Abstract: To improve the segmentation accuracy of regions of interest in the diagnosis of various neurological diseases and promote further application of convolutional neural networks (CNN)-based semantic segmentation, this paper summarized the applications of the semantic segmentation method based on CNN in a variety of neuroimage research. First, the current CNN architectures, as well as a variety of classical models based on CNN semantic segmentation and their variant were reviewed. Then, a comprehensive coverage of their application in neuroimage area was provided. Finally, future directions and challenges in this important field of research were discussed.