Investigative Ophthalmology & Visual Science Cover Image for Volume 64, Issue 9
June 2023
Volume 64, Issue 9
Open Access
ARVO Imaging in the Eye Conference Abstract  |   June 2023
Comparison of Artificial Intelligence and Physician Evaluation in Diabetic Retinopathy Screening
Author Affiliations & Notes
  • Samantha D'Amico
    William Carey University College of Osteopathic Medicine, Hattiesburg, Mississippi, United States
    Department of Surgery, Division of Ophthalmology, University of Vermont Larner College of Medicine, Burlington, Vermont, United States
  • Ian McClain
    Ophthalmology, University of Colorado Anschutz Medical Campus School of Medicine, Aurora, Colorado, United States
  • Brian Kim
    Department of Surgery, Division of Ophthalmology, University of Vermont Larner College of Medicine, Burlington, Vermont, United States
  • Footnotes
    Commercial Relationships   Samantha D'Amico, None; Ian McClain, None; Brian Kim, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, PB0035. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Samantha D'Amico, Ian McClain, Brian Kim; Comparison of Artificial Intelligence and Physician Evaluation in Diabetic Retinopathy Screening. Invest. Ophthalmol. Vis. Sci. 2023;64(9):PB0035.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Increases in the prevalence of diabetes and the finite number of trained fundus image graders present limitations to diabetic retinopathy (DR) screening capacity. The use of artificial intelligence (AI) algorithms to grade DR images can improve screening efforts by providing real-time feedback and reducing the burden on interpreting physicians. We compared the interpretation of fundus images between an AI algorithm and expert grader in patients with diabetes.

Methods : The research protocol was approved by the University of Vermont (UVM) Institutional Review Board. A retrospective search identified patients diagnosed with diabetes seen by UVM Ophthalmology and billed for retinal fundus photography between 1/17/2010 and 10/31/2019. Retrospective images were reviewed for quality and one 50 degree macula-centered image of each eye was included. The selected image of each eye was graded by an established AI algorithm and a retinal specialist masked to algorithm grade. Each image was assigned an AI-generated level of DR (no disease, mild non-proliferative DR (NPDR), moderate NPDR, severe NPDR/proliferative DR). If moderate or worse, recommendation for referral for further examination was generated.

Results : Of the 627 patients initially identified, 178 met inclusion criteria. Based on reference standard grading, 132 patients had NPDR of sufficient quantity to warrant referral. The AI system correctly identified 120 of these patients (Table 1). There were 22 false positives and 12 false negatives. Compared to the reference standard, the AI system showed a sensitivity of 90.9% (95% CI, 84.7 - 95.2%) and specificity of 52.2% (95% CI, 37.0 - 67.1%).

Conclusions : Analysis of this AI system for grading DR found a sensitivity and specificity similar to previous studies. False positives were due to other retinal pathologies present in addition to DR or mild disease, therefore further training of the algorithm can help achieve higher specificity. However, even with a specificity of 52.2%, the AI algorithm could significantly reduce the workload of physicians. Implementation of telemedicine and AI in screening for DR can help increase screening rates and avoid preventable blindness.

This abstract was presented at the 2023 ARVO Imaging in the Eye Conference, held in New Orleans, LA, April 21-22, 2023.

 

Table 1: Artificial intelligence outcome compared to reference standard grading.

Table 1: Artificial intelligence outcome compared to reference standard grading.

 

Table 2: Statistical analysis of artificial intelligence outcome compared to reference standard grading.

Table 2: Statistical analysis of artificial intelligence outcome compared to reference standard grading.

×
×

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.

×