Classification of COVID-19 Chest CT Images Based on Ensemble Deep Learning

Table 1

Overview of methods and quantitative results toward COVID-19 classification.

Author	Dataset	No. of images	Method	Quantitative results indicators
Gao [9]	Internal	791	XGBoost	Acc = 94.34%; Sens = 83.33%
Wang [10]	Internal	540	3D CNN	Acc = 90.1% ;ROC = 95.5%
Han [11]	Internal	460	Attention mechanism + 3D multiple instance learning	Acc = 97.9%; AUC = 99.0%
J. HORRY [12]	COVID-CT dataset	746	VGG19	Acc = 84%
A. Waheed [13]	COVID-19 chest X-ray dataset [14–16]	932	VGG16 + ACGAN	Acc = 95%; Sens = 90%
Pathak [17]	Chest CT images [18]	1790	DBM	Acc = 98.37%; AUC = 98.32%
Y. Oh [19]	JSRT [20]	502	ResNet-18	Acc = 88.9%; Spec = 96.4
Wang [21]	RSNA [22]; chest X-ray [23]	18567	ResNet-101 + ResNet-102	Acc = 96.1%
Ouyang [24]	Internal	2796	Attention mechanism + 3D CNN	Acc = 87.5%; AUC = 94.4%; Sens = 86.9%
T. Siswantining [25]	Internal	170	CNN + SVM + NN	Acc = 95%
Dong [26]	Internal	640	DCNN	Acc = 93.64 ± 1.42% Sens = 93.28 ± 1.5% Spec = 94.0 ± 1.56%
Zhang [1]	CC-CCI [1]	61775	3D Resnet-18	Acc = 92.49%; Sens = 94.93%; Spec = 91.13%

Internal is the nonpublic dataset.