Deep learning models are now the go-to method for segmenting medical images. However, using these models usually requires a large set of manually labeled data, which is expensive and time-consuming.