Copyright (c) 2025 Documentos de Trabajo ECBTI
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Automatic detection of diabetic retinopathy using a computer vision model and deep learning
The research focuses on the early and accurate detection of diabetic retinopathy using advanced artificial intelligence and computer vision techniques. Public databases such as DRIVE, STARE and KAGGLE are used, which contain retinal images of healthy individuals and those with diabetic retinopathy. The study applies several computer vision techniques, including grayscale conversion, adaptive binarization, contour detection, and exudate highlighting. For the classification of retinal images, convolutional neural network (CNN) models were developed using K-Fold cross-validation. A pre-trained network, VGG16, was integrated and data augmentation techniques were used to optimize classification accuracy. The effectiveness of the algorithm was evaluated and adjusted, and the results were analyzed to contribute to early and accurate detection of the disease. The model demonstrated a validation accuracy of 90.91%, indicating its ability to correctly generalize to new data.
