An Enhanced Approach for Automated Glaucoma Diagnosis in Retinal Fundus Images

Abstract

AbstractAccurate automated system for glaucoma diagnosis is very important since glaucoma is the second disease that leads to vision loss. Therefore, it is necessary to utilize all computational intelligence approaches to detect glaucoma accurately at its early stages in retinal images. The most significant sign of glaucoma in retinal fundus images, which helps doctors to diagnose if the patient has normal or glaucomatous image, is the ratio of vertical optic cup diameter to vertical optic disc diameter. Therefore, it is important to accurately segment both the optic disc and optic cup regions in the retinal fundus image. In this paper, our contribution proposed an approach that utilizes a combination of significant large-scale features to perform supervised superpixel classification by linear support vector machine for segmenting the optic disc and optic cup regions. We got an accuracy of 98.6% and 99.2% for disc and cup segmentation, respectively, and glaucoma diagnosis accuracy of 99%.KeywordsGlaucoma diagnosisRetinal fundus imageOptic discOptic cupSegmentation