Table 6 Classification results of radiomics method and deep learning
Methods | Models | AUC | ACC (%) | SEN (%) | SPEC (%) | Precision (%) | Recall (%) | F1 (%) |
|---|---|---|---|---|---|---|---|---|
Radiomics [16] | 2D-domain | 0.8151 | 76.00 | 87.50 | 55.56 | 77.78 | 87.50 | 82.35 |
3D-domain | 0.8402 | 74.00 | 78.13 | 66.67 | 80.65 | 78.13 | 79.37 | |
Combined-domain | 0.8107 | 72.00 | 81.25 | 55.56 | 76.47 | 81.25 | 78.79 | |
The proposed method | 2D-ResNet34 | 0.8264 | 76.00 | 78.13 | 72.22 | 83.33 | 78.13 | 80.65 |
3D-ResNet-Anisotropic | 0.8455 | 76.00 | 75.00 | 77.78 | 85.71 | 75.00 | 80.00 | |
Feature-Ensemble | 0.8247 | 76.00 | 87.50 | 55.56 | 77.78 | 87.50 | 82.35 | |
Decision-Ensemble-UA | 0.8837 | 82.00 | 84.38 | 77.78 | 87.10 | 84.38 | 85.71 |