Abstract: Semi-supervised learning is becoming increasingly popular in medical image segmentation because of its ability to exploit large amounts of unlabelled data to extract additional information.