Recently, with the developments of deep learning technology, deep neural networks have been widely applied in the field of medical image segmentation. Due to its good segmentation performance, U-Net has gradually become a research focus in the field of image segmentation. First, the improved works of U-Net are summarized from two perspectives: structural and non-structural improvements. Then, four medical images of retinal vessels, pulmonary nodules, liver and liver tumors, and brain tumors are used as examples to demonstrate the characteristics and segmentation difficulties of various images and to summarize the application of U-Net and its improved networks in relevant images. Finally, the problems encountered in the improvement of U-Net are discussed, and future developments are forecasted.