Rapid Detection of Diabetic Retinopathy Through Deep Learning (#1325)

Authors

Patiño-Pérez, Darwin

Munive-Mora, Celia

Chóez-Acosta, Luis

Collantes-Farah, Alex

Molina-Calderón, Miguel

Reyez-Sánchez, Zynnia

Abstract

The disease called diabetic retinopathy (DR) has generated a series of studies and investigations that seek to be able to detect it in time, in order to counteract the damage that it can cause to the retina in the eye of people and prevent blindness; For its diagnosis, a specialist is required to detect it by analyzing an ocular image. Since rapid detection is required, an investigation was proposed with the objective of implementing and analyzing different techniques based on automatic learning or Machine Learning (ML), appropriate to solve this problem and evaluate which of them is the most efficient for the detection and classification of DR, highlighting among them those based on deep learning or deep learning that currently have become a component of artificial intelligence most used in the field of biomedicine. It was determined that one of the deep learning-based prediction models learned to perform rapid DR detection with very fine details, such as microaneurysms and some larger features such as hard exudates, by analyzing all retinographies using the process classification, reaching 85.9% accuracy in the detection of DR.Keywords: Diabetic retinopathy, Machine learning, Deep learning, Artificial neural networks, Convolutional neural networks