Table 1 Performance evaluation of meniscus segmentation according to labeled data and unlabeled data ratios in supervised and semi-supervised learning network

Methods	Network	Ratio(%)		Metrics(%)
Methods	Network	Labeled	Unlabeled	Balanced accuracy	F1-score	Recall	Precision
Supervised learning	2D nnU-Net[18]	22	-	94.42±2.0	89.40±1.6	89.16±4.2	90.09±4.8
Supervised learning	2D nnU-Net[18]	83	-	96.07±1.7	90.95±1.9	92.48±3.7	89.86±5.2
Semi supervised learning	MT[15]	22	63	94.92±2.0	89.90±1.9	90.17±4.2	90.06±5.1
	MT+AMB			94.92±1.6	83.27±3.9	90.75±3.4	77.47±7.6
	MT+AMB+w^s			95.09±1.7	89.18±1.9	90.59±3.7	88.20±5.2
	MT+AMB+w^t			95.56±1.8	89.72±1.8	91.52±3.8	88.39±5.1
	MT+AMB+w^s+w^t			96.42±1.6	90.02±2.1	93.28±3.5	87.39±5.6
	MT+AMB+w^s+w^t	22	226	96.87±1.6	88.81±2.6	94.32±3.4	84.35±6.3
	MT+AMB+w^s+w^t	83	165	97.02±1.5	91.11±2.3	94.45±3.2	88.39±5.8