Table 4.
Results of average AUC, sensitivity, and specificity in the 10-fold cross-validation
| Model | FTC vs. (FV-PTC + BN) |
FV-PTC vs. (FTC + BN) |
(FTC + FV-PTC) vs. BN |
||||||
|---|---|---|---|---|---|---|---|---|---|
| AUC (SD) | SEN. (SD) | SPE. (SD) | AUC (SD) | SEN. (SD) | SPE. (SD) | AUC (SD) | SEN. (SD) | SPE. (SD) | |
| ResNet50 | 0.790 (0.015) | 0.684 (0.032) | 0.830 (0.013) | 0.783 (0.017) | 0.683 (0.031) | 0.819 (0.010) | 0.803 (0.014) | 0.865 (0.012) | 0.658 (0.022) |
| RadImageNet | 0.830 (0.012) | 0.713 (0.032) | 0.857 (0.014) | 0.816 (0.013) | 0.662 (0.035) | 0.834 (0.014) | 0.836 (0.010) | 0.854 (0.016) | 0.693 (0.025) |
| ThyNet | 0.838 (0.011) | 0.716 (0.030) | 0.859 (0.014) | 0.822 (0.010) | 0.735 (0.024) | 0.832 (0.012) | 0.840 (0.009) | 0.887 (0.014) | 0.699 (0.017) |
| Swin-transformer | 0.933 (0.015) | 0.860 (0.054) | 0.935 (0.009) | 0.920 (0.014) | 0.831 (0.021) | 0.938 (0.016) | 0.935 (0.010) | 0.927 (0.016) | 0.896 (0.021) |