A Soft Label Method for Medical Image Segmentation with Multirater Annotations

Table 2

Quantitative results with different strategies on the MNIST and RIGA test set.

Methods		MNIST		RIGA
		Hard	Mode-UNet	62.89	57.30	96.90	82.41	94.62	75.09
MV-UNet	89.14		80.59	97.03	84.92	94.35	73.47
STAPLE-UNet	82.26		74.51	96.28	85.37	92.84	75.68
Soft	Average-UNet	90.54	82.85	97.04	85.40	94.52	76.58
	GLS-UNet	87.32	78.29	96.14	86.83	93.71	75.95
	-UNet	90.50	81.03	96.85	84.71	94.33	77.18
	-UNet	87.67	80.13	96.77	86.13	93.82	77.90
	Mixup-UNet	86.61	78.58	96.83	84.72	94.02	75.18
	SVLS-UNet	90.32	82.05	97.40	86.09	94.95	76.87
SOTA	LNL	84.52	76.33	97.67	87.56	95.46	78.76
	MRNet	93.63	88.09	97.60	86.54	95.78	78.19
	Ours	94.76	90.82	97.98	89.85	96.04	81.97

The best results are highlighted, and the second best results are italic.