Table 1.
Quantitative results (mean standard deviation) of the cross-validated LA and PV segmentation, compared to the performance using the WHS, 2D U-Net, 3D U-Net, 2D V-Net and 3D V-Net. AC: Accuracy, SE: Sensitivity, SP: Specificity and DI: Dice score.
| Methods | AC (%) | SE (%) | SP (%) | DI (%) |
|---|---|---|---|---|
| WHS | 99.62 ± 0.21 | 80.86 ± 18.07 | 99.88 ± 0.14 | 84.54 ± 15.11 |
| 2D U-Net | 98.60 ± 0.42 | 93.50 ± 3.73 | 99.11 ± 0.43 | 91.97 ± 2.42 |
| 3D U-Net | 98.48 ± 0.05 | 93.02 ± 3.35 | 99.04 ± 0.44 | 90.58 ± 2.64 |
| 2D V-Net | 98.36 ± 0.58 | 92.20 ± 4.91 | 98.98 ± 0.50 | 90.66 ± 3.15 |
| 3D V-Net | 98.47 ± 0.46 | 94.43 ± 3.33 | 98.89 ± 0.44 | 91.37 ± 2.48 |
| Vesal et al. | 98.30 ± 0.71 | 94.97 ± 3.02 | 98.65 ± 0.73 | 90.58 ± 3.40 |
| SV+CLSTM | 98.49 ± 0.40 | 92.41 ± 4.59 | 99.17 ± 0.45 | 91.67 ± 3.12 |
| MV | 98.04 ± 0.89 | 90.95 ± 4.69 | 98.76 ± 0.59 | 89.03 ± 4.14 |
| S-LA/PV | 98.55 ± 0.51 | 95.32 ± 3.08 | 98.88 ± 0.50 | 91.87 ± 2.68 |
| MVTT | 98.62 ± 0.46 | 92.92 ± 4.47 | 99.20 ± 0.38 | 92.11 ± 2.39 |