August 2019
Volume 60, Issue 11
Open Access
ARVO Imaging in the Eye Conference Abstract  |   August 2019
Success rate of a vision screening program using a handheld fundus camera and a deep learning image quality and diabetic retinopathy screening algorithm
Author Affiliations & Notes
  • Katherine Makedonsky
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Meike Mack
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Mary Durbin
    Carl Zeiss Meditec, Inc., Dublin, California, United States
  • Footnotes
    Commercial Relationships   Katherine Makedonsky, Carl Zeiss Meditec, Inc. (E); Meike Mack, Carl Zeiss Meditec, Inc. (E); Mary Durbin, Carl Zeiss Meditec, Inc. (E)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science August 2019, Vol.60, PB098. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Katherine Makedonsky, Meike Mack, Mary Durbin; Success rate of a vision screening program using a handheld fundus camera and a deep learning image quality and diabetic retinopathy screening algorithm. Invest. Ophthalmol. Vis. Sci. 2019;60(11):PB098.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Diabetic Retinopathy (DR) affects one-third of adults with diabetes over the age of 40 according to a CDC survey. Early detection of DR is vital; however, access to care in remote areas of the world may be difficult. A platform called VISUHEALTH enables eye care providers to access retinal images and evaluate them for eye disease. VISUSCOUT® 100 (ZEISS, Jena, Germany) is a handheld fundus camera used to capture images of the retina that may be uploaded to the VISUHEALTH platform. The purpose of this study is to evaluate the success rate of an image quality and AUTO-DR algorithm within the VISUHEALTH platform on images acquired at a vision screening event.

Methods : Central images were acquired on 89 patients using the VISUSCOUT 100 at a vision screening event (Fig 1). Images for each patient were imported into the VISUHEALTH platform where an algorithm (AUTO-DR) determined the image quality and diabetic retinopathy status of the patient. An optometrist retrospectively evaluated the images and determined the success rate of the algorithm implemented within VISUHEALTH.

Results : 89 patients (Mean Age: 53 years, Standard deviation: 14) were screened with the VISUSCOUT 100. 20 of 89 patients were diabetics, and of those 17 were over the age of 40. The AUTO-DR algorithm identified 4 patients with ungradable images and 10 patients with diabetic retinopathy. An optometrist evaluated the images and determined that there were 14 patients with ungradable images, and 5 patients who had diabetic retinopathy. All of the patients with diabetic retinopathy were diabetics over the age of 40. Of the gradable images, the sensitivity of the DR algorithm was 100% and the specificity was 94%. The image quality algorithm sensitivity was 97% and specificity was 86%.

Conclusions : Based on the results from the vision screening, 29% of the diabetics over age 40 had DR. Screening for diabetic retinopathy remains very important in order to prevent future blindness. The VISUSCOUT® 100 is an easy to use camera that can be incorporated in the vision screening setting, however it is advised that an optometrist review images flagged as having diabetic retinopathy to prevent false positives. Images may be considered ungradable in the event of a cataract or pupils under 3.5 mm, therefore a referral is advised to properly evaluate the patient.

This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.

 

DR image acquired with VISUSCOUT® .

DR image acquired with VISUSCOUT® .

×
×

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.

×