Table 4 . Performance metrics of SVM classifier trained individually with each deep feature extracted from the pre-trained CNNs.

	Accuracy (std)	AUC (std)	Sensitivity (std)	Specificity (std)	Precision (std)	F1 score (std)	Time (s) (std)
AlexNet	90.9% (0.002)	0.95 (0)	0.922 (0.005)	0.896 (0.003)	0.891 (0.005)	0.907 (0.003)	20.991 (3.066)
GoogleNet	89.2% (0.004)	0.95 (0)	0.914 (0.029)	0.86 (0.009)	0.849 (0.006)	0.881 (0.016)	3.867 (0.274)
ResNet-18	92.5% (0.005)	0.97 (0)	0.933 (0.005)	0.918 (0.07)	0.916 (0.007)	0.925 (0.006)	1.947 (0.25)
ShuffleNet	91.1% (0.002)	0.98 (0.001)	0.919 (0.003)	0.904 (0.004)	0.902 (0.005)	0.911 (0.003)	2.54 (0.168)

Note:

Bold values indicate the highest results.