The optic disc is a critical area for detecting glaucoma, as this condition primarily affects the optic disc and its surrounding areas. To improve the efficiency of glaucoma detection, I developed a simple yet effective threshold-based algorithm for automatically detecting the optic disc in fundus images.
By using this threshold-based method to label my data and then training a deep neural network for final optic disc localization, I was able to achieve good results. The extracted regions can also be used to create an even more robust cropping system.