Methodology for the Creation of a Medical Database: Case of Fundus imaging

Abstract

Currently, recognition systems based on the use of Artificial Intelligence techniques, are being used more frequently. For these systems to be trained, it is necessary to have a series of training images, known as dataset of images. This research study demonstrates a novel method to create a dataset of different types of images, with a demonstration applied to fundus images, in order to recognize exudates that are the first symptoms of diabetic retinopathy. The proposed method considers original fundus images, which identify characteristics of some pathology, and in this case, diabetic retinopathy. With these images, groups of images are generated to be able to build a dataset of images and that these can be used in the design of classification algorithms. As a result, this study presents a new dataset corresponding to image areas with presence of hard exudates. The proposed method can be scaled to different types and modalities of images.