A Novel COVID-19 Diagnosis Support System Using the Stacking Approach and Transfer Learning Technique on Chest X-Ray Images

Table 12

Results for the staking technique-based model compared with some previous works.

Works	Methods used	Accuracy	Precision	Sensitivity	Specificity
[35]	ResNet50	93.01%	95.18%	91.45%	94.77%
[36]	CovXNet	89.6%	88.5%	90.3%	87.6%
		90.2%	90.8%	89.9%	89.1%
[37]	CoroNet	95%	95%	96.9%	97.5%
		89.6%	90%	89.92%	96.4%
[38]	AlexNet	78.92%	N/A	89.21%	68.63%
	VGG16	83.33%	N/A	80.39%	86.27%
	VGG19	85.29%	N/A	92.16%	78.43%
	SqueezeNet	82.84%	N/A	78.43%	87.52%
	GoogLeNet	85.29%	N/A	81.37%	90.20%
	MobileNetV2	92.16%	N/A	97.06%	87.25%
	ResNet18	91.61%	N/A	95.10%	88.23%
	ResNet50	94.12%	N/A	90.20%	100%
	ResNet101	99.51%	N/A	100%	99.02%
	Xception	99.02%	N/A	98.04%	100%
[39]	VGG16	79.01%	N/A	N/A	N/A
	DenseNet121	89.96%	N/A	N/A	N/A
	Xception	88.03%	N/A	N/A	N/A
	NASNet	85.03%	N/A	N/A	N/A
	EfficientNet	93.48%	N/A	N/A	N/A
[40]	AlexNet	99.13%	N/A	99.47%	99.15%
[41]	Majority voting method	99.31%	100%	100%	N/A
[42]	DenseNet	97.99%	98.38%	98.15%	N/A
[43]	Majority voting method	99.26%	97.87%	100%	98.89%
[44]	ResNet50V2	95.49%	96.85%	99.19%	98.27%
	VGG16	92.70%	97.50%	94.35%	98.69%
	Inception V3	92.97%	97.60%	98.39%	98.67%
[45]	VGG16	91.69%	92.33%	95.92%	100%
[46]	VGG16	87.84%	82.00%	82.33%	91.20%
	Inception V3	91.32%	87.54%	89.00%	94.00%
	EfficientNetB0	92.93%	88.30%	90.00%	95.00%
Proposed model	Stacking technique	99.23% [95% CI: 98.3–100]	98.96% [95% CI: 98–100]	99.34% [95% CI: 98.4–100]	99.75% [95% CI: 98.5–100]

The bold values mean the performance evaluation metrics obtained with our proposed model based on stacking technique. The values between [ ] lead to the confidence interval (CI) that is the standard used to quantify the uncertainty of estimating the obtained performance evaluation metrics.