Review of Retinal Image Analysis Methods for OCT/OCTA Based on Deep Learning

Deep learning is a branch of artificial intelligence that relies on deep neural networks for data processing and analysis.In recent years,deep learning has made significant breakthroughs in the field of medical imaging,especially in image classification,segmentation and efficacy evaluation.In the field of ophthalmology,there is an increasing need to apply deep learning techniques for efficient and accurate analysis of OCT and OCTA.Compared with traditional manual methods,deep learning methods show higher accuracy and automation in dealing with complex fundus structure and pathological changes.However,most of the previous reviews focuse on single imaging mode or single task research,and often ignore the correlation between different imaging technology,as well as the acceptability and correlation between tasks.This paper not only summarizes the commonly used data sets,systematically reviews the segmentation methods of retina-related disease biomarkers based on different OCT and OCTA devices,but also summarizes the typical classification methods of retina-related diseases from the perspective of different disease characteristics.Finally,this paper also looks forward to the future research direction from the perspectives of data privacy and security,model interpretability,and model universality,which provides a valuable reference for subsequent research.