July 2018
Volume 59, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2018
Comparative analysis of the screening efficacy of a non-medical grader utilizing the Glaucoma Score versus automated grading of fundus photographs to detect glaucomatous eyes.
Author Affiliations & Notes
  • Claire Le Roux
    Ophthalmology, The Eye Centre, East London, South Africa
  • Tunde Peto
    Moorfield's Eye Hospital, London, United Kingdom
  • William Eric Sponsel
    Glaucoma Service, WESMDPA/UIW/UTSA, San Antonio, Texas, United States
  • Stephen Cook
    Ophthalmology, The Eye Centre, East London, East Cape, South Africa
  • Footnotes
    Commercial Relationships   Claire Le Roux, None; Tunde Peto, None; William Sponsel, None; Stephen Cook, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 4067. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Claire Le Roux, Tunde Peto, William Eric Sponsel, Stephen Cook; Comparative analysis of the screening efficacy of a non-medical grader utilizing the Glaucoma Score versus automated grading of fundus photographs to detect glaucomatous eyes.. Invest. Ophthalmol. Vis. Sci. 2018;59(9):4067.

      Download citation file:

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

  • Supplements

Purpose : Purpose: While screening the population for glaucoma, it is difficult to know who to refer for expensive clinical investigation. The detection of early disease is highly desirable. There is a severe shortage of ophthalmologists throughout Africa. Glaucoma Score (Fig 1) has been proven to be a useful tool to help make this determination by a non-medical grader. Automated grading could also be extremely useful, and recent developments have made this possible.

This study shows a comparison between the outcome of screening by a non-physician grader using the Glaucoma Score methodology (CL), comparing it with the outcome from an automated grader (TP) versus the clinically established doctor diagnosis as a gold standard (SC).

Methods : Between Jan-Jun 2017, 56 fundus photos from fully investigated patients attending The Eye Centre were analyzed using the Glaucoma Score algorithm by trained ophthalmic assistant (CL). The images were also analyzed using the automated grader (TP).The image set was arranged into (N) unlikely to be glaucoma, (S) glaucoma suspect, and (G) likely glaucoma. Outcomes were compared with the doctor’s diagnosis after throrough clinical assessment (SC).

Results : 56 Fundus photos were graded using the Glaucoma Prediction Score, and the automated grading system. Glaucoma Score versus doctor’s established diagnosis showed a 92.9% detection rate, while automated grading vs doctor’s diagnosis showed a 66.1% detection rate.

Conclusions : There was a strong correlation between the non-physician grader (CL) utilizing the Glaucoma Score and the Doctor’s diagnosis. The correlation between automated grader output and doctor diagnosis was less strong. This emphasizes the importance of multi-factorial risk factor inclusion (available to both the doctor and non-physician grader). In spite of the poorer correlation, the automated grader did identify a substantial majority of the appropriately referable cases.

Automated grading continues to develop as a useful tool for glaucoma screening. We advocate the use of a combination of human grading and automated grading with the glaucoma score methodology to facilitate effective detection of early glaucoma in Africa. The scarcity of Ophthalmologists necessitates task shifting. These tools will enhance the ability of non-medical staff to provide appropriate referral.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.




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.