Absract. In Russia, retinal diseases rank second in the structure of visual impairment. Among all retinal diseases, diabetic retinopathy is a common cause of blindness, being a microvascular complication of diabetes mellitus. One of the factors leading to blindness and disability in patients with diabetic retinopathy is the late visit to an ophthalmologist due to the asymptomatic course of the initial stage, as well as a shortage of medical professionals, especially in remote areas. The implementation of a telemedicine system for digital screening of diabetic retinopathy may help address this issue. Purpose of the study: to develop a telemedicine model for screening diabetic retinopathy using digital tools. Materials and methods. To develop the telemedicine system for digital screening, fundus photography was conducted for patients diagnosed with diabetes mellitus. A total of 1,895 individuals over the age of 18 were included. Fundus photography was performed using a portable digital retinal camera, Optomed AURORA (Finland), in two projections. The photographs were transmitted via a secure data transmission channel from medical institutions in the Tomsk region to a centralized server at the retinal pathology center, where images were analyzed using a convolutional neural network (CNN) with classification into two categories: diabetic retinopathy and normal. The CNN was trained on datasets of fundus images labeled as normal and pathological (n = 131,905 images), including 95,339 images without pathology and 34,146 images with pathology. Since the presented dataset was not balanced across classes, a weighted error calculation algorithm with a coefficient of 2.64 in favor of the smaller class was used during training. Subsequently, a physician from the center evaluated the retinal photographs and determined the stage of the disease. The results were sent to the patient via the Medical Information System "BARS." Results. The developed telemedicine system enabled the examination of 1,895 patients with diabetes mellitus from various municipal districts of the Tomsk region, including areas without an ophthalmologist. Fundus photography was conducted by nurses in preventive medicine offices of healthcare institutions. Diabetic retinopathy was detected in 9.8% of the examined patients with diabetes. Early (non-proliferative) diabetic retinopathy was identified in 73.8% of cases, pre-proliferative stage was recorded in 8.6% of cases, and the proportion of proliferative stage was 17.6%. Depending on the stage, recommendations for further observation and treatment of diabetic retinopathy were provided to the patients. During the screening model testing, 121 cases of asymptomatic diabetic retinopathy were identified, including 23 individuals in the late proliferative stage. Conclusion. The developed telemedicine model for screening diabetic retinopathy allows for the identification of cases of diabetic retinopathy among residents of remote areas without the need for visits to ophthalmology offices, and only in cases of detected pathological changes in the fundus are patients referred for further ophthalmological examination or treatment.