Table 2 Performance of Transfer Learning Group and General Group in the retrospective datasets.
AUROC (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | F1 | |
|---|---|---|---|---|
Transfer Learning Group | ||||
DenseNet-169 | 0.996 (0.995–0.997) | 0.973 (0.929–0.991) | 0.985 (0.953–0.996) | 0.977 |
DenseNet-121 | 0.994 (0.992–0.995) | 0.947 (0.894–0.975) | 0.995 (0.968–0.999) | 0.969 |
ResNet-101 | 0.972 (0.971–0.976) | 0.913 (0.853–0.951) | 0.970 (0.933–0.988) | 0.935 |
VGG-16 | 0.951 (0.947–0.953) | 0.873 (0.807–0.920) | 0.890 (0.836–0.928) | 0.865 |
General Group | ||||
DenseNet-169 | 0.985 (0.984–0.988) | 0.953 (0.903–0.979) | 0.980 (0.946–0.994) | 0.963 |
DenseNet-121 | 0.992 (0.991–0.993) | 0.947 (0.894–0.975) | 0.990 (0.961–0.998) | 0.966 |
ResNet-101 | 0.976 (0.973–0.979) | 0.913 (0.853–0.951) | 0.945 (0.901–0.971) | 0.919 |
VGG-16 | 0.879 (0.872–0.885) | 0.673 (0.591–0.746) | 0.950 (0.907–0.974) | 0.774 |
