Table 3.
Comparison of different segmentation models for vessel segmentation in YOLOv5-generated ROIs
| Method | Dice | Mean | ||||
|---|---|---|---|---|---|---|
| PA | Ao | SVC | IoU | HD | Dice | |
| FCN [15] | 82.90 | 81.64 | 69.03 | 67.40 | 3.54 | 77.86 |
| U-Net [16] | 83.80 | 83.28 | 72.94 | 70.18 | 3.48 | 80.01 |
| U-Net++ [20] | 83.31 | 84.25 | 72.41 | 69.92 | 3.47 | 79.99 |
| Attention U-Net [21] | 82.93 | 81.41 | 66.72 | 67.13 | 3.61 | 77.02 |
| Deeplabv3 [22] | 85.44 | 87.35 | 76.52 | 73.25 | 3.30 | 83.11 |
| T-S-deeplab [32] | 81.36 | 86.89 | 74.24 | 71.38 | 3.46 | 81.35 |
| Ours | 85.55 | 89.12 | 77.54 | 74.51 | 3.25 | 84.07 |
The optimal value is highlighted in bold, while the suboptimal value is underlined (in column)
Comparing with the underlined values, this further illustrates the superiority of our proposed method