Secondary Logo

Journal Logo

Original Articles

Telemedicine and Pediatric Retinal Disease

Jeng-Miller, Karen W. MD, MPH; Yonekawa, Yoshihiro MD

Author Information
International Ophthalmology Clinics: Winter 2020 - Volume 60 - Issue 1 - p 47-56
doi: 10.1097/IIO.0000000000000297
  • Free

Introduction

Ophthalmic diagnostic technologies have revolutionized our understanding about the pathoanatomy and pathophysiology of many disease processes. This is especially applicable in the field of retina where fundus photography, optical coherence tomography, optical coherence tomography angiography, fluorescein angiography, and many other imaging modalities have allowed us to evaluate, document, and disseminate the findings and evolution of retinal diseases. Many have recognized the utility of ophthalmic imaging for use in telemedicine applications. Telemedicine is defined as the use of digital communication and electronic information sharing to provide health care at a distance.1

Telemedicine has the power to transform patient experiences for ophthalmic care, especially in a primary care setting. Patients normally have better access to general practitioners and, with teleophthalmology, can have the opportunity to receive screening eye care during their routine visits to determine the need for referral to subspecialty care.2 There are various studies demonstrating the efficacy of teleophthalmology in screening for various blinding eye conditions such as diabetic retinopathy,3,4 age-related macular degeneration,5 and glaucoma.6

One particularly useful application for telemedicine is in the field of pediatric retina. Pediatric retina remains a small field with limited specialists available to engage in the diagnosis, treatment, and management of pediatric retinal diseases. The advent of information sharing through various retinal imaging modalities allows for remote screening and guidance of potential treatment decisions for various pediatric retinal diseases. This paper will review the applications for telemedicine for pediatric retinal disease and future directions for care.

Retinopathy of Prematurity (ROP): Current Screening, Shortcomings and the Role of Telemedicine

ROP is a primary cause of blindness in premature infants and continues to have a perceptible burden on our health economy. ROP occurs due to an immature retinal blood supply in premature and low–birth weight infants, resulting in the ischemic peripheral retina. The ischemic retina induces production of vascular endothelial growth factors resulting in neovascularization, retinal detachments, and in some cases, blindness.

ROP is characterized with regard to zone, stage, and plus disease seen on clinical examination (most commonly bedside binocular indirect ophthalmoscopy) which helps guides treatment decisions.7 Treatment options include ablative therapy (cryotherapy and/or laser photocoagulation), intravitreal injections, and intraocular surgery. Randomized trials have demonstrated that early treatment, especially in high-risk prethreshold ROP, significantly reduced unfavorable outcomes.8 This demonstrates the importance of adequate screening and subsequent appropriate treatment of infants at risk.

As of 2012 in the United States, the incidence of ROP increased by 5.18% over the last 12 years.9 Furthermore, the frequency of ROP was calculated at 30.22% in newborns with a birth weight between 750 and 999 g.9 The number of premature infants requiring screening is likely increasing because of technological advancements in the neonatal intensive care unit, fortunately, allow for improved survival outcomes of younger infants.

ROP screenings with a full dilated binocular indirect ophthalmoscopic examination are recommended for infants with birth weights <1500 g or a gestational age of 30 weeks or below and infants with birth weights between 1500 and 2000 g or gestational age above 30 weeks with an unstable clinical course.10,11 Although the demand for screenings of these babies continues to rise, there is a discrepancy in the number of ophthalmologists available to perform these examinations. An American Academy of Ophthalmology survey in 2006 revealed that over half of pediatric and retinal ophthalmologists surveyed stated they would cease ROP care/screenings or planned to cease screenings and treatment of ROP in the near future.12 Reasons among these ophthalmologists for doing so included high medicolegal liability, high costs of insurance, poor reimbursement, and complexity of care for these children.12 Furthermore, in cases of ROP, 78% of neonatal intensive care units that do not offer treatment indicate the main reasoning to be a lack of ophthalmologic services.13 With a persistent and increasing demand for ROP screening and a decline in the number of physicians available in the vicinity of these patients, telemedicine becomes an attractive model to provide care to some of the most vulnerable populations.

