Table 3.

DTL models performance.

DTL Models	AUC (95%CI)	Accuracy	Sensitivity	Specificity	Precision	Recall	F1-score
Vgg19
Training	0.999 (0.998–1.000)	0.984	0.990	0.975	0.981	0.990	0.985
Validation	0.968 (0.943–0.994)	0.913	0.922	0.902	0.922	0.922	0.922
resnet50
Training	0.988 (0.982–0.994)	0.944	0.971	0.910	0.931	0.971	0.951
Validation	0.888 (0.830–0.946)	0.833	0.923	0.717	0.809	0.923	0.862
GoogLeNet
Training	0.993 (0.985–1.000)	0.975	0.980	0.967	0.974	0.980	0.977
Validation	0.912 (0.866–0.958)	0.819	0.883	0.738	0.810	0.883	0.845
Inception-v3
Training	0.976 (0.964–0.989)	0.935	0.951	0.914	0.933	0.951	0.942
Validation	0.835 (0.769–0.901)	0.761	0.792	0.721	0.782	0.792	0.787

DTL, Deep transfer learning; AUC, area under the curve; 95%CI, 95 % confidential interval.