Abdominal Organ Segmentation via Self Training

Hexin Dong, Zifan Chen, Jie Zhao, Mingze Yuan, Fei Yu, Jiangdong Zhang, Li Zhang

Keywords: Semantic Segmentation, Self Training

TL;DR: This paper proposes a semi-supervised abdominal organ segmentation approach based on self-training.

Abstract:

This paper proposes a semi-supervised abdominal organ segmentation approach based on self-training. Self-training(ST) helps improve the segmentation network’s decision boundary with unlabeled images and generated pseudo labels. Experiment results show that our
method has significantly outperformed the non-ST baseline, improving the mean Dice score from 0.8195 to 0.8568.