This application describes a system to estimate the likelihood of glaucoma from fundus images. The system may include receiving an optic nerve head region of an original image. A variation to the optic nerve head region are applied. A first primary open angle glaucoma likelihood estimate based on the first varied optic nerve head region is generated using a trained convolutional neural network. A second primary open angle glaucoma likelihood estimate based on the second varied optic nerve head region is generated using a trained convolutional neural network. The first primary open angle glaucoma likelihood estimate and the second primary open angle glaucoma likelihood estimate are combined into a final primary open angle glaucoma likelihood estimate. The final primary open angle glaucoma likelihood estimate is presented to a user.