Table 2 Performance evaluation of various-sized kidney tumor segmentation according to labeled data ratios in supervised and semi-supervised learning network by kidney tumor size

Methods	Network	Ratio(%)		Balanced accuracy(%)
Methods	Network	Labeled	Unlabeled	Small (tumor<2.3)	Medium (2.3≤tumor<4.3)	Large (4.3≤tumor<7.4)	Extremely Large (7.4≤tumor)
Supervised learning	U-Net[24]	100	0	57.7±12.7	62.3±10.5	74.7±13.9	84.8±12.4
	U-Net[24]	25	0	54.9±12.1	58.2±10.0	68.6±14.4	76.9±8.7
	nnU-Net[25]	100	0	62.7±14.4	70.3±14.2	87.8±13.5	94.9±3.8
	nnU-Net[25]	25	0	53.3±8.5	59.1±14.8	75.4±14.8	88.3±11.1
Semi supervised learning	MT[23]	25	75	56.4±10.9	60.6±12.0	63.1±12.8	74.0±10.1
Semi supervised learning	MT with Kidney Local Guided Map (Ours)	25	75	88.74±16.0	88.9±14.6	88.2±13.4	78.5±11.1