June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
Geographic atrophy measured by machine learning and manual segmentation on optical coherence tomography in non-neovascular age-related macular degeneration
Author Affiliations & Notes
  • Phuoc-Hanh Le
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Sumit Sharma
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Kimberly Baynes
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Thuy K Le
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Kubra Sarici
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Leina Lunasco
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Gagan Kalra
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Katherine Wise
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Carmen Calabrise
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Nora LaMunyon
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Christopher J. Mugnaini
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Josie Friedl
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Danielle Burton
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Justis P Ehlers
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Sunil K Srivastava
    Cole Eye Institute, Cleveland Clinic, Cleveland, Ohio, United States
  • Footnotes
    Commercial Relationships   Phuoc-Hanh Le None; Sumit Sharma Allergan, Bausch and Lomb, Genentech, Regeneron, Alimera, Clearside, Eyepoint, Code C (Consultant/Contractor); Kimberly Baynes None; Thuy Le None; Kubra Sarici None; Leina Lunasco None; Gagan Kalra None; Katherine Wise None; Carmen Calabrise None; Nora LaMunyon None; Christopher Mugnaini None; Josie Friedl None; Danielle Burton None; Justis Ehlers Aerpio, Novartis, Zeiss, Alcon, Leica, Santen, Allergan, Genentech, Regeneron, Adverum, Allegro, Thrombogenics, Stealth, Code C (Consultant/Contractor), Aerpio, Boehringer-Ingelheim, Alcon, Allergan, Regeneron, Genentech, Thrombogenics, Novartis, Code F (Financial Support), Leica, Code P (Patent); Sunil Srivastava Novartis, Regeneron, Bausch and Lomb, Eyepoint, Eyevensys, Abbvie, Zeiss, Code C (Consultant/Contractor), Eyepoint, Regeneron, Allergan, Santen, Code F (Financial Support)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 3022 – F0292. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Phuoc-Hanh Le, Sumit Sharma, Kimberly Baynes, Thuy K Le, Kubra Sarici, Leina Lunasco, Gagan Kalra, Katherine Wise, Carmen Calabrise, Nora LaMunyon, Christopher J. Mugnaini, Josie Friedl, Danielle Burton, Justis P Ehlers, Sunil K Srivastava; Geographic atrophy measured by machine learning and manual segmentation on optical coherence tomography in non-neovascular age-related macular degeneration. Invest. Ophthalmol. Vis. Sci. 2022;63(7):3022 – F0292.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Non-neovascular age-related macular degeneration (NNAMD) is associated with loss of ellipsoid zone (EZ) on optical coherence tomography (OCT). Geographic atrophy (GA) in NNAMD can be identified and measured using sub-RPE (retinal pigment epithelium) illumination area (IA) on commercially available software. In this study we compare sub-RPE IA changes using a commercial software program to changes in EZ loss on OCT identified by machine learning analysis with manual correction in patients with NNAMD with GA.

Methods : This is a retrospective review of patients diagnosed with NNAMD and GA with growth. IA was measured by the Advanced RPE Analysis feature of FORUM Viewer (Carl Zeiss Meditec, Inc.) which measures the area (mm2) exhibiting sub-RPE hyper-reflectance within a 2.5-mm radius from the fovea on OCT due to loss of outer retina. IA was manually corrected where needed. OCTs were also analyzed by machine learning algorithms followed by manual correction of layer segmentation using custom software. EZ loss was measured by the percentage of the 6x6-mm scan where EZ thickness was zero µm. Correlation was calculated using Pearson correlation coefficient.

Results : 60 eyes of 60 patients were included in the study with a mean follow-up length of 5.1 years. At baseline, mean age was 79 and mean IA was 3.0 mm2. Mean EZ loss measured 1.67 mm2 after machine learning analysis and 5.3 mm2 after manual correction. IA strongly correlated to EZ loss measured by machine learning (r = .85) and manual correction (r = .93) on 60 OCTs. Area where EZ measured 10 or less microns was 3.4 mm2 after machine learning and 6.1 mm2 after manual correction. Mean annual growth of IA was 1.15 mm2/yr while mean annual change in EZ loss on manually corrected OCTs was 1.4 mm2/yr.

Conclusions : EZ measurements on OCT can be used to estimate GA in NNAMD. IA correlates strongly to EZ loss measured by machine learning and manual segmentation. GA area measures smaller than area of EZ loss and can take several years of growth to reach the area of EZ loss. IA and OCT layer segmentation should be compared to measurements of GA on fundus autofluorescence to determine their potential as outcomes measures in clinical research.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

(A) IA measured by Zeiss compared to EZ thickness in µm measured by (B) machine learning analysis and (C) manual segmentation on the same scan in one eye.

(A) IA measured by Zeiss compared to EZ thickness in µm measured by (B) machine learning analysis and (C) manual segmentation on the same scan in one eye.

×
×

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.

×