Quantificação automática do glaucoma utilizando imagens de fundo do olho

Abstract

Glaucoma is a chronic eye disease that affects millions of people worldwide. In color eye fundus images, glaucoma appears like an erosion in the optic disk, called excavation. Usually this excavation is manually quantified by a expert. This paper proposes a morphologic-based method for detect and quantify the severity of glaucoma using color eye fundus images. To accomplish this task, it is necessary to detect the optical disk and the excavation with highest precision. The detection of these structures is based on mathematical morphology techniques. The proposed method was evaluated using images from the DIARETDB1, MESSIDOR and MESSIDOR-2 databases. In addition, local images were also used. As results, the detection method of the excavation has reached an average sensitivity and specificity of 84.06% and 99.95%, respectively. The detection method of the optic disk has reached average sensivity and specificity of 92.92% and 99.87%, respectively. Finally, the automatic quantify of the severity of glaucoma, reached an average of error of 0.02 for the vertical quantify and 0.05 for horizontal quantify.