Abstract: In semi-supervised medical image segmentation, two main challenges arise. First, the quality of pseudo-labels generated by segmentation networks in data-limited scenarios is often poor, ...