Abstract:Deep learningbased retinal vessel segmentation methods were summarized, while the problems of existing methods were sorted out to provide reference for further research on retinal vessel segmentation tasks. Literature related to deep learningbased retinal vessel segmentation methods from 2016 to 2022 was searched, summarized and generalized. First, it introduces the background of retinal vessel segmentation. Then, it summarizes the dataset and data preprocessing, evaluation indexes, and segmentation methods in the related literature. In the dataset and data preprocessing section, commonly used retinal image datasets are introduced, and commonly used feature enhancement techniques and data enhancement techniques are summarized. In the evaluation indexes section, commonly used evaluation indexes of retinal vessel segmentation methods are presented. In the section of segmentation methods, it categorizes these methods into those that consider network feature extraction ability, those that consider vessel characteristics, and those that consider practical applications. Among them, the methods considering the feature extraction ability of the network can be classified into methods with restricted receptive fields, methods with extended receptive fields, and methods with feature refinement, the methods considering vascular characteristics often use multilabels, loss functions, etc. to improve the problem of the difference between thick and thin vessels or the imbalance of categories; the methods considering practical applications pay more attention to the lightweighting and generalization ability of the network. Finally, it organizes and summarizes the problems of retinal vessel segmentation methods and possible future research directions. Retinal vessel segmentation methods based on deep learning will remain one of the focuses and hotspots of medical image processing research for some time in the future, and it is valuable for understanding the current status of the retinal blood vessel segmentation task and solving practical application problems.