Short Report
Evaluation of artificial intelligence-based telemedicine screening for retinopathy of prematurity

https://doi.org/10.1016/j.jaapos.2020.01.014Get rights and content

Retrospective evaluation of a deep learning–derived retinopathy of prematurity (ROP) vascular severity score in an operational ROP screening program demonstrated high diagnostic performance for detection of type 2 or worse ROP. To our knowledge, this is the first report in the literature that evaluated the use of artificial intelligence for ROP screening and represents a proof of concept. With further prospective validation, this technology might improve the accuracy, efficiency, and objectivity of diagnosis and facilitate earlier detection of disease progression in patients with potentially blinding ROP.

Section snippets

Methods

Institutional Review Board approval was obtained at the Oregon Health & Science University, Portland, to retrospectively evaluate images from an existing ROP screening telemedicine program at Salem Hospital, Salem, Oregon. We included all subjects who met ROP screening criteria between September 2015 and June 2018 and had one or more telemedicine examination. All subjects had wide-angle retinal images taken by a neonatal nurse practitioner trained to use the RetCam (Natus Medical Incorporated,

Results

A total of 81 subjects met inclusion criteria (summarized in Figure 1), and 613 telemedicine eye examinations were analyzed. The mean (with standard deviation) number of telemedicine examinations for each infant was 3.8 ± 2.3 (range, 1-10). The mean postmenstrual age (PMA) at birth for included infants was 29.2 ± 2.1 weeks, with mean weight at birth of 1240 ± 235 g. Four eyes of 2 patients (2%) developed RR-ROP during telemedicine screening. Mean age for infants requiring transfer was 25 weeks;

Discussion

These results demonstrate proof of concept that AI may have a role in the screening of patients at risk for ROP. This study has several limitations. Our sample size was limited by actual ROP screening needs at a community hospital over 3 years, and may or may not be representative of other ROP screening populations. We included all image encounters in the relevant time period, and both image quality and field of view of the manually selected images were variable. Further work is needed to

Acknowledgments

The authors thank Dr. Dongseok Choi for his assistance with the analysis for this project, and the staff in the community hospital NICU for their invaluable help with this telemedicine ROP program.

References (7)

There are more references available in the full text version of this article.

Cited by (35)

  • A deep learning framework for the detection of Plus disease in retinal fundus images of preterm infants

    2021, Biocybernetics and Biomedical Engineering
    Citation Excerpt :

    The clinicians often recommend initial stage ROP screening for premature babies having low birth weight and who are born before 31 weeks of gestation. As tele-medicine based diagnostics are becoming wide spread, the amount of images that needs to be screened are too high and manual examination of these images become very difficult [3,4]. Recent advancements in neonatal care increased the survival rate of low birth weight preterm infants.

View all citing articles on Scopus

Funding support: NIH grants R01EY19474, P30EY10572, P30 EY001792, R01EY19474 and K12EY27720 (Bethesda, MD), NSF grants SCH-1622679 and 1622542 (Arlington, VA), and unrestricted departmental funding and a Career Development Award (JPC) from Research to Prevent Blindness (New York, NY). The sponsor or funding organizations had no role in the design or conduct of this research.

Conflict of Interest: JKC was a consultant for Infotech Soft (Miami, FL). RVPC is a consultant for Alcon (Ft. Worth, TX), Novartis (Basel, Switzerland) and is on the advisory board for Phoenix technology (Pleasanton, CA). MFC is an unpaid member of the Scientific Advisory Board for Clarity Medical Systems (Pleasanton, CA), a consultant for Novartis (Basel, Switzerland), and an equity owner in Inteleretina, LLC (Honolulu, HI).

Presented at the Annual Meeting of the Association for Research in Vision and Ophthalmology, Vancouver, British Columbia, April 28-May 2, 2019.

View full text