Table 1. Quantitative comparison of meniscus segmentation performance across model variants with 10 and 83 training samples. Mean and standard deviations are provided, with the highest values highlighted in bold.

Methods	Data		Whole meniscus	Lateral meniscus				Medial meniscus
Methods	Labled	Unlabeld	DSC(%)	DSC(%)	Recall(%)	Precision(%)	HD95 (pixel)	DSC(%)	Recall(%)	Precision(%)	HD95(pixel)
2D nnU-Net[12]	22	-	89.40±2.0	-	-	-	-	-	-	-	-
2D nnU-Net[12]	83	-	90.95±1.9	-	-	-	-	-	-	-	-
Jeong et al.[3](Semi-supervised learning)	22	226	90.02±2.0	-	-	-	-	-	-	-	-
Jeong et al.[3](Semi-supervised learning)	83	165	91.11±2.3	-	-	-	-	-	-	-	-
Model A	10		0	0	0	0	-	0	0	0	-
Model A	83		5.76± 5.77	0	0	0	-	11.51±11.54	16.91±21.37	10.33±8.70	93.55±41.39
Model B	10		28.96±8.20	23.77±8.37	24.55±10.42	26.24±7.77	35.98±44.99	34.15±8.02	38.54±13.68	32.02±7.37	26.18±28.44
Model B	83		73.55±4.98	69.77±5.74	71.31±6.19	68.92±8.50	6.37±3.04	71.78±5.51	69.21±7.05	75.61±8.61	6.59±3.89
Model C	10		74.09±5.18	72.07±8.15	81.38±10.26	66.73±8.12	17.79±30.16	71.63±8.62	87.28±8.29	60.44±10.05	20.07±38.66
Model C	83		89.89±2.03	89.38±2.54	90.83±4.24	88.98±6.65	2.25±0.57	90.11±2.42	93.09±4.89	87.61±7.14	2.29±0.74
Model D	10		86.20±3.49	84.87±4.77	81.77±8.67	89.87±6.66	3.58±1.21	85.55±5.14	79.08±9.93	94.22±4.84	3.45±1.61
Model D	83		91.17±2.37	91.49±2.68	93.71±3.94	89.80±5.94	2.31±0.82	90.54±2.58	92.31±4.20	89.38±6.45	2.14±0.79