 4.27 | — | — | — | 84.46 | |
| RWKV-UNet-S2025.01 | 91.19 | 96.22 | 87.87 | 89.49 | — | — | — | — | — | |
| U-CycleMLP2025.12 | 91.11 | 95.63 | 88.44 | 89.28 | — | — | — | — | — | |
| SDAUTModel Dimension=2D2024.03 | 91.08 | 95.28 | 88.58 | 89.37 | — | — | — | — | — | |
| WS Dice lossNetwork=TransUNet2022.07 | 91.03 | 94.55 | 81.16 | 88.72 | — | — | — | — | — | |
| MISSFormer2025.01 | 90.86 | 94.99 | 88.04 | 89.55 | — | — | — | — | — | |
| MISSFormerPre-training=Scratch2021.09 | 90.86 | 94.99 | 88.04 | 89.55 | — | — | — | — | — | |
| MISSFormerArchitecture=MISSFormer2023.03 | 90.86 | 94.99 | 88.04 | 89.55 | — | — | — | — | — | |
| MissFormerModel Dimension=2D2024.03 | 90.86 | 94.99 | 88.04 | 89.55 | — | — | — | — | — | |
| UnetParam (M)=31.04, FLOPs (G)=30.75, Training Size=192 × 1922024.02 | 90.82 | 95.6 | 88.95 | 87.9 | 4.7 | — | — | — | 83.84 | |
| M-ENASMode=Evolutionary, Params=5.61M, GFLOPs=69.45, GPU Days=0.622026.05 | 90.59 | 93.54 | 87.42 | 90.81 | 5.9 | 5.89 | 6.72 | 5.09 | — | |
| WS Dice lossNetwork=U-Net2022.07 | 90.57 | 94.51 | 81.03 | 87.08 | — | — | — | — | — | |
| CoTrModel Dimension=2D2024.03 | 90.52 | 95.29 | 88.44 | 87.81 | — | — | — | — | — | |
| PVT-CASCADE2025.01 | 90.45 | 95.19 | 88.96 | 87.2 | — | — | — | — | — | |
| MT-UNet2025.01 | 90.43 | 95.62 | 89.04 | 86.64 | — | — | — | — | — | |
| MT-UNetArchitecture=MT-UNet2023.03 | 90.43 | 95.62 | 89.04 | 86.64 | — | — | — | — | — | |
| MixedUNetModel Dimension=2D2024.03 | 90.43 | 95.62 | 89.04 | 86.64 | — | — | — | — | — | |
| MT-UNet2025.12 | 90.43 | 95.62 | 89.04 | 86.64 | 1.35 | — | — | — | — | |
| VM UNetParam (M)=27.43, FLOPs (G)=2.32, Training Size=192 × 1922024.02 | 90.38 | 95.11 | 88.16 | 87.88 | 5.57 | — | — | — | 83 | |
| LeViT-UNet-384s2025.01 | 90.32 | 93.76 | 87.64 | 89.55 | — | — | — | — | — | |
| LeViT-UNet-384svariant=384s2021.09 | 90.32 | 93.76 | 87.64 | 89.55 | — | — | — | — | — | |
| LeViT-UNet-384sVariant=384s2021.09 | 90.32 | 93.76 | 87.64 | 89.55 | — | — | — | — | — | |
| LeViT-UNet2025.12 | 90.32 | 93.76 | 87.64 | 89.55 | — | — | — | — | — | |
| RefineSegAnnotations=coarse2025.08 | 90.1 | 93.8 | 88.4 | 88.1 | — | — | — | — | — | |
| SwinUNet2025.01 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| SwinUnetPre-training=ImageNet2021.09 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| SwinUNet2021.09 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| SwinUNet2021.09 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| SwinUNetArchitecture=SwinUNet2023.03 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| SwinUNetModel Dimension=2D2024.03 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| Swin-UNet2025.12 | 90 | 95.83 | 85.62 | 88.55 | 1.6 | — | — | — | — | |
| Swin-UNet2025.12 | 90 | 95.83 | 85.62 | 88.55 | — | — | — | — | — | |
| Mamba UNetParam (M)=19.12, FLOPs (G)=2.60, Training Size=192 × 1922024.02 | 89.77 | 95.39 | 87.72 | 86.21 | 6.44 | — | — | — | 82.19 | |
| TransUNet2025.01 | 89.71 | 95.73 | 84.54 | 88.86 | — | — | — | — | — | |
| TranUnetPre-training=ImageNet2021.09 | 89.71 | 95.73 | 84.53 | 88.86 | — | — | — | — | — | |
| TransUNet2021.09 | 89.71 | 95.73 | 84.54 | 88.86 | — | — | — | — | — | |
| TransUNet2021.09 | 89.71 | 95.73 | 84.54 | 88.86 | — | — | — | — | — | |
| TransUNetArchitecture=TransUNet2023.03 | 89.71 | 95.73 | 84.53 | 88.86 | — | — | — | — | — | |
| TransUNetModel Dimension=2D2024.03 | 89.71 | 95.73 | 84.53 | 88.86 | — | — | — | — | — | |
| TransUNet2025.12 | 89.71 | 95.73 | 84.53 | 88.86 | — | — | — | — | — | |
| SwinUNetParam (M)=27.17, FLOPs (G)=5.95, Training Size=224 × 2242024.02 | 89.7 | 94.3 | 87.51 | 87.29 | 11.37 | — | — | — | 81.95 | |
| U-Net2025.12 | 89.68 | 94.68 | 87.21 | 87.17 | 2.61 | — | — | — | — |