Table 2 Performance of models in the internal validation dataset
Model number | Parameters | Architecture | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | AUC (95% CI) |
|---|---|---|---|---|---|---|
1 | Class weight = [0.1, 25] Epoch = 80 | ResNet101 | 0.961 (0.930–0.991) | 0.793 (0.607–0.978) | 0.968 (0.946–0.991) | 0.969 (0.942–0.995) |
2 | ResNet50 | 0.962 (0.950–0.974) | 0.779 (0.674–0.884) | 0.973 (0.954–0.991) | 0.973 (0.957–0.990) | |
3 | InceptionResNetV2 | 0.960 (0.942–0.978) | 0.769 (0.494-1.000) | 0.970 (0.941-1.000) | 0.946 (0.879-1.000) | |
4 | InceptionV3 | 0.954 (0.933–0.976) | 0.756 (0.664–0.848) | 0.966 (0.948–0.985) | 0.958 (0.931–0.986) | |
5 | Class weight = [0.1, 30] Epoch = 60 | ResNet101 | 0.956 (0.923–0.989) | 0.769 (0.692–0.846) | 0.967 (0.939–0.994) | 0.958 (0.952–0.965) |
6 | ResNet50 | 0.963 (0.925-1.000) | 0.883 (0.777–0.988) | 0.965 (0.912-1.000) | 0.972 (0.952–0.992) | |
7 | InceptionResNetV2 | 0.967 (0.942–0.992) | 0.809 (0.686–0.933) | 0.967 (0.919-1.000) | 0.963 (0.941–0.986) | |
8 | InceptionV3 | 0.939 (0.865-1.000) | 0.833 (0.581-1.000) | 0.945 (0.853-1.000) | 0.944 (0.900-0.988) |