Various studies have demonstrated the efficacy and safety of photographic screening for ROP.14–22 One of the earliest studies evaluating telemedicine in ROP screening was the photographic screening for retinopathy of prematurity study (PHOTO-ROP).23 Over a year of study, infants were screened first by remote digital fundus imaging followed by indirect binocular ophthalmoscopy. Study-defined clinically significant ROP was detected with 92% sensitivity and 37.21% specificity when compared with traditional ophthalmoscopy; in patients with ROP requiring early treatment of ROP, image analysis demonstrated 92% sensitivity and 67.39% specificity.23

Importantly, the telemedicine model has also been proven to identify referral-warranted ROP by trained nonphysician readers, thus potentially providing a larger safety net for these at-risk infants.24 In this study, trained imaging technicians photographed retinal images, which were evaluated by nonphysician readers to identify referral-warranted ROP (zone 1 ROP, stage 3 ROP or worse, and/or plus disease). This was compared against bedside binocular indirect ophthalmoscopy evaluations from ophthalmologists with expertise in ROP. Results demonstrated that trained nonphysician readers can accurately identify referral-warranted ROP.24 Although trained ROP image readers can reduce the burden of in-person examinations, ophthalmologists are still needed to be available for adjudication in discrepant findings and treatment when necessary. Furthermore, readers will require quality assurance protocols and stringent training standards, requiring time, cost, and ongoing supervision. Although this concept is an intriguing alternative in the telemedicine realm, the cost-effectiveness and efficacy require further evaluation.

As a result of the robust data for the efficacy of telemedicine ROP screening, several enacted telemedicine programs have formed and demonstrated promising results. A 5-year study by researchers at Stanford found that no cases of treatment-warranted ROP were missed on telemedicine screening; the screening was 100% sensitive, 99.8% specific, and had a 100% negative predictive value for detecting these cases.25 A recent 2018 study conducted in the United States and Mexico found no significant difference in diagnostic accuracy for the detection of clinically significant ROP when comparing traditional binocular indirect ophthalmoscopy versus remote image grading of wide-angle fundus photographs.26 Although traditional binocular indirect ophthalmoscopy had a slightly higher accuracy when diagnosing zone II and stage 3 ROP, the model continues to validate the use of telemedicine for diagnosis of clinically significant ROP.26

The telemedicine model has also had success overseas in varying resource-available countries such as Germany, France, Chile, and India.27–30 The multicenter study in Germany spanned 6 years and found that wide-field digital image screening was 100% sensitive in capturing suspected treatment-requiring ROP.27 A 2017 study in Hungary demonstrated digital image screening for ROP to have a high diagnostic performance and also provided a return on initial investment within 5 years of project initiation.31

In the context of ROP treatment, years of findings have demonstrated that telemedicine is an effective modality for ROP screening. The 2013 screening guidelines for ROP acknowledge the utility of digital image screening for ROP and the use of this practice as an alternative means for ROP screening.10 As a result, there is mounting evidence that telemedicine can address important gaps in clinical coverage without significantly sacrificing quality care.

Potential Role of Telemedicine in Pediatric Retinal Examinations and Diseases

The role of telemedicine in pediatric retinal disease has been most widely examined in the ROP realm. However, little work has been done regarding baseline pediatric fundus screening examinations or applications for telemedicine in other pediatric retinal disorders. Digital fundus imaging has great potential for information sharing in other pediatric retinal disorders and degenerations, such as familial exudative vitreoretinopathy, Coats’ disease, incontinentia pigmenti, among others, in which early management and treatment can be important for vision preservation. Telemedicine can serve as a consulting tool among general practitioners and even general ophthalmologists with experts to help guide in the diagnosis, urgency of referral, and need for treatment in these conditions.

