Table 1.

Overview of related literature

Publication	n	Methods	Input	Performance
Cai et al. [7]	99	random forest	automated lung and lesion segmentation, 40 clinical parameters	cv10 AUC = 0.96
Meng et al. [8]	366	CNN	CT, sex, age, severity grade, chronic disease status	test set AUC = 0.94
Ning et al. [9]	1521	DNN, CNN, logistic regression	CT, 130 clinical features	cv10 AUC = 0.86
Fu et al. [10]	64	SVM	CT-derived radiomics	leave-one-out AUC = 0.83
Li et al. [11]	217	CNN, logistic regression, SVM, decision tree, random forest	CT-derived radiomics	test set AUC = 0.86
Yue et al. [12]	52	logistic regression, random forest	CT-derived radiomics	independent test set AUC = 0.97
Wu et al. [13]	725	logistic regression	semi-automatically derived CT findings, 7 clinical features	independent test set AUC = 0.84 to 0.93
Fang et al. [14]	1040	CNN, RNN	CT, 61 clinical parameters	cv5 AUC = 0.92, domain adaptation AUC = 0.86
Wang et al. [15]	1051	CNN, DNN, random forest	CT, 15 clinical parameters	test set AUC = 0.81 to 0.83
Wang et al. [16]	188	logistic regression	CT-derived radiomics, 24 clinical parameters	test set AUC = 0.87
Revel et al. [17]	10735	logistic regression	manually-derived CT findings, 7 clinical parameters	AUC = 0.64
Lassau et al. [19]	1003	DNN	CT, 5 clinical parameters	test set AUC = 0.79
Shiri et al. [20]	14339	random forest	CT-derived radiomics	test set AUC = 0.83
Kienzle et al. [21]	2476	CNN	CT	test set F1 = 0.49
Duan et al. [23]	44	random forest	CT-derived radiomics	cv10 AUC = 0.99 to 1.00

n: total number of patients, cvx: x-fold cross-validation, DNN: deep neural network, CNN: convolutional neural network, RNN: recurrent neural network, SVM: support vector machine