Table 3.

Inter-reader variability of lesion matching versus the performance of the automated lesion matching method (auto) by disease burden. Cases were divided into three disease-burden cohorts: low-(<10 lesions), intermediate- (10–29 lesions) and high- (30 or more lesions) burden. Data are reported as median (range). P-values are tests for significant differences between IRV and automated matching performance (Wilcoxon paired tests).

	Precision	Recall	F1 score	N d
Low burden (N = 14)
IRV	1.00 (0.25, 1.00)	1.00 (0.50, 1.00)	1.00 (0.40, 1.00)	0 (0, 3)
Auto	1.00 (0.50, 1.00)	1.00 (0.33, 1.00)	1.00 (0.40, 1.00)	0 (0, 3)
p	0.85	0.41	1.00	1.00
Intermediate burden (N = 17)
IRV	1.00 (0.80, 1.00)	1.00 (0.77, 1.00)	1.00 (0.79, 1.00)	0 (0, 7)
Auto	0.92 (0.59, 1.00)	0.91 (0.71, 1.00)	0.91 (0.65, 1.00)	2 (0, 14)
p	0.17	0.27	0.17	0.11
High burden (N = 9)
IRV	0.95 (0.69, 1.00)	0.91 (0.74, 1.00)	0.93 (0.72, 1.00)	5 (0, 58)
Auto	0.91 (0.66, 1.00)	0.86 (0.60, 1.00)	0.89 (0.63, 1.00)	17 (0, 59)
p	0.12	0.26	0.26	0.18