Table A4.

Sensitivity and specificity values for classification of severe and non-severe COVID-19 based on chest radiographs in the held-out independent test sets by the Inception-v4 models trained with different training sets and batch sizes of 16 and 32. The mean and standard deviation were estimated from the models trained with five random initializations for a given training set and batch size.

Batch Size	Models	Performance Metric	MIDRC	BrixIA	COVIDGR	UMICH
16	M	Sensitivity	0.75 ± 0.11	0.71 ± 0.15	0.80 ± 0.16	0.77 ± 0.12
		Specificity	0.79 ± 0.04	0.76 ± 0.06	0.82 ± 0.04	0.77 ± 0.04
	MB	Sensitivity	0.82 ± 0.04	0.78 ± 0.03	0.72 ± 0.09	0.82 ± 0.04
		Specificity	0.78 ± 0.04	0.81 ± 0.03	0.75 ± 0.07	0.85 ± 0.03
	MBC	Sensitivity	0.82 ± 0.05	0.78 ± 0.01	0.79 ± 0.03	0.81 ± 0.05
		Specificity	0.79 ± 0.04	0.79 ± 0.02	0.82 ± 0.06	0.84 ± 0.03
32	M	Sensitivity	0.77 ± 0.04	0.78 ± 0.04	0.69 ± 0.08	0.84 ± 0.02
		Specificity	0.77 ± 0.05	0.78 ± 0.04	0.71 ± 0.08	0.78 ± 0.02
	MB	Sensitivity	0.81 ± 0.02	0.79 ± 0.02	0.75 ± 0.06	0.84 ± 0.05
		Specificity	0.80 ± 0.03	0.79 ± 0.03	0.70 ± 0.03	0.83 ± 0.03
	MBC	Sensitivity	0.78 ± 0.04	0.81 ± 0.03	0.85 ± 0.09	0.83 ± 0.03
		Specificity	0.81 ± 0.03	0.79 ± 0.02	0.79 ± 0.10	0.83 ± 0.05