Automated classification of optic nerve head topography images for glaucoma screening

Glaucoma prevalence is 5% and consequently it is an important public health issue to improve possibilities of an early Glaucoma treatment. We present a research strategy and first results of the development of an automated classification scheme of optic nerve head topography images for Glaucoma screening. Feature detection using the Heidelberg Retina Tomograph currently requires the manual determination of the disc margin of the optic nerve head. Procedures for the automated detection of the optic nerve head as well as vessel segmentation are discussed. Based on data accumulated in the Erlangen Glaucoma Register we evaluate linear and tree based discriminant analysis for the classification of glaucoma. In a cross-sectional study including 196 subjects we found that the use of stabilized methods like bagging and the incorporation of various anamnestic data decreases the estimated error rates.