Table 5.

Diagnosis results (%) of the DenseANet on the CP and NCP. The numbers in bold black font represent the best classification results. Note that the p-values are calculated between DenseANet and Baseline+ARL.

	Models	Acc	AUC	Se	Sp	Pr	F1
Results of 2D slices	ResNet101	88.65 ± 1.95	95.50 ± 0.61	93.03 ± 0.93	83.43 ± 1.47	86.98 ± 1.98	89.75 ± 1.35
	ResNet101+ARL	90.00 ± 1.01	96.06 ± 0.52	93.03 ± 0.89	86.39 ± 1.93	89.05 ± 1.24	90.91 ± 1.16
	Baseline	90.00 ± 1.13	97.13 ± 0.49	98.01 ± 0.30	80.47 ± 2.64	85.65 ± 2.47	91.24 ± 0.95
	Baseline+ARL	90.54 ± 1.37	96.75 ± 0.56	92.04 ± 1.02	88.76 ± 1.53	90.69 ± 1.17	91.19 ± 1.09
	DenseANet	91.28 ± 0.87	97.24 ± 0.41	98.18 ± 0.24	82.33 ± 1.97	86.85 ± 1.47	92.03 ± 0.69
	p-value	0.014	0.034	1.68e-5	8.61e-7	2.38e-4	0.002
Results of patients	ResNet101	90.88 ± 1.28	97.02 ± 0.49	92.45 ± 0.82	88.15 ± 1.17	93.08 ± 0.88	92.58 ± 0.75
	ResNet101+ARL	92.13 ± 0.89	97.41 ± 0.47	91.97 ± 0.97	92.43 ± 1.02	95.45 ± 0.47	93.81 ± 0.62
	Baseline	93.70 ± 0.77	97.72 ± 0.56	97.78 ± 0.66	86.67 ± 1.34	92.67 ± 0.91	95.07 ± 0.74
	Baseline+ARL	92.73 ± 0.93	97.98 ± 0.38	92.56 ± 1.04	93.04 ± 0.85	95.82 ± 0.58	93.94 ± 0.98
	DenseANet	94.67 ± 0.55	98.41 ± 0.27	98.31 ± 0.27	87.15 ± 1.16	93.47 ± 0.62	95.43 ± 0.38
	p-value	4e-4	0.015	1.33e-10	4.75e-7	0.001	0.002