不改变原意使其更有逻辑改写：Diabetic retinopathy DR is the most prevalent blinding disease in the world According to a study by the International Diabetes Federation IDF 1 the chance of retinopathy is up to 60 for

DR is a widespread disease that causes blindness and is prevalent globally. The IDF conducted a study [1] which found that the likelihood of developing retinopathy for diabetics over 10 years is up to 60%. Numerous studies have been conducted on DR recognition, which can be divided into two approaches: traditional image processing and deep learning models. Traditional methods involve extracting features from retinal fundus images, such as microangiomas, small hemorrhages, hard exudates, cotton wool spots, and other lesion areas. These extracted features are then used to train classifiers for image classification [2]. While traditional methods have achieved some success, the complex internal structure of retinal images makes it difficult to extract pathological features. Additionally, a significant amount of prior knowledge is required for lesion segmentation studies, making it challenging for general researchers to capture all lesion area features