TABLE 3.
Quantitative results of different feature fusion methods.
| Test | HD95 (mm) | IoU | clDice | Dice |
|---|---|---|---|---|
| Baseline | 10.320 | 0.629 | 0.92 | 0.769 |
| Baseline+C | 9.550 | 0.642 | 0.894 | 0.777 |
| Baseline+DRC | 6.355 | 0.656 | 0.912 | 0.788 |
Among them, C means to use the traditional convolution module, and DRC means to use the deep residual convolution module. Bold values are reports the optimal result.