The Segment Anything Model (SAM) can be effectively used as a data annotation tool to generate pseudo labels for training medical image segmentation models, achieving comparable performance to fully supervised models.
