. Author manuscript; available in PMC: 2023 Apr 8.

Published in final edited form as: Med Phys. 2022 Nov 5;50(2):894–905. doi: 10.1002/mp.16053

TABLE 2.

Area under curve (AUC) mean and standard deviation values of multi-scale, domain knowledge-guided attention + random forest (MSGA+RF performance on validation set for various λ^h and λ^m (task importance) parameters

		λ^m
		0	1	10	50	100	200

*λ^h*	200	0.95 (0.04)	0.98 (0.01)	0.99 (0.01)	0.97 (0.01)	0.97 (0.04)	0.98 (0.02)
	100	0.97 (0.03)	0.98 (0.02)	0.97 (0.03)	0.95 (0.06)	0.96 (0.04)	0.97 (0.02)
	50	0.97 (0.03)	0.96 (0.03)	0.97 (0.03)	0.94 (0.05)	0.97 (0.02)	0.98 (0.02)
	10	0.95 (0.06)	0.98 (0.02)	0.97 (0.03)	0.95 (0.05)	0.99 (0)	0.96 (0.02)
	1	0.99 (0.02)	0.98 (0.02)	0.97 (0.05)	0.94 (0.05)	0.97 (0.03)	0.92 (0.08)
	0	0.97 (0.03)	0.98 (0.01)	0.99 (0.01)	0.94 (0.04)	0.95 (0.03)	0.95 (0.06)

Note: λ^h and λ^m are the relative task importance parameters in the overall loss function, representing high- and medium-resolution attentions, respectively. Three top performing combinations based on MSGA (λ^h = 200 and λ^m = 1; λ^h = 50 and λ^m = 200; = 10 and = 100) are in bold font.