Recently, a 2016 study piloted at Stanford University School of Medicine conducted a year-long wide-angle digital retinal photography screening of all newborns born at a Lucile Packard Children’s Hospital.32 This study named the Newborn Eye Screen Test (NEST) study, examined the birth prevalence of fundus hemorrhages in all babies born at age 37 weeks gestation or above. The study used the RetCam III camera (Natus, Pleasanton, CA) to obtain photographs and found that fundus hemorrhages are fairly common among newborns, with vaginal delivery having a higher associated risk of fundus hemorrhages. As dilated funduscopic examinations are rarely conducted in non at-risk newborns, it is unclear of the visual significance of these fundus hemorrhages. However, with increasing ease of digital image capture for newborn and pediatric patients, more insight can be gathered on retinovascular development and pathophysiology for future management and treatment developments of diseases.

Advantages of Telemedicine in Pediatric Retinal Disease

There are many advantages to employing telemedicine in screening pediatric retinal disease. The most obvious advantage is overcoming geographical challenges and providing improved access to health care for these patients. This is especially important in more rural areas or resource-poor areas/countries in which specialists needed for screening and evaluations are often not accessible.

Telemedicine also can be a cost-effective solution for communication and treatment coordination between local physicians and expert consultants.33–35 A US study modeled 2 scenarios for ROP screening: (1) standard ophthalmic examinations by an ophthalmologist versus (2) digital fundus photography taken by nonophthalmic personnel using a wide-angle imaging device and interpreted by a remote ophthalmologist.33 The study demonstrated that the costs per quality-adjusted life-year gained were $3193 with telemedicine versus $5617 with standard ophthalmoscopy examination. Further sensitivity analyses demonstrated that telemedicine had improved cost-effectiveness over standard ophthalmoscopy over various parameters, such as accuracy, incidence of treatment-requiring ROP, and percentage of interpretable telemedicine images.33 A separate study conducted in the United Kingdom found that models of telemedicine via image capture and grading by visiting nurses and telemedicine via image capture by nurses and image grading by remote ophthalmologists were more cost-effective strategies than traditional bedside ophthalmoscopy.34 Finally, telemedicine examinations require less physician time in comparison to traditional bedside ophthalmoscopy due to many factors such as patient cooperation/positioning, coordination with neonatal intensive care unit schedules, and travel time. Richter et al36 found that mean times for ophthalmoscopic diagnosis of ROP ranged from 4 to over 6 minutes per infant whereas telemedicine diagnosis time ranged from 1 to 1.75 minutes per infant.

Many arguments in favor of telemedicine for pediatric retinal diseases center on benefits such as health care accessibility, cost savings, and time effectiveness. However, there are other societal costs and benefits that are more difficult to quantify that should be considered when advocating for telemedicine programs. From the physician perspective, telemedicine can provide subjectively greater provider satisfaction due to time savings and efficiency of diagnosis. Furthermore, the repository of images gained from telemedicine can be important in information sharing for second opinions, peer education and research.37 Digital fundus photographs are also important in the education of health care staff and families. For physicians in more rural areas, the repository of images can be used for ongoing clinical education and potential improvements in ROP awareness and screening.38 Fundus photographs can also be used for family education to demonstrate pictorially the disease patterns in ROP and clinical improvements with continued monitoring and treatment.38 Finally, fundus photography serves as clear documentation along the clinical course of these neonates. This can be especially important in the medicolegal aspects of ROP care. In contrast to popular belief, malpractice claims in the ROP realm are comparably rare but, when present, are often very costly.39 Fundus photo documentation can serve as an objective way to demonstrate proper care and judgment in cases should doubt arise.

