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

European Radiology Experimental

Table 3 Intraclass correlation coefficient values for the clinical metrics computed for all readers and stratified by COVID-19 expertise and scan quality

Metric	# Rad	owaA	raaA	twrssC	twrmaC	twrssR	twrmaR	Sample	Strata
CTSS	14	0.63	0.96	0.70	0.97	0.63	0.96	120	All
	3	0.63	0.84	0.72	0.88	0.64	0.84	120	> 400
	3	0.64	0.84	0.68	0.86	0.65	0.85	120	< 100
	8	0.60	0.92	0.69	0.95	0.61	0.93	120	100−400
	14	0.69	0.97	0.75	0.98	0.69	0.97	83	>Acceptable
Lesion type	14	0.45	0.92	0.49	0.93	0.45	0.92	120	All
	3	0.41	0.68	0.49	0.75	0.44	0.70	120	> 400
	3	0.38	0.65	0.48	0.73	0.41	0.67	120	< 100
	8	0.46	0.87	0.48	0.88	0.46	0.87	120	100−400
	14	0.43	0.91	0.47	0.93	0.43	0.91	83	>Acceptable
Bilateral	14	0.76	0.98	0.76	0.98	0.76	0.98	120	All
	3	0.75	0.90	0.76	0.91	0.75	0.90	120	> 400
	3	0.73	0.89	0.73	0.89	0.73	0.89	120	< 100
	8	0.77	0.96	0.77	0.96	0.77	0.96	120	100−400
	14	0.79	0.98	0.80	0.98	0.79	0.98	83	>Acceptable
Basal predominant	14	0.52	0.94	0.53	0.94	0.52	0.94	120	All
	3	0.53	0.77	0.54	0.78	0.53	0.77	120	> 400
	3	0.60	0.82	0.61	0.82	0.60	0.82	120	< 100
	8	0.48	0.88	0.50	0.89	0.49	0.88	120	100−400
	14	0.53	0.94	0.54	0.94	0.53	0.94	83	>Acceptable

Scans being labeled at least once as “low quality” were excluded
# Rad Number of radiologists in the stratum, Sample Number of scans in the stratum, owaA One-way absolute agreement, raaA Random-average absolute agreement, twrssC Two-way-random single-score consistency, twrmaC Two-way-random mixed-average consistency, twrssR Two-way-random single-score reliability, twrmaR Two-way-random mixed-average reliability