Table 3.

Comparison of quantitative classification results based on different models.

Model	Specificity	Sensitivity	Accuracy	AUC	F1
ResNet50	88.74%	77.39%	84.94%	0.91	0.78
Xception	87.16%	77.44%	84.06%	0.90	0.76
InceptionV3	89.06%	77.44%	85.13%	0.91	0.78
CNN3	79.22%	63.19%	74.44%	0.78	0.60
Traditional model	74.61%	58.10%	70.55%	0.73	0.49