. 2021 Dec 13;116:108291. doi: 10.1016/j.asoc.2021.108291

Table 2.

Comparison results of our CIFD-Net method vs. state-of-the-art architectures on the CC-CCII dataset.

Annotation	Method	Patient Acc. (%)	Precision (%)	Sensitivity (%)	Specificity (%)	$F_{1}$ -score (%)	AUC (%)
Patient-level	ResNet-50 [55]	53.70 $_{\pm 0.02}^{* *}$	61.42 $_{\pm 0.08}^{* *}$	77.13 $_{\pm 0.10}^{* *}$	10.37 $_{\pm 0.17}^{* *}$	68.38 $_{\pm 0.05}^{* *}$	46.30 $_{\pm 0.10}^{* *}$
	COVID-Net [3]	53.62 $_{\pm 0.03}^{* *}$	61.35 $_{\pm 0.01}^{* *}$	77.18 $_{\pm 0.05}^{* *}$	10.06 $_{\pm 0.25}^{*}$	68.36 $_{\pm 0.02}^{* *}$	44.53 $_{\pm 0.18}^{*}$
	COVNet [2]	67.64 $_{\pm 0.04}^{* *}$	76.03 $_{\pm 0.08}^{*}$	73.17 $_{\pm 0.07}^{* *}$	57.34 $_{\pm 0.18}^{*}$	74.57 $_{\pm 0.04}^{* *}$	66.13 $_{\pm 0.15}^{*}$
	VB-Net [4]	76.75 $_{\pm 0.04}^{*}$	85.25 $_{\pm 0.10}^{* *}$	77.61 $_{\pm 0.07}^{* *}$	75.22 $_{\pm 0.19}^{*}$	81.25 $_{\pm 0.05}^{* *}$	89.48 $_{\pm 0.16}^{*}$
	CIFD-Net (Ours)	89.25 $_{\pm 0.02}^{* *}$	89.98 $_{\pm 0.13}^{*}$	93.86 $_{\pm 0.06}^{* *}$	80.67 $_{\pm 0.13}^{*}$	91.91 $_{\pm 0.07}^{* *}$	93.22 $_{\pm 0.06}^{* *}$

Image-level	ResNet-50 [55]	67.29 $_{\pm 0.04}^{* *}$	68.23 $_{\pm 0.06}^{* *}$	92.95 $_{\pm 0.05}^{* *}$	20.40 $_{\pm 0.16}^{* *}$	78.71 $_{\pm 0.02}^{* *}$	53.43 $_{\pm 0.11}^{* *}$
	COVID-Net [3]	64.83 $_{\pm 0.08}^{*}$	66.28 $_{\pm 0.07}^{*}$	93.18 $_{\pm 0.02}^{* *}$	12.48 $_{\pm 0.04}^{* *}$	77.46 $_{\pm 0.03}^{* *}$	51.47 $_{\pm 0.09}^{* *}$
	COVNet [2]	70.79 $_{\pm 0.03}^{* *}$	83.09 $_{\pm 0.07}^{*}$	68.95 $_{\pm 0.11}^{* *}$	74.10 $_{\pm 0.08}^{* *}$	75.37 $_{\pm 0.05}^{* *}$	73.08 $_{\pm 0.07}^{* *}$
	CIFD-Net (Ours)	84.83 $_{\pm 0.02}^{* *}$	91.19 $_{\pm 0.03}^{* *}$	84.74 $_{\pm 0.07}^{* *}$	84.99 $_{\pm 0.11}^{* *}$	87.86 $_{\pm 0.04}^{* *}$	89.63 $_{\pm 0.08}^{* *}$

* indicates the $p$ -value $< 0.05$ , and ** represents the $p$ -value $< 0.01$ .