深度学习方法在糖尿病视网膜病变方向上的应用主要是病变识别以及等级划分两项任务。基于CNN的深度学习方法在DR病变识别领域表现出有益的性能这些深度网络可以从大规模的数据集中提取特征。Liu等人开发了一种改进的双卷积层的CNN结构并应用于不同程度的DR患者进行等级划分。Tahira等人提出了一种基于快速区域的卷积神经网络FRCNN模糊k-means FKM聚类算法用于DR的自动定位和识别。Yang等人

The application of deep learning methods in the field of diabetic retinopathy (DR) mainly focuses on lesion recognition and grading. Deep neural networks based on convolutional neural networks (CNNs) have shown promising performance in DR lesion recognition by extracting features from large-scale datasets. Liu et al. developed an improved CNN structure with double convolutional layers for DR grading in patients with different severity levels. Tahira et al. proposed a fast region-based convolutional neural network (FRCNN) with fuzzy k-means (FKM) clustering algorithm for automatic localization and recognition of DR. Yang et al. proposed an automatic DR analysis algorithm using a two-stage deep learning approach that integrates local and global networks to identify lesion location and type and classify the severity of DR. Wang et al. proposed a collaborative neural network method with attention mechanism that effectively aggregates contextual and local image information for automatic segmentation of DR lesion regions in retinal images. However, this method is limited in extracting fine-grained lesion features and requires a large amount of fine-grained annotations to improve performance.