July 2018
Volume 59, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2018
A Receiver Operating Characteristic Analysis for Accuracy of Remote Image Grading of Retinopathy of Prematurity (ROP) in the Telemedicine Approaches to Evaluating of Acute-Phase ROP (e-ROP) Study
Author Affiliations & Notes
  • Wei Pan
    Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Graham E Quinn
    Ophthalmology, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, United States
  • Ebenezer Daniel
    Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Agnieshka Baumritter
    Ophthalmology, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, United States
  • Gui-Shuang Ying
    Ophthalmology, University of Pennsylvania, Philadelphia, Pennsylvania, United States
  • Footnotes
    Commercial Relationships   Wei Pan, None; Graham Quinn, None; Ebenezer Daniel, None; Agnieshka Baumritter, None; Gui-Shuang Ying, Chengdu Kanghong Biotech co. Ltd (C), Ziemer Ophthalmic Systems AG (C)
  • Footnotes
    Support  National Eye Institute of the National Institutes of Health, Department of Health and Human Services U10 EY017014 and R21EY025686.
Investigative Ophthalmology & Visual Science July 2018, Vol.59, 2775. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Wei Pan, Graham E Quinn, Ebenezer Daniel, Agnieshka Baumritter, Gui-Shuang Ying; A Receiver Operating Characteristic Analysis for Accuracy of Remote Image Grading of Retinopathy of Prematurity (ROP) in the Telemedicine Approaches to Evaluating of Acute-Phase ROP (e-ROP) Study. Invest. Ophthalmol. Vis. Sci. 2018;59(9):2775.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To evaluate the accuracy of remote image grading for retinopathy of prematurity (ROP) by trained readers using receiver operating characteristic (ROC) analysis.

Methods : Secondary analyses of data from infants with birth weight (BW) of <1251g. Infants underwent serial diagnostic examinations by an ophthalmologist and digital imaging by non-physician staff using a wide-field digital camera. Two masked non-physician trained readers in a remote image reading center independently evaluated retinal image sets (6-images per eye) for ROP stage, zone, and posterior pole vascular abnormality. The reading supervisor adjudicated discrepancies. For each diagnostic exam and image grading, an eye at each session was classified into one of four categories: no ROP, mild ROP, type 2 ROP and type 1 ROP following ETROP criteria. ROC analyses were performed using 3 cutpoints (mild ROP, type 2 ROP, type 1 ROP) of image grading by comparing to the reference standard of diagnostic examination. The analyses were performed based on per eye per session, per infant per session, per eye any session and per infant any session.

Results : Among 7822 sessions of 1238 infants enrolled from 13 North American centers, diagnostic exam found any ROP in 4866 sessions of 1528 eyes from 790 infants, type 2 ROP in 540 sessions of 185 eyes from 94 infants, and type 1 ROP in 252 sessions of 251 eyes in 141 infants. The sensitivity and specificity was 65%-75% and 89%-94% for any ROP, 78%-95% and 79-85% for type 2 ROP or worse, 49-67% and 95%-97% for type 1 ROP (Table 1). The area under ROC curve ranged from 0.79-0.85 for any ROP, 0.87-0.92 for type 2 ROP or worse, and 0.91-1-0.92 for type 1 ROP (Table 2).

Conclusions : ROC analyses confirm that remote image evaluation by non-physician trained readers has high accuracy in detecting ROP, particularly ROP type 2 or worse.

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.

×