Information systems for Convolutional Neural Network in diabetes detection using fundus imaging

Villegas-Cubas J.E.; Capuñay-Uceda O.E.; Coronado-Navarro A.F.; Delgado-Chavarri A.H.; Osores-Granda O.E. · RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao (2022)

Resumen

Diabetes is considered a critical illness, and its early detection improves patient care and quality of life. This research develops a convolutional neural network for diabetes detection using fundus images from the IDRID image set. The performance analysis of the convolutional neural network applies a cross-validation and is compared with other models used in previous studies based on methods such as Random Forest, Extra Trees, SVM and AdaBoost. Our CNN model consists of an input layer, three convolutional layers, three pooling layers, two fully connected layers, and an output layer with two neurons. The results of the evaluation of the convolutional neural network model with 103 images produced an accuracy of 83.50%, which represents a better performance compared to previous studies.

Instituciones: Universidad Tecnológica del Perú