notesum.ai

Published at November 18

SP${ }^3$ : Superpixel-propagated pseudo-label learning for weakly semi-supervised medical image segmentation

cs.CV

cs.AI

Released Date: November 18, 2024

Authors: Shiman Li¹, Jiayue Zhao¹, Shaolei Liu¹, Xiaokun Dai¹, Chenxi Zhang¹, Zhijian Song¹

Aff.: ¹Not specified

Arxiv: http://arxiv.org/abs/2411.11636v1

Refer to caption

Method	Dice	JI	95HD	ASD	Dice	JI	95HD	ASD
labeled ratio	10%				20%
Sparse Anno. (LB)	0.5232±0.0781	0.3754±0.0730	116.54±9.03	53.55±9.01	0.5774±0.0238	0.4349±0.0215	107.78±12.49	43.93±6.88
Dense Anno. (UB)	0.6824±0.0584	0.5709±0.0632	16.93±3.39	3.74±0.96	0.7579±0.0540	0.6511±0.0636	12.10±3.63	2.53±0.96
S2L_WSSS [24]	0.6383±0.0455	0.5043±0.0378	43.10±20.37	14.96±6.46	0.6657±0.0700	0.5370±0.0766	33.24±17.94	10.16±5.48
USTM_WSSS [27]	0.6814±0.0476	0.5526±0.0519	64.41±12.65	19.49±5.26	0.7135±0.0487	0.5810±0.0514	57.68±4.24	16.72±2.30
URPC_WSSS [25]	0.6805±0.1208	0.5498±0.1218	48.74±30.06	18.30±18.42	0.7267±0.0766	0.6228±0.0787	13.09±3.76	2.31±0.71
ICT_WSSS [62]	0.7010±0.0764	0.5790±0.0781	18.83±4.59	6.35±2.38	0.7344±0.073	0.6153±0.0781	33.50±19.00	11.91±9.02
SOUSA [16]	0.7224±0.0602	0.5980±0.0666	37.85±20.11	12.98±8.78	0.7703±0.0808	0.6519±0.0899	39.10±16.41	12.78±8.72
Ours	0.7937±0.0306	0.6740±0.0377	13.11±1.56	3.89±0.52	0.8186±0.0439	0.7127±0.0531	7.92±2.79	2.26±0.75