Table 4. Network performances on the MRI ACDC challenge in Dice score (%).
| Networks | Average | RV | Myo | LV |
|---|---|---|---|---|
| UTNet (41) | 88.30 | 90.41 | 89.15 | 94.39 |
| TransUNet† (35) | 89.71 | 88.86 | 84.54 | 95.73 |
| Swin-UNet† (37) | 90.00 | 88.55 | 85.62 | 95.83 |
| LeVit-Unet (46) | 90.32 | 89.55 | 87.64 | 93.76 |
| Mixed Transformer (40) | 90.43 | 86.64 | 89.04 | 95.62 |
| MISSFormer (39) | 90.86 | 89.55 | 88.04 | 94.99 |
| UNETR (28) | 87.15 | 84.52 | 84.36 | 92.57 |
| VT-UNet† (45) | 91.13 | 89.44 | 88.42 | 95.53 |
| nnFormer† (33) | 91.78 | 90.22 | 89.53 | 95.59 |
| PHTrans (54) | 91.79 | 90.13 | 89.58 | 95.76 |
| MedFormer (51) | 92.14 | 90.95 | 89.71 | 95.76 |
| D-Former (31) | 92.29 | 91.33 | 89.60 | 95.93 |
†, the model was pre-trained on ImageNet. MRI, magnetic resonance imaging; ACDC, Automated Cardiac Diagnosis Challenge; RV, right ventricle; Myo, left ventricular myocardium; LV, left ventricle.