Table 3.
Performance metrics | Model A | Model B | A-B | P value | Effect size |
---|---|---|---|---|---|
ResNet50 | |||||
Accuracy | SAM | SGD | 0.006 | 0.003 | 0.761 |
Precision | 0.006 | 0.006 | 0.677 | ||
Recall | 0.004 | 0.046 | 0.440 | ||
F1 score | 0.005 | 0.018 | 0.556 | ||
AUC | 0.008 | < .0001 | 1.052 | ||
ResNet50v2 | |||||
Accuracy | SAM | SGD | 0.007 | < .0001 | 1.456 |
Precision | 0.007 | < .0001 | 0.874 | ||
Recall | 0.009 | < .0001 | 0.995 | ||
F1 score | 0.008 | < .0001 | 1.103 | ||
AUC | 0.006 | 0.001 | 0.899 | ||
ResNet50v2 versus ResNet50 optimizer; SAM | |||||
Accuracy | ResNet50v2 | ResNet50 | 0.004 | 0.004 | 0.835 |
Precision | 0.006 | 0.004 | 0.742 | ||
Recall | 0.005 | 0.013 | 0.560 | ||
F1 score | 0.006 | 0.004 | 0.712 | ||
AUC | 0.007 | < .0001 | 0.932 |
AUC, area under the receiver operating characteristics curve.