Table 4 Comparisons with the state-of-the-art baselines on the GLAS dataset
From: P-TransUNet: an improved parallel network for medical image segmentation
Method | mDice | mIou | Recall | Precision |
|---|---|---|---|---|
U-Net [5] | 0.7976 | 0.6763 | – | – |
ResUNet [38] | 0.8088 | 0.6911 | 0.8511 | 0.8001 |
MedT [9] | 0.8102 | 0.6961 | – | – |
U-Net++ [6] | 0.8245 | 0.7023 | 0.8324 | 0.8179 |
Attention U-Net [11] | 0.8159 | 0.7006 | – | – |
TransUNet [15] | 0.8634 | 0.7736 | 0.8573 | 0.8268 |
DS-TransUNet [20] | 0.8719 | 0.7845 | – | – |
TransAttUnet [39] | 0.8837 | 0.8008 | 0.8919 | 0.8849 |
FCBFormer [36] | 0.8745 | 0.7903 | 0.8786 | 0.8523 |
Our method | 0.8922 | 0.8124 | 0.8933 | 0.8957 |