June 2021
Volume 62, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2021
Topographic conversion model of outer retinal atrophy progression in intermediate age-related macular degeneration using deep learning
Author Affiliations & Notes
  • Antoine Rivail
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Gregor Sebastian Reiter
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Sophie Riedl
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Julia Mai
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Wolf-Dieter Vogl
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Christoph Grechenig
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Zhichao Wu
    Centre for Eye Research Australia, Royal Victorian Eye and Ear Hospital, East Melbourne, Victoria, Australia
    Department of Surgery, The University of Melbourne, Melbourne, Victoria, Australia
  • Robyn H Guymer
    Centre for Eye Research Australia, Royal Victorian Eye and Ear Hospital, East Melbourne, Victoria, Australia
    Department of Surgery, The University of Melbourne, Melbourne, Victoria, Australia
  • Ursula Schmidt-Erfurth
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Hrvoje Bogunovic
    Department of Ophthalmology and Optometry, Medizinische Universitat Wien, Wien, Wien, Austria
  • Footnotes
    Commercial Relationships   Antoine Rivail, None; Gregor Reiter, None; Sophie Riedl, None; Julia Mai, None; Wolf-Dieter Vogl, None; Christoph Grechenig, None; Zhichao Wu, None; Robyn Guymer, Apellis (C), Bayer (C), Genentech (C), Novartis (C), Roche (C); Ursula Schmidt-Erfurth, Genentech (C), Heidelberg Engineering (C), Kodiak (C), Novartis (C), Roche (C); Hrvoje Bogunovic, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2021, Vol.62, 130. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Antoine Rivail, Gregor Sebastian Reiter, Sophie Riedl, Julia Mai, Wolf-Dieter Vogl, Christoph Grechenig, Zhichao Wu, Robyn H Guymer, Ursula Schmidt-Erfurth, Hrvoje Bogunovic; Topographic conversion model of outer retinal atrophy progression in intermediate age-related macular degeneration using deep learning. Invest. Ophthalmol. Vis. Sci. 2021;62(8):130.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To investigate morphological risk factors and their precise quantification using deep learning algorithms for INL/OPL subsidence (at least subsidence of INL-OPL and OPL-ONL junctions) in patients with incipient macular atrophy (iMA) using a local survival model.

Methods : A series of anatomical features were extracted with deep learning methods in a cohort of eyes with bilateral large drusen at baseline: photoreceptor (PR) thickness, outer nuclear layer (ONL) thickness and drusen thickness (DT). Areas of INL/OPL subsidence were manually annotated in every eye which developed iMA during the 36-months follow-up. For each eye, the retina was divided in three zones at the visits preceding the first observed subsidence: zone of INL/OPL subsidence at first appearance, zone over drusen (>40μm height) and reference zone (rest of the 3mm circle, centered on fovea); an example is presented in figure 1. Average values for the anatomical features were computed for each zone. A mixed effects Cox survival model (one group per eye) was fitted to estimate the hazard ratio for each zone.

Results : Out of 280 eyes, 54 eyes developed iMA, of which 52 eyes, presented with INL/OPL subsidence prior to iMA, were selected. For each anatomical feature (PR, ONL and DT), the average zone values were normalized and fed to the survival model. The fitted model coefficients were -0.83 for PR (p<0.001), -0.56 for ONL (p<0.01) and -0.25 for DT (p=0.15). Negative values indicate higher risk for thinner PR, ONL and DT. Therefore, PR thinning and ONL thinning were found to be significant risk factors of INL/OPL subsidence development in iMA patients with a hazard ratio at one standard deviation under the mean of 2.86 and 2.13, respectively. However, drusen thickness was not found to be a significant factor.

Conclusions : Survival analysis of anatomical features in early INL/OPL subsidence eyes allowed us to identify PR thinning and ONL thinning as risk factors for the development of INL/OPL subsidence. This indicates that these subtle changes occur in the outer retina before becoming clinically apparent as macular atrophy. This highlights the importance of protecting PR integrity for potential early intervention in AMD. AI algorithms are essential to identify and precisely measure subclinical early morphological changes on OCT.

This is a 2021 ARVO Annual Meeting abstract.

 

×
×

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.

×