Table 3.

The best-performing classifiers among various deep models.

Model	Best-performing classifier		AUC	95% CI	Sensitivity	Specificity	PPV	NPV
ViT	MLP	train	0.80	0.74-0.85	0.59	0.86	0.78	0.71
		test	0.78	0.67-0.89	0.50	0.79	0.67	0.66
VGG16	MLP	train	0.85	0.80-0.90	0.66	0.82	0.76	0.74
		test	0.79	0.67-0.90	0.75	0.79	0.75	0.79
ShuffleNet_v2	SVM	train	0.92	0.89-0.95	0.79	0.92	0.89	0.84
		test	0.81	0.70-0.92	0.55	0.92	0.84	0.71
ResNet18	MLP	train	0.87	0.82-0.91	0.77	0.81	0.77	0.80
		test	0.87	0.78-0.96	0.83	0.74	0.73	0.83
MobileNet_v2	MLP	train	0.83	0.78-0.88	0.63	0.85	0.78	0.73
		test	0.74	0.62-0.87	0.62	0.79	0.72	0.71
MnasNet-0.5	LightGBM	train	0.92	0.89-0.96	0.75	0.91	0.88	0.81
		test	0.75	0.63-0.88	0.41	0.88	0.75	0.64
GoogleNet	SVM	train	0.93	0.90-0.96	0.74	0.91	0.88	0.81
		test	0.80	0.68-0.91	0.62	0.77	0.69	0.70
DenseNet121	SVM	train	0.96	0.94-0.98	0.70	0.80	0.75	0.76
		test	0.75	0.63-0.87	0.62	0.77	0.69	0.70
AlexNet	MLP	train	0.87	0.82-0.91	0.73	0.88	0.84	0.80
		test	0.84	0.73-0.94	0.72	0.88	0.84	0.79