Automatic landmark detection in fundus photography

Abstract This chapter provides background, reasoning and sample methods for automatically detecting the optic disc and fovea centralis in retinal fundus photographs. Publicly available datasets used to train and test algorithms are enumerated and evaluation criteria are specified. Whether automating processes for disease detection or adding tools to semiautomatic retinal analysis software, the accurate detection of retinal landmarks provides the basis for future processing. This can include, but is not limited to, providing contextual information such as eye side and field of view, enhancing the ability to detect other features in the retina such as the retinal blood vessels or disease lesions, providing the areas of interest for detecting disease itself or making other retinal measurements possible. For these purposes, researchers having been developing methods for their detection for many years. This chapter is not meant as an exhaustive review, but to provide diverse examples and highlight popular approaches.