. 2021 Aug 25;9(9):1099. doi: 10.3390/healthcare9091099

Table 3.

COVID-19 screening performance of all models using a mixed dataset, presented with 95% confidence intervals (CI, $α = 0.05$ ). $T_{a}$ —Training Set; $T_{s}$ —Test Set.

Algorithm	Accuracy (%)			Precision (%)			Recall (%)			F-1 Score (%)
Algorithm	T $_{a}$	T $_{s}$	CI	T $_{a}$	T $_{s}$	CI	T $_{a}$	T $_{s}$	CI	T $_{a}$	T $_{s}$	CI
VGG16	95	91	93 ± 1.4	95	93	94 ± 1.3	95	91	93 ± 1.4	95	92	93.5 ± 1.34
InceptionResNetV2	94	93	93.5 ± 1.34	95	93	94 ± 1.3	94	93	93.5 ± 1.34	94	93	93.5 ± 1.35
ResNet50	88	85	86.5 ± 1.86	87	85	86 ± 1.89	88	85	86.5 ± 1.86	87	85	86 ± 1.89
MobileNetV2	99	91	95 ± 1.2	99	92	95.5 ± 1.13	99	91	95 ± 1.2	99	91	95 ± 1.2
ResNet101	88	86	87 ± 1.83	88	87	87.5 ± 1.80	88	86	87 ± 1.83	88	86	87 ± 1.83
VGG19	94	91	92.5 ± 1.43	94	92	93 ± 1.4	94	91	92.5 ± 1.43	94	92	93 ± 1.4