Algoritmo de segmentação para triagem inicial do glaucoma com imagens de fundo do olho esquerdo

Abstract

Glaucoma, a neurodegenerative disease, is among the leading causes of irreversible visual impairment globally. Commonly associated with increased intraocular pressure, this disorder causes progressive and permanent damage to the optic nerve. Given the challenges faced in early diagnosis using traditional methods, this work proposes an approach based on classic Digital Image Processing (DIP) techniques to aid in disease identification. To this end, an algorithm was developed to segment the optic disc and cup in fundus images of the left eye, calculating the cup-to-disc ratio (CDR). This algorithm is aimed at initial glaucoma screening in fundus images, aiming to expand assessment capacity and contribute to early interventions. The public database used was Kaggle, composed of 5,000 images with various pathologies. Applying the methodology to images of the left eye of glaucoma patients, the developed algorithm achieved an accuracy rate of 84.85% in identifying regions of interest. Despite the limitations of the database images, the proposed screening offers a promising starting point for improvements and future investigations, serving as a foundation for advances in the field of diagnostic imaging.