Architectures and Applications of U-net in Medical Image Segmentation: A Review
Jundi Wang, Lei Han, Dongsheng Ran
Abstract
Recently, with the increasing application of deep learning in the medical field, convolutional neural networks, represented by U-Net, has been widely applied in medical image segmentation. The improved U-shaped network structure based on U-Net has gradually become a hot topic in medical image segmentation research. This article summarizes the improvement work related to U-Net from three perspectives: modifying skip connections, adding or replacing blocks and concatenating multiple neural networks. Then, taking the segmentation of retina, lungs, brain, abdomen, and other organs as examples, the characteristics and difficulties of various organ segmentation were introduced. Finally, a summary and outlook were made.