This study aims to integrate clinical indicators and features of fundus images, combined with metabolomics, to construct an early warning model for diabetic retinopathy (DR) in type 2 diabetes. By combining clinical indicators with metabolomics, the investigators aim to establish a precise DR typing model based on the age of diabetes onset (early-onset diabetes, late-onset diabetes) and based on the coexistence with two types of diabetic macrovascular complications. A multidisciplinary collaboration will be conducted for comprehensive management of DR to control the progression of moderate-stage DR. Cloud-based patient rooms combined with continuous glucose monitoring (CGM) will further explore the role of integrated diabetes retinopathy ward management models in the management of patients undergoing diabetes retinopathy surgery, pioneering a new model for the management of advanced DR.