Enhancing the Accuracy of Glaucoma Detection from OCT Probability Maps using Convolutional Neural Networks

Annu Int Conf IEEE Eng Med Biol Soc. 2019 Jul:2019:2036-2040. doi: 10.1109/EMBC.2019.8856899.


We describe and assess convolutional neural network (CNN) models for detection of glaucoma based upon optical coherence tomography (OCT) retinal nerve fiber layer (RNFL) probability maps. CNNs pretrained on natural images performed comparably to CNNs trained solely on OCT data, and all models showed high accuracy in detecting glaucoma, with receiver operating characteristic area under the curve (AUC) scores ranging from 0.930 to 0.989. Attention-based heat maps of CNN regions of interest suggest that these models could be improved by incorporation of blood vessel location information. Such CNN models have the potential to work in tandem with human experts to maintain overall eye health and expedite detection of blindness-causing eye disease.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Glaucoma / diagnostic imaging*
  • Humans
  • Nerve Fibers
  • Neural Networks, Computer*
  • Probability
  • ROC Curve
  • Tomography, Optical Coherence*