TY - JOUR
T1 - Mutual consistency learning for semi-supervised medical image segmentation
AU - Wu, Yicheng
AU - Ge, Zongyuan
AU - Zhang, Donghao
AU - Xu, Minfeng
AU - Zhang, Lei
AU - Xia, Yong
AU - Cai, Jianfei
N1 - Funding Information:
This work was supported in part by the Monash FIT Start-up Grant, and in part by the National Natural Science Foundation of China under Grant 62171377 , and in part by the Key Research and Development Program of Shaanxi Province under Grant 2022GY-084. We also appreciate the efforts to collect and share the datasets (Xiong et al., 2021; Clark et al., 2013; Bernard 485 et al., 2018) and several public benchmarks (Yu et al., 2019; Li et al., 2020b; Luo et al., 2021a,b; Luo, 2020).
Publisher Copyright:
© 2022 Elsevier B.V.
PY - 2022/10
Y1 - 2022/10
N2 - In this paper, we propose a novel mutual consistency network (MC-Net+) to effectively exploit the unlabeled data for semi-supervised medical image segmentation. The MC-Net+ model is motivated by the observation that deep models trained with limited annotations are prone to output highly uncertain and easily mis-classified predictions in the ambiguous regions (e.g., adhesive edges or thin branches) for medical image segmentation. Leveraging these challenging samples can make the semi-supervised segmentation model training more effective. Therefore, our proposed MC-Net+ model consists of two new designs. First, the model contains one shared encoder and multiple slightly different decoders (i.e., using different up-sampling strategies). The statistical discrepancy of multiple decoders’ outputs is computed to denote the model's uncertainty, which indicates the unlabeled hard regions. Second, we apply a novel mutual consistency constraint between one decoder's probability output and other decoders’ soft pseudo labels. In this way, we minimize the discrepancy of multiple outputs (i.e., the model uncertainty) during training and force the model to generate invariant results in such challenging regions, aiming at regularizing the model training. We compared the segmentation results of our MC-Net+ model with five state-of-the-art semi-supervised approaches on three public medical datasets. Extension experiments with two standard semi-supervised settings demonstrate the superior performance of our model over other methods, which sets a new state of the art for semi-supervised medical image segmentation. Our code is released publicly at https://github.com/ycwu1997/MC-Net.
AB - In this paper, we propose a novel mutual consistency network (MC-Net+) to effectively exploit the unlabeled data for semi-supervised medical image segmentation. The MC-Net+ model is motivated by the observation that deep models trained with limited annotations are prone to output highly uncertain and easily mis-classified predictions in the ambiguous regions (e.g., adhesive edges or thin branches) for medical image segmentation. Leveraging these challenging samples can make the semi-supervised segmentation model training more effective. Therefore, our proposed MC-Net+ model consists of two new designs. First, the model contains one shared encoder and multiple slightly different decoders (i.e., using different up-sampling strategies). The statistical discrepancy of multiple decoders’ outputs is computed to denote the model's uncertainty, which indicates the unlabeled hard regions. Second, we apply a novel mutual consistency constraint between one decoder's probability output and other decoders’ soft pseudo labels. In this way, we minimize the discrepancy of multiple outputs (i.e., the model uncertainty) during training and force the model to generate invariant results in such challenging regions, aiming at regularizing the model training. We compared the segmentation results of our MC-Net+ model with five state-of-the-art semi-supervised approaches on three public medical datasets. Extension experiments with two standard semi-supervised settings demonstrate the superior performance of our model over other methods, which sets a new state of the art for semi-supervised medical image segmentation. Our code is released publicly at https://github.com/ycwu1997/MC-Net.
KW - Medical image segmentation
KW - Mutual consistency
KW - Semi-supervised learning
KW - Soft pseudo label
UR - http://www.scopus.com/inward/record.url?scp=85134601962&partnerID=8YFLogxK
U2 - 10.1016/j.media.2022.102530
DO - 10.1016/j.media.2022.102530
M3 - Article
C2 - 35839737
AN - SCOPUS:85134601962
SN - 1361-8415
VL - 81
JO - Medical Image Analysis
JF - Medical Image Analysis
M1 - 102530
ER -