Limitations of Telemedicine for Pediatric Retinal Diseases

Telemedicine in pediatric retinal diseases relies heavily on image quality. Obtaining quality images can be difficult in neonates due to various factors: infants can have darkly pigmented fundi, small palpebral fissures, media opacities, or motion artifacts.38 Furthermore, quality can vary based on camera operator experience and education. In various published studies, noninterpretable images can range from 8% to 21%.15,16 Another factor in image analysis is the ability to examine all necessary quadrants for an accurate diagnosis to be made. A study by Lajoie et al14 demonstrated that single image telemedicine examinations are comparable to multiple-image telemedicine examinations when detecting and determining the follow-up interval of plus disease. However, most agree that the detection of ROP relies on a full set of imaging.40 It is often easiest to obtain images of the posterior pole and superior retina (due to Bell reflex) but often more difficult to obtain inferior retinal images due to eye and speculum positioning.40

Other prohibitive factors for telemedicine implementation include the cost of a wide-field imaging camera and the medicolegal implications of telemedicine. A pediatric wide-field imaging camera can be expensive and is often difficult to incorporate into the budget of smaller institutions. A 2008 ROP workforce analysis found that only 18% of ophthalmologists who provide care for children below 1 year of age have a retinal imaging device.41 Although most reported that the reason for a lack of a retinal imaging device was that there was no need for one, a minority did cite that the reimbursement was inadequate to offset the cost of the device.41

Another consideration is the medicolegal liability of telemedicine in ROP screening. Telemedicine is relatively new in clinical practice and laws and legislation surrounding it are not robustly defined.38 Furthermore, the general consensus among ophthalmologists continues to support traditional indirect ophthalmoscopy, with only a minority believing that telemedicine was a safe practice for ROP screening.13 Last, telemedicine can be potentially dissatisfying for family members who prefer a face to face discussion with an ophthalmologist regarding their child’s care and result in confusion regarding diagnosis and necessity of treatment.37 These factors warrant further study in the ongoing debate of telemedicine.

Future Directions and Conclusions

Telemedicine in the realm of pediatric retinal diseases has many benefits. However, limitations for implementation include proper workforce and support training and technological shortcomings. However, there are many ongoing technological developments that can help bolster the argument for future widespread use of telemedicine. The cost of pediatric retinal cameras is a high deterrent for telemedicine implementation. There are many investigations into inexpensive fundus camera alternatives that can potentially alleviate these costs.42–44 Although the timeline for the clinical implementation of these devices, especially in pediatric clinical evaluation, is unclear, inexpensive imaging devices is an important avenue for the development of cost-effective and widespread telemedicine programs.

Image learning algorithms and artificial intelligence can also aid in screening and analyses for ROP.45–49 A 2016 study reported on the results of ROPtool, an image analysis tool, which evaluates for vascular abnormalities and tortuosity in fundus photographs. The ROPtool sensitivity was 91% and had a specificity of 82%.50 Deep learning is a recent machine learning technique using multiple processing layers to learn a representation of data and has improved the art of speech recognition, visual object recognition, and object detection.51 A 2018 study by Brown et al52 demonstrated that deep learning technology could diagnose plus disease in ROP with proficiency comparable to human experts, achieving 91% accuracy and out-performing 6 of 8 ROP experts. With improvements in image acquisition and image learning, telemedicine will continue to become a more validated and attractive model for screening in pediatric retinal disease.

In conclusion, telemedicine for pediatric retinal disease is an important paradigm in clinical practice that should continue to be investigated and implemented. Further developments can refine the implementation and delivery of care in telemedicine models.

