CT-based radiomics combined with signs: a valuable tool to help radiologist discriminate COVID-19 and influenza pneumonia

Table 2 1000-times Bootstrap estimate of the area under the ROC curve and the model optimism estimation for the radiomics and CT sign models

Model	Apparent AUC^a	AUC Bootstrap-Train^b (mean, 95% CI)	AUC Bootstrap-Test^c (mean, 95% CI)	Average optimism^d	Optimism-corrected AUC^e
Radiomics	0.888	0.899 (0.897–0.901)	0.838 (0.835–0.841)	0.061	0.827
CT sign	0.906	0.908 (0.907–0.910)	0.888 (0.886–0.890)	0.020	0.886

^aThe AUC of predicting model developed in original whole dataset
^bThe averaged model performance in the resampled training set after 1000-times bootstrap
^cThe averaged model performance in the r “out-of-bag” test set after 1000-times bootstrap
^dThe model’s averaged optimism as the difference between the bootstrap training set AUC and the test AUC
^eThe corrected AUC by subtracting the average optimism from the apparent AUC

ISSN: 1471-2342