Segmentação automática de regiões de interesse em imagens de retinografia

Abstract

One of the main causes of irreversible blindness in Brazil and worldwide, glaucoma is a silent disease caused by increased intraocular pressure (IOP). In the early stages of this disease, it does not present symptoms, making its diagnosis and early treatment difficult. With the increase in intraocular pressure, the optic nerve suffers damage as well as other parts of the eye. These lesions cause physical changes that can be observed through imaging examinations such as Retinography (Funds of the Eye Imaging). Fundus of the eye images contain valuable information about the status of the retina, vessels, the optic nerve, and diagnostic information. This work presents a tool for segmenting regions of interest in retinal examinations. The proposed methodology for this tool comprises image processing and computer vision techniques for segmenting 460 images from the datasetRetinal fundus images for glaucoma analysis (RIGAdataset) in the MESSIDOR image set. The regions of interest obtained in the retinography image are the optic disc and the excavation, where a Dice score of 86.8% was obtained in the optic disc segmentation and Dice 68.7% in the cup segmentation.