June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Long-term rate of optic disc rim loss in glaucoma patients measured from optic disc photographs with a deep neural network
Author Affiliations & Notes
  • Sangwook Jin
    Jules Stein Eye Institute, Los Angeles, California, United States
  • Esteban Morales
    Jules Stein Eye Institute, Los Angeles, California, United States
  • Ella Bouris
    Jules Stein Eye Institute, Los Angeles, California, United States
  • Joseph Caprioli
    Jules Stein Eye Institute, Los Angeles, California, United States
  • Footnotes
    Commercial Relationships   Sangwook Jin None; Esteban Morales None; Ella Bouris None; Joseph Caprioli None
  • Footnotes
    Support  Research to Prevent Blindness (departmental grant), Simms/Mann Family Foundation, Payden Glaucoma Research Fund
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 374. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Sangwook Jin, Esteban Morales, Ella Bouris, Joseph Caprioli; Long-term rate of optic disc rim loss in glaucoma patients measured from optic disc photographs with a deep neural network. Invest. Ophthalmol. Vis. Sci. 2023;64(8):374.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : This study uses neural-network generated rim-to-disc area ratio (RDAR) measurements to measure the rate of optic disc rim loss in a large cohort of glaucoma patients

Methods : 16277 optic disc photographs (ODPs) had adequate quality for automated segmentation. Each eye had ≥2 ODPs and ≥2 years of follow up. Data collected included demographics, diagnosis, intraocular pressure (IOP), RDAR, and mean deviation (MD) values. A deep neural network was used to automate the optic disc rim segmentation and output RDAR for each ODP (Rasheed, et al, Ophthalmol Science 2022). Linear regression of RDAR was used to measure the rate of rim loss.

Results : 13695 ODPs from 4129 eyes of 2420 patients were included. Most patients were female (1441/2430, 60%), Caucasian (1364/2420, 56%), and had primary open-angle glaucoma (2942/4129 eyes, 71%), or were suspicious for glaucoma (845/4129, 20%). Mean (±SD) age at baseline was 62.7(±12.6) and median (IQR) duration of follow up was 5.0(6.4) years. Mean IOP was 14.38(±3.22) mmHg. Median baseline and final MD were -2.66(4.01) dB and -2.12(5.66) dB, respectively. Mean baseline and final RDAR were significantly different for glaucoma patients, 0.57(±0.18) and 0.49(±0.19), respectively (p = 0.00), and glaucoma suspects, 0.67(±0.13) and 0.63(±0.15). For patients with glaucoma, the average rate of change of RDAR was -0.02 per year. This difference was weakly correlated with change in MD (R2 = 0.03, p = 0.00). For glaucoma suspects, the rate was -0.01/year and was not significantly correlated with MD (R2 = 0, p = 0.27).

Conclusions : RDAR was significantly lower after a follow up period of 5 years, decreasing by 0.07±0.14, or -0.02/year, in patients with glaucoma. Patients who were suspect for glaucoma had a slower rate of rim area loss, -0.01/year. Automated segmentation of ODPs and calculation of RDAR with a neural network can be used to evaluate the extent and rate of optic disc rim loss and is further evidence of long-term trend towards nerve fiber loss in patients with glaucoma, even under treatment.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Linear regression of RDAR for all eyes (red=glaucoma, green=glaucoma suspect)

Linear regression of RDAR for all eyes (red=glaucoma, green=glaucoma suspect)

 

Frequency distribution of change in RDAR between baseline and final optic disc photograph

Frequency distribution of change in RDAR between baseline and final optic disc photograph

×
×

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.

×