The study is dedicated to the investigation of diabetic retinopathy images by digital processing methods and further pathological outcome levels classification. The application of image processing methods to the problem of diabetic retinopathy (DR) analysis is considered in the paper. In order to investigate the possibilities of machine learning for the problem of classification of retinal images, the dataset of retinal images, which represent 5 classes: absence of DR, moderate, mild, proliferate stages, and severe DR, was used in this work. The aim of this study is to identify and compare the different image processing methods used for diabetic retinopathy detection, as well as to choose the classification method that provides the highest accuracy in the identification of the human retina condition. The convolutional neural networks with tuned parameters such as EfficientNet and ResNet were applied to determine the best classification models for computerized disease screening. The accuracy and losses of the different models were determined and compared. Based on this, a combination of image preprocessing steps and neural network models, which provide the highest accuracy of diabetic retinopathy condition recognition, reaching 91.4% for the task of recognition of 5 classes (absence of DR and 4 stages of DR) is proposed. Intermediate stages in the development of diabetic retinopathy are the most difficult to distinguish: the best model showed 85.2% of correctly defined cases of moderate stage of diabetic retinopathy and 83% of correctly defined cases of mild stage. Overall, this article highlights the significance of artificial intelligence (AI) and deep learning in the detection and classification of diabetic retinopathy. It underscores the need for improved screening methods, especially in underserved areas, and emphasizes the potential of these technologies in preserving vision, reducing healthcare professionals' workload, and promoting widespread adoption in clinical practice. The article also acknowledges the challenges associated with image variability and the potential impact on AI model performance, calling for further research and improvement in image quality and consistency.