Table 7.

COVID-19 classification accuracy (in percentage) by several state of the art models. Values in parentheses indicate 95% confidence intervals (CI).

Model	Variant	Sensitivity (CI)	Specificity (CI)	Accuracy (CI)
AlexNet	–	71.0 (57.9–81.6)	90.3 (79.5–96.5)	80.7 (68.3–89.5)
ResNet	18	71.0 (57.9–81.6)	93.5 (83.5–98.5)	82.3 (70.1–90.7)
	34	80.7 (68.3–89.5)	90.3 (79.5–96.5)	85.5 (73.7–93.1)
	50	83.9 (71.9–91.9)	90.3 (79.5–96.5)	87.1 (75.6–94.3)
	101	77.4 (64.7–89.9)	87.1 (75.6–94.3)	82.3 (70.1–90.7)
	152	77.4 (64.7–89.9)	90.3 (79.5–96.5)	83.9 (71.9–91.9)
DenseNet	121	77.4 (64.7–89.9)	93.5 (83.5–98.5)	85.5 (73.7–93.1)
	169	67.9 (83.5–98.5)	93.5 (83.5–98.5)	81.4 (68.7–90.2)
	201	90.3 (79.5–96.5)	93.5 (83.5–98.5)	91.9 (81.5–97.5)
SqueezeNet	–	66.7 (54.5–78.9)	93.5 (83.5–98.5)	81.4 (68.7–90.2)
ResNeXt	–	77.4 (64.7–86.9)	90.3 (79.5–96.5)	83.9 (71.9–91.9)