Table 5.
Test Dice Score (%) of medical instrument segmentation.
| Methods | Strong-weak data split(697 in total) | ||
|---|---|---|---|
| 697-0 | 348-349 (50%) | 174-523 (25%) | |
| U-Net | 70.60 ± 0.87 | 65.73 ±0.39 | 62.55 ±0.11 |
| U-Net+Unary sSE | 71.25 ± 0.76 | 67.31 ± 0.88 | 65.93 ± 1.08 |
| MS-Net | – | 67.61 ± 0.56 | 65.98 ±0.75 |
| MS-Net+Unary sSE | – | 68.59 ± 0.35 | 66.15 ± 0.75 |
| MS-Net2 | – | 68.51 ± 0.54 | 66.38 ±0.85 |
| MS-Net2+Unary sSE | – | 68.92 ± 0.47 | 66.55 ± 0.82 |
| MSDN− | 70.86 ±0.72 | 68.71 ± 1.33 | 62.22 ± 0.56 |
| MSDN | – | 69.05 ±0.65 | 66.89 ±0.89 |