Table 2.

Test metrics for the five neural networks and ensemble model

Base CNN model	Validation loss	Test loss	Test accuracy	AUC a
VGG16	0.269	0.468	0.86	0.918 (0.864–0.971)
Densenet121	0.21	0.368	0.85	0.929 (0.884–0.974)
NasNetLarge	0.179	0.416	0.88	0.934 (0.887–0.981)
InceptionV3	0.262	0.278	0.88	0.953 (0.918–0.988)
Xception	0.231	0.372	0.85	0.925 (0.872–0.977)
Ensemble			0.92	0.961 (0.929–0.992)

AUC, area under the receiver operator characteristic curve; CNN, convolutional neural network.

^aAUC. The 95% confidence intervals are shown in brackets.