References

1. Field MJ, Grigsby J. Telemedicine and remote patient monitoring. JAMA. 2002;288:423–425.
2. Rathi S, Tsui E, Mehta N, et al. The current state of teleophthalmology in the United States. Ophthalmology. 2017;124:1729–1734.
3. Kirkizlar E, Serban N, Sisson JA, et al. Evaluation of telemedicine for screening of diabetic retinopathy in the Veterans Health Administration. Ophthalmology. 2013;120:2604–2610.
4. Zimmer-Galler IE, Kimura AE, Gupta S. Diabetic retinopathy screening and the use of telemedicine. Curr Opin Ophthalmol. 2015;26:167–172.
5. Li B, Powell A-M, Hooper PL, et al. Prospective evaluation of teleophthalmology in screening and recurrence monitoring of neovascular age-related macular degeneration: a randomized clinical trial. JAMA Ophthalmol. 2015;133:276–282.
6. Thomas S-M, Jeyaraman MM, Jeyaraman M, et al. The effectiveness of teleglaucoma versus in-patient examination for glaucoma screening: a systematic review and meta-analysis. PLoS One. 2014;9:e113779.
7. International Committee for the Classification of Retinopathy of Prematurity. The International Classification of Retinopathy of Prematurity revisited. Arch Ophthalmol. 2005;123:991–999.
8. Early Treatment for Retinopathy of Prematurity Cooperative Group. Revised indications for the treatment of retinopathy of prematurity: results of the early treatment for retinopathy of prematurity randomized trial. Arch Ophthalmol. 2003;121:1684–1694.
9. Ludwig CA, Chen TA, Hernandez-Boussard T, et al. The epidemiology of retinopathy of prematurity in the United States. Ophthalmic Surg Lasers Imaging Retina. 2017;48:553–562.
10. Fierson WM. American Academy of Pediatrics Section on Ophthalmology; American Academy of Ophthalmology; American Association for Pediatric Ophthalmology and Strabismus; American Association of Certified Orthoptists. Screening examination of premature infants for retinopathy of prematurity. Pediatrics. 2013;131:189–195.
11. Pathipati AS, Moshfeghi DM. Telemedicine applications in pediatric retinal disease. J Clin Med. 2017;6:36.
12. Altersitz K, Piechocki M. Survey: Physicians being driven away from ROP treatment. Ocul Surg News US Ed; 2006. Available at: www.healio.com/ophthalmology/retina-vitreous/news/print/ocular-surgery-news/%7Bedfa784c-a2ac-4f93-b473-5be5d07d82ca%7D/survey-physicians-being-driven-away-from-rop-treatment. Accessed December 15, 2018.
13. Vartanian RJ, Besirli CG, Barks JD, et al. Trends in the screening and treatment of retinopathy of prematurity. Pediatrics. 2017;139:e20161978.
14. Lajoie A, Koreen S, Wang L, et al. Retinopathy of prematurity management using single-image vs. multiple-image telemedicine examinations. Am J Ophthalmol. 2008;146:298–309.
15. Wu C, Petersen RA, VanderVeen DK. RetCam imaging for retinopathy of prematurity screening. J AAPOS. 2006;10:107–111.
16. Photographic Screening for Retinopathy of Prematurity (Photo-ROP) Cooperative Group, Balasubramanian M, Capone A Jr, Hartnett ME, et al. The Photographic Screening for Retinopathy of Prematurity Study (Photo-ROP): study design and baseline characteristics of enrolled patients. Retina. 2006;26:S4–S10.
17. Chiang MF, Keenan JD, Starren J, et al. Accuracy and reliability of remote retinopathy of prematurity diagnosis. Arch Ophthalmol. 2006;124:322–327.
18. Roth DB, Morales D, Feuer WJ, et al. Screening for retinopathy of prematurity employing the retcam 120: sensitivity and specificity. Arch Ophthalmol. 2001;119:268–272.
19. Ells AL, Holmes JM, Astle WF, et al. Telemedicine approach to screening for severe retinopathy of prematurity: a pilot study. Ophthalmology. 2003;110:2113–2117.
