June 2020
Volume 61, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2020
A Deep Learning-Based Mapping of Structure to Function in Glaucoma
Author Affiliations & Notes
  • Eduardo Bicalho Mariottoni
    Vision, Imaging and Performance (VIP) Laboratory, Duke Eye Center, Durham, North Carolina, United States
    Department of Ophthalmology and Visual Sciences, Paulista School of Medicine, Universidade Federal de São Paulo, São paulo, Brazil
  • Shounak Datta
    Duke university, Durham, North Carolina, United States
  • David Dov
    Duke university, Durham, North Carolina, United States
  • Alessandro Adad Jammal
    Vision, Imaging and Performance (VIP) Laboratory, Duke Eye Center, Durham, North Carolina, United States
  • Samuel Berchuck
    Duke university, Durham, North Carolina, United States
  • Ivan Tavares
    Department of Ophthalmology and Visual Sciences, Paulista School of Medicine, Universidade Federal de São Paulo, São paulo, Brazil
  • Lawrence Carin
    Duke university, Durham, North Carolina, United States
  • Felipe Medeiros
    Vision, Imaging and Performance (VIP) Laboratory, Duke Eye Center, Durham, North Carolina, United States
  • Footnotes
    Commercial Relationships   Eduardo Mariottoni, None; Shounak Datta, None; David Dov, None; Alessandro Jammal, None; Samuel Berchuck, None; Ivan Tavares, Alcon Pharmaceuticals (F), Glaukos Corporation (F), Novartis (F); Lawrence Carin, None; Felipe Medeiros, Aeri Pharmaceuticals (C), Allergan (C), Annexon (C), Biogen (C), Biozeus (C), Carl-Zeiss Meditec (F), Carl-Zeiss Meditec (C), Galimedix (C), Google (F), Heidelberg Engineering (F), IDx (C), NGoggle, Inc. (P), Novartis (C), Reichert (F), Reichert (C), Stealth Biotherapeutics (C)
  • Footnotes
    Support  National Institutes of Health/National Eye Institute grant EY029885
Investigative Ophthalmology & Visual Science June 2020, Vol.61, 3213. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Eduardo Bicalho Mariottoni, Shounak Datta, David Dov, Alessandro Adad Jammal, Samuel Berchuck, Ivan Tavares, Lawrence Carin, Felipe Medeiros; A Deep Learning-Based Mapping of Structure to Function in Glaucoma. Invest. Ophthalmol. Vis. Sci. 2020;61(7):3213.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : To develop a deep learning (DL) structure-function (SF) map relating retinal nerve fiber layer (RNFL) damage on spectral domain optical coherence tomography (SDOCT) to functional loss on standard automated perimetry (SAP).

Methods : The study included 26,499 pairs of SAP and SDOCT from 15,173 eyes of 8,878 patients with glaucoma or suspected of having the disease extracted from the Duke Glaucoma Registry. The dataset was randomly divided at the patient level in training and test sets. A DL convolutional neural network was initially trained and validated to predict the 52 sensitivity thresholds of the 24-2 SAP from the 768 RNFL thickness points of the SDOCT peripapillary scan. Simulated localized RNFL defects of varied locations and depths were created by modifying the normal average peripapillary RNFL profile. The simulated profiles were then fed to the previously trained DL model and the topographic SF relationships between structural defects and SAP functional losses were investigated.

Results : The DL predictions had an average correlation coefficient of 0.60 (P < 0.001) with the measured values from SAP, and a mean absolute error of 4.25 dB. Simulated RNFL defects led to well-defined arcuate or paracentral visual field losses in the opposite hemifield, which varied according to the location and depth of the simulations (Fig. 1 and 2).

Conclusions : A DL algorithm was capable of predicting SAP sensitivity thresholds from SDOCT RNFL thickness measurements and generate a SF map from simulated defects. Such map may improve the understanding of how SDOCT losses translate into detectable SAP damage.

This is a 2020 ARVO Annual Meeting abstract.

 

Patterns of visual field loss predicted by the deep learning algorithm when simulating retinal nerve fiber layer (RNFL) defects in the superior hemiretina. The RNFL profile is shown on the left, with dashed vertical lines showing the location of each simulated RNFL defect. For each simulated defect in a particular location, there were 3 simulated depths representing the 10th (p10), 5th (p5) and 1st (p1) percentiles. The corresponding predicted standard automated perimetry pattern deviation plots are shown on the right.

Patterns of visual field loss predicted by the deep learning algorithm when simulating retinal nerve fiber layer (RNFL) defects in the superior hemiretina. The RNFL profile is shown on the left, with dashed vertical lines showing the location of each simulated RNFL defect. For each simulated defect in a particular location, there were 3 simulated depths representing the 10th (p10), 5th (p5) and 1st (p1) percentiles. The corresponding predicted standard automated perimetry pattern deviation plots are shown on the right.

 

Patterns of visual field loss predicted by the deep learning algorithm when simulating retinal nerve fiber layer (RNFL) defects in the inferior hemiretina. The simulated defects on the RNFL and the corresponding predicted standard automated perimetry pattern deviation plots are represented in the same fashion as Figure 1.

Patterns of visual field loss predicted by the deep learning algorithm when simulating retinal nerve fiber layer (RNFL) defects in the inferior hemiretina. The simulated defects on the RNFL and the corresponding predicted standard automated perimetry pattern deviation plots are represented in the same fashion as Figure 1.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×