Table 4.
A comparison of prediction accuracy on cine MR dataset from different methods. All accuracy measures are represented by mean ± standard deviation, which are computed over different patients and time frames.
| Epi | LA | LV | RA | RV | Ao | PA | WH | ||
|---|---|---|---|---|---|---|---|---|---|
|
| |||||||||
| Dice (↑) | Ours | 0.656 ± 0.169 | 0.708 ± 0.187 | 0.822 ± 0.104 | 0.672 ± 0.114 | 0.643 ± 0.228 | 0.543 ± 0.255 | 0.445 ± 0.225 | 0.693 ± 0.112 |
| 2D UNet | 0.543 ± 0.263 | 0.517 ± 0.283 | 0.734 ± 0.218 | 0.274 ± 0.218 | 0.644 ± 0.184 | 0.393 ± 0.215 | 0.487 ± 0.286 | 0.598 ± 0.166 | |
| 3D UNet | 0.546 ± 0.244 | 0.702 ± 0.22 | 0.782 ± 0.134 | 0.598 ± 0.169 | 0.631 ± 0.144 | 0.495 ± 0.175 | 0.285 ± 0.249 | 0.627 ± 0.131 | |
| Voxel2Mesh | 0.438 ± 0.178 | 0.529 ± 0.275 | 0.669 ± 0.135 | 0.54 ± 0.206 | 0.598 ± 0.273 | 0.395 ± 0.246 | 0.223 ± 0.195 | 0.527 ± 0.167 | |
| ASSD (mm) (↓) | Ours | 4.009 ± 1.118 | 4.775 ± 2.522 | 4.534 ± 2.195 | 5.299 ± 1.883 | 5.468 ± 1.856 | 6.713 ± 3.233 | 7.463 ± 3.14 | 5.466 ± 1.613 |
| 2D UNet | 4.585 ± 3.501 | 6.665 ± 5.147 | 5.204 ± 3.3 | 10.638 ± 6.918 | 4.12 ± 2.493 | 8.36 ± 7.738 | 7.914 ± 9.257 | 6.784 ± 3.951 | |
| 3D UNet | 3.498 ± 2.47 | 4.841 ± 5.061 | 3.228 ± 2.945 | 8.537 ± 5.393 | 5.234 ± 2.466 | 10.022 ± 6.599 | 11.643 ± 8.608 | 6.715 ± 3.091 | |
| Voxel2Mesh | 5.104 ± 1.767 | 7.105 ± 3.082 | 6.763 ± 2.528 | 6.945 ± 3.163 | 7.775 ± 4.613 | 9.181 ± 4.593 | 12.079 ± 7.703 | 7.85 ± 2.881 | |