AUTOMATIC METHOD FOR GLAUCOMA CLASSIFICATION USING TEXTURE ANALYSIS, XGBOOST AND GRID SEARCH

Glaucoma is an irreversible pathology, generated by increased intraocular pressure. Early detection is critical and can pre- vent total vision loss. Clinical examinations are commonly used to detect the disease. Still, the time and cost of identi- fication is quite high. This paper presents a computational methodology that aims to assist specialists in the discov- ery of glaucoma through Computer Vision techniques. The proposed methodology consists in the application of several texture descriptors combined with a parameter optimiza- tion done through Grid search with the XGBoost classifier. A result was obtained with accuracy of 82.37% and ROC of 82.08%.