Deep-learning Approach To Detect Childhood Glaucoma Based On Periocular Photograph

This article presents the first automated facial diagnostic system for the detection of childhood glaucoma using a deep-learning model. This system uses a convolutional neural network based on transfer learning from a large-scale pretraining dataset to extract facial features. The system achieved a high accuracy in the binary classification of childhood glaucoma by pediatric ophthalmologists and glaucoma specialists. Visual inspection plays an important role in the diagnosis of glaucoma in childhood, and this deep-learning model allows clinicians to make non-contact diagnoses of childhood glaucoma with comparable accuracy.