20. Schwartz SD, Harrison SA, Ferrone PJ, et al. Telemedical evaluation and management of retinopathy of prematurity using a fiberoptic digital fundus camera. Ophthalmology. 2000;107:25–28.
21. Chiang MF, Wang L, Busuioc M, et al. Telemedical retinopathy of prematurity diagnosis: accuracy, reliability, and image quality. Arch Ophthalmol. 2007;125:1531–1538.
22. Chiang MF, Starren J, Du YE, et al. Remote image based retinopathy of prematurity diagnosis: a receiver operating characteristic analysis of accuracy. Br J Ophthalmol. 2006;90:1292–1296.
23. Photographic Screening for Retinopathy of Prematurity (Photo-ROP) Cooperative Group. The photographic screening for retinopathy of prematurity study (photo-ROP). Primary outcomes. Retina. 2008;28:S47–S54.
24. Quinn GE, Ying G, Daniel E, et al. Validity of a telemedicine system for the evaluation of acute-phase retinopathy of prematurity. JAMA Ophthalmol. 2014;132:1178–1184.
25. Fijalkowski N, Zheng LL, Henderson MT, et al. Stanford University Network for Diagnosis of Retinopathy of Prematurity (SUNDROP): five years of screening with telemedicine. Ophthalmic Surg Lasers Imaging Retina. 2014;45:106–113.
26. Biten H, Redd TK, Moleta C, et al. Diagnostic accuracy of ophthalmoscopy vs telemedicine in examinations for retinopathy of prematurity. JAMA Ophthalmol. 2018;136:498–504.
27. Lorenz B, Spasovska K, Elflein H, et al. Wide-field digital imaging based telemedicine for screening for acute retinopathy of prematurity (ROP). Six-year results of a multicentre field study. Graefes Arch Clin Exp Ophthalmol. 2009;247:1251–1262.
28. Vinekar A, Mangalesh S, Jayadev C, et al. Impact of expansion of telemedicine screening for retinopathy of prematurity in India. Indian J Ophthalmol. 2017;65:390–395.
29. Ossandón D, Zanolli M, Stevenson R, et al. A national telemedicine network for retinopathy of prematurity screening. J AAPOS. 2018;22:124–127.
30. Chan H, Cougnard-Grégoire A, Korobelnik JF, et al. Screening for retinopathy of prematurity by telemedicine in a tertiary level neonatal intensive care unit in France: review of a six-year period. J Fr Ophtalmol. 2018;41:926–932.
31. Kovács G, Somogyvári Z, Maka E, et al. Bedside ROP screening and telemedicine interpretation integrated to a neonatal transport system: economic aspects and return on investment analysis. Early Hum Dev. 2017;106–107:1–5.
32. Callaway NF, Ludwig CA, Blumenkranz MS, et al. Retinal and optic nerve hemorrhages in the newborn infant: one-year results of the newborn eye screen test study. Ophthalmology. 2016;123:1043–1052.
33. Jackson KM, Scott KE, Graff Zivin J, et al. Cost-utility analysis of telemedicine and ophthalmoscopy for retinopathy of prematurity management. Arch Ophthalmol. 2008;126:493–499.
34. Castillo-Riquelme MC, Lord J, Moseley MJ, et al. Cost-effectiveness of digital photographic screening for retinopathy of prematurity in the United Kingdom. Int J Technol Assess Health Care. 2004;20:201–213.
35. Isaac M, Isaranuwatchai W, Tehrani N. Cost analysis of remote telemedicine screening for retinopathy of prematurity. Can J Ophthalmol. 2018;53:162–167.
36. Richter GM, Sun G, Lee TC, et al. Speed of telemedicine vs ophthalmoscopy for retinopathy of prematurity diagnosis. Am J Ophthalmol. 2009;148:136.e1–142.e2.
37. Richter GM, Williams SL, Starren J, et al. Telemedicine for retinopathy of prematurity diagnosis: evaluation and challenges. Surv Ophthalmol. 2009;54:671–685.
38. Thanos A, Yonekawa Y, Todorich B, et al. Screening and treatments using telemedicine in retinopathy of prematurity. Eye Brain. 2