Cluster fusion based cross teaching for semi-supervised medical image segmentation

Huaikun Zhang, Xiangyu Lu, Pei Ma, Jizhao Liu, Jing Lian, Yide Ma

Abstract: Highlights•A novel cluster based semi-supervised architecture for medical image segmentation.•The proposed scheme considers semantic segmentation as a cluster assignment process.•CSCT introduces cluster swapping between sub-networks, enhancing overall performance.•Outperforms state-of-the-art: Better Dice, Jaccard, HD, and ASD on three public datasets.