Abstract
Purpose :
To present performance results of an automatic diabetic retinopathy tele-screening system operating at a network of comprehensive diabetic care clinics in Monterrey, Mexico.
Methods :
VisionQuest has developed an Automatic Retinal Imaging Screening System (ARISS) which is being used autonomously at five diabetes care clinics that are part of Clinicas del Azucar, Monterrey, Mexico, since February 2016. An average of 500 cases monthly are processed by ARISS. Images are uploaded to a central server and are processed and returned by ARISS to the clinics within 20 minutes. A Quality Assurance (QA) program monitors ARISS performance on a weekly basis. The QA program samples 20% of the cases classified as “refer in 12 months” (no DR or Mild NPDR), 100% of the cases classified as “return in 6 months” (Moderate NPDR), and 25% of the cases classified as “refer to ophthalmology” (Severe NPDR or worse, and clinically significant macular edema). ARISS has an image quality module that classifies a case as “inadequate” if at least one eye does not have a macula centered image with sufficient quality due to small pupils, media opacities, or technical errors. Sampled cases are re-read by a certified reader. Results of the QA program are reported periodically to the clinics.
Results :
To date, the QA has been performed on 518 cases of which 22% were rejected for low quality. Of these, 15% were due to photographer errors, 71% due to physiological issues (cataracts, small pupils, etc.) and the rest, 14%, were deemed adequate by the reader. ARISS’ sensitivity for referable DR was 100% while the specificity was 85%. Workload reduction for the clinics’ reading center is estimated by calculating the number of cases that do not need an ophthalmologist referral, which stood at 60%.
Conclusions :
The ARISS demonstrated clinically safe and effective performance for the detection of diabetic retinopathy in a clinical setting. The software reduces reading workload by up to 60% and referrals to ophthalmology by almost 70%. Resource scarcity and limited access to care in places like Mexico make automation a viable alternative to screen their diabetic population.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.