Automatic methods for diagnosis of glaucoma using texture descriptors based on phylogenetic diversity

Abstract Glaucoma, a multifactorial optic neuropathy causing numerous ocular conditions in retinal and optic nerve cells, is an asymptomatic and chronic pathology; hence, early detection is essential for treatment. This study presents two automatic methods for performing optic disc delineation and glaucoma quantification. These proposed methods are based on the Otsu and k-means algorithms, which were incorporated for delimiting the optic disc region. The methods were exhaustively tested using red, green, and blue channel images extracted from the image of a retina. After segmentation, we performed characterization using texture properties based on phylogenetic diversity indices. The classification was then performed using multiple classifiers. The methodology obtained promising results. In the best case, the results obtained using the Otsu algorithm reached 100% sensitivity, 99.3% specificity, 99.6% accuracy, and a ROC curve of 0.996.