Table 2.
Thyroid classification results with federated learning with internal test data.
Deep learning algorithm | Accuracy (%) | Specificity (%) | Sensitivity (%) | PPVa (%) | NPVb (%) | F1 score (%) | AUROC (%) |
VGG19 | 79.5 | 64.3 | 88.2 | 81.2 | 75.7 | 84.5 | 82.0 |
ResNet50 | 77.4 | 57.8 | 88.6 | 78.6 | 74.3 | 83.3 | 78.9 |
ResNext50 | 73.9 | 31.5 | 98.2 | 71.5 | 91.1 | 82.7 | 86.0 |
SE-ResNet50 | 77.9 | 56.3 | 90.2 | 78.3 | 76.8 | 83.8 | 79.9 |
SE-ResNext50 | 77.2 | 42.1 | 97.3 | 74.6 | 90.0 | 84.4 | 87.6 |
aPPV: positive predictive value.
bNPV: negative predictive value.