Table 2 Segmentation results of ablation methods on the test set
From: U-Net combined with multi-scale attention mechanism for liver segmentation in CT images
Method | DSC [%] | IOU [%] | Precision [%] | Recall [%] |
|---|---|---|---|---|
UNet | 89.23 ± 4.98 | 80.93 ± 8.12 | 92.97 ± 3.87 | 86.49 ± 9.59 |
UNet + MSRB | 97.10 ± 1.02 | 94.38 ± 1.91 | 95.76 ± 1.83 | 98.52 ± 1.51 |
UNet + AASPP | 96.83 ± 2.35 | 93.95 ± 4.09 | 96.21 ± 2.81 | 97.59 ± 3.55 |
UNet + RASM | 97.13 ± 1.07 | 94.45 ± 1.98 | 95.60 ± 2.16 | 98.75 ± 1.18 |
MSA-UNet | 98.00 ± 0.38 | 96.08 ± 0.74 | 97.17 ± 0.85 | 98.85 ± 0.70 |