Medical segmentation plays an important role in adaptive radiotherapy. However, current medical segmentation methods require tons of annotated samples to achieve robust result, thus posing great pressure on data annotation. With this project, we focus on developing effecicent medical segmentation alghrithm, thus recuding labrious annotation burden. Specifically, we plan to tackle this challenging problem from three pespective, i.e., few-shot segmentation, weakly-supervised segmentation，and interactive segmetnation. By combining these tasks in the medical filed, we aims to build efficent and robust medical segmentation models. Finally, we hope our research could be secesffuly applied in the clinic use.