No abstract

Diabetic Retinopathy (DR) in Mexico is especially challenging due to i) high prevalence of diabetes in the country, ii) low rate of ophthalmologists and iii) the lack of public policies to address the DR screening. In this context, two Mexican institutions and one international financial support a three years project, to implement a DR screening program to harness the power of Artificial Intelligence (AI). In this work, we present our preliminary Deep Learning (DL) models to cropping and classifying Retinal Fundus Images (RFI) from three public datasets and one private dataset. Some of our models can achieve 93% in test accuracy and up to 98% of sensitivity. We are going to perform transfer learning with a new local dataset. We expect to improve the user-experience based on AI and reduce the DR detection time.