Comparison of the proposed segmentation model (M1) with the 3D U-Net and 3D spatially weighted U-Net in terms of similarity, difference and positioning performance
| Methods | DSC (%) | Jaccard (%) | Precision (%) | Sensitivity (%) | HD95 | ASD |
| 3D U-Net | 30.2±4.9 | 26.7±6.3 | 76.2±4.1 | 29.4±7.8 | 197.6±14.7 | 4.3±0.4 |
| 3D spatially weighted U-Net | 67.1±2.4 | 69.8±2.7 | 81.0±3.3 | 83.8±5.3 | 72.7±11.2 | 4.7±0.8 |
| M1 | 87.6±5.0 | 82.2±6.2 | 99.0±1.0 | 83.0±6.6 | 6.9±3.8 | 1.6±1.1 |
Data shown are mean±SD.
ASD, average surface distance; DSC, Dice similarity coefficient; HD95, Hausdorff distance 95%; M1, two-step method (3D spatially weighted U-Net and 2D classical U-Net).