A Step Forward in Using Artificial Intelligence to Identify Serious Retinopathy of Prematurity—A Start With a Long Road Ahead | Neonatology | JAMA Network Open | JAMA Network
[Skip to Navigation]
Sign In
Views 1,115
Citations 0
Invited Commentary
Ophthalmology
May 5, 2021

A Step Forward in Using Artificial Intelligence to Identify Serious Retinopathy of Prematurity—A Start With a Long Road Ahead

Author Affiliations
  • 1Division of Ophthalmology, The Children’s Hospital of Philadelphia, University of Pennsylvania, Philadelphia
JAMA Netw Open. 2021;4(5):e219245. doi:10.1001/jamanetworkopen.2021.9245

Elsewhere in JAMA Network Open, Wang et al1 have advanced our understanding of the utility of quantitative image analysis for premature infants at risk of blindness due to retinopathy of prematurity (ROP). Using a cloud-based platform in a stepwise approach, Wang et al1 selected parameters that are likely to be available in images using current systems; this approach creates new understanding of the ophthalmologic findings in ROP.

Among the most intriguing parameters selected for inclusion in these analyses is the presence of hemorrhage in the digital image, whether it is preretinal, retinal, intravitreal, or subretinal. Even during the development of the International Classification of ROP revisited in 2005,2 retinal hemorrhage was not included in the classification, although its presence was recognized as a worrisome indicator of serious disease. The novel binary approach to hemorrhage in eyes of at-risk premature infants that Wang et al1 chose is further supported by studies from Hutchinson et al3 and Daniel et al,4 and this choice allows for relatively easy categorization, even with dense vitreous haze.

By largely concentrating on what data can be obtained from digital images that include the optic disc using the RetCam imaging system (Clarity Medical Systems), Wang et al1 could determine whether the ROP occurred in zone I or II/III. They suggest that enough information can be obtained from a posterior retinal image to decide whether an ophthalmologist experienced in ROP should evaluate the child and determine the need for treatment.

Similarly, the stage of ROP appears to be less relevant than the actual presence of some degree of retinopathy in the image, and that is sufficient to enter into the calculation of risk. This observation is particularly trenchant, as the current indication for treatment of ROP often depends on distinguishing between whether ROP in zone II is stage 2 or stage 3 (ie, whether extraretinal neovascularization is present). Using optical coherence tomography, Maldonado and Toth5 have shown the distinction between the stages is subtle and likely dependent on technology.

Wang et al1 also provide a clear definition of the region of the retina vessels in which to make a quantitative assessment of vessel abnormality, again making the binary distinction between normal and preplus/plus disease. Indeed, as shown by eTable 2 and eFigure 5 in the Supplement, this distinction was made quite reliably, with very few false-negative results.

This quantitative approach has potential in several areas. First, by providing quantitative measures, outcomes of clinical investigations into the course of ROP and the treatment of serious disease can be more rigorously designed without the need for repeated examinations of infants. However, even with the approach presented by Wang et al1 and others likely to be developed, I am concerned that we have just begun to understand how to provide optimal ROP care for the increasing number of premature infants surviving worldwide. In 2010, Blencowe et al6 estimated that there were 20 000 infants blinded annually from ROP, with more than 10 000 additional infants with visual handicaps due to ROP. With an increasing number of premature infants surviving in regions of the world with rapidly developing neonatal intensive care systems, in addition to children with ever lower birth weights and gestational ages surviving in countries with well-established systems, there is an increasing need for cooperation between neonatologists, pediatricians, neonatal nurses, and ophthalmologists to develop adequate detection and treatment programs. As technology and imaging systems become less expensive, more widely available, capable of obtaining images of more peripheral retina, and with increased understanding of the pitfalls and advantages of these novel image analysis approaches, the approach described in the study by Wang et al1 begins to provide hope for improving care of our children with the most risk of blindness.

Back to top
Article Information

Published: May 5, 2021. doi:10.1001/jamanetworkopen.2021.9245

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Quinn GE. JAMA Network Open.

Corresponding Author: Graham E. Quinn, MD, MSCE, Division of Ophthalmology, The Children’s Hospital of Philadelphia, University of Pennsylvania, Main Building, 9th Floor, Philadelphia, PA 19104 (quinn@chop.edu).

Conflict of Interest Disclosures: None reported.

References
1.
Wang  J, Ji  J, Zhang  M,  et al.  Automated explainable multidimensional deep learning platform of retinal images for retinopathy of prematurity screening.   JAMA Netw Open. 2021;4(5):e218758. doi:10.1001/jamanetworkopen.2021.8758Google Scholar
2.
International Committee for the Classification of Retinopathy of Prematurity.  The International Classification of Retinopathy of Prematurity revisited.   Arch Ophthalmol. 2005;123(7):991-999. doi:10.1001/archopht.123.7.991PubMedGoogle ScholarCrossref
3.
Hutcheson  KA, Nguyen  AT, Preslan  MW, Ellish  NJ, Steidl  SM.  Vitreous hemorrhage in patients with high-risk retinopathy of prematurity.   Am J Ophthalmol. 2003;136(2):258-263. doi:10.1016/S0002-9394(03)00190-9PubMedGoogle ScholarCrossref
4.
Daniel  E, Ying  GS, Siatkowski  RM, Pan  W, Smith  E, Quinn  GE; e-ROP Cooperative Group.  Intraocular hemorrhages and retinopathy of prematurity in the telemedicine approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) Study.   Ophthalmology. 2017;124(3):374-381. doi:10.1016/j.ophtha.2016.10.040PubMedGoogle ScholarCrossref
5.
Maldonado  RS, Toth  CA.  Optical coherence tomography in retinopathy of prematurity: looking beyond the vessels.   Clin Perinatol. 2013;40(2):271-296. doi:10.1016/j.clp.2013.02.007PubMedGoogle ScholarCrossref
6.
Blencowe  H, Lawn  JE, Vazquez  T, Fielder  A, Gilbert  C.  Preterm-associated visual impairment and estimates of retinopathy of prematurity at regional and global levels for 2010.   Pediatr Res. 2013;74(suppl 1):35-49. doi:10.1038/pr.2013.205PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    ×