A multicenter evaluation of a deep learning software (LungQuant) for lung parenchyma characterization in COVID-19 pneumonia

European Radiology Experimental

Table 4 AUC, Youden’s cutoff, and sigmoid-fit inflection point on the LungQuant outputs versus the respective dichotomized clinical metrics

Metric	Strata	LungQuant cutoff	AUC	Inflection [95% CL] linear constrained	Inflection [95% CL] unconstrained
CTSS	All	0.10	0.98	0.20 [0.19 0.21]	0.23 [0.23 0.24]
	> 400 COVID-19 cases	0.06	0.95	0.17 [0.16 0.18]	0.23 [0.22 0.25]
	> 10 years experience	0.09	0.97	0.20 [0.20 0.21]	0.25 [0.23 0.27]
Bilateral	All	0.60	0.85	0.64 [0.52 0.76]	0.77 [0.76 0.78]
	> 400 COVID-19 cases	0.60	0.89	0.64 [0.53 0.76]	0.87 [0.87 0.87]
	> 10 years experience	0.60	0.85	0.63 [0.50 0.75]	0.67 [0.66 0.68]
Basal predominant	All	0.34	0.90	0.32 [0.31 0.33]	0.29 [0.28 0.29]
	> 400 COVID-19 cases	0.32	0.89	0.33 [0.32 0.34]	0.29 [0.29 0.30]
	> 10 years experience	0.34	0.91	0.31 [0.30 0.33]	0.28 [0.27 0.28]
Type	All	0.15	0.81	0.18 [0.12 0.25]	0.26 [0.23 0.29]
	> 400 COVID-19 cases	0.16	0.80	0.19 [0.12 0.25]	0.29 [0.25 0.33]
	> 10 years experience	0.15	0.73	0.16 [0.09 0.22]	0.22 [0.19 0.25]

AUC Area under the curve at receiver operating characteristics analysis, CL Confidence limits, CTSS Computed tomography severity score, COVID-19 Coronavirus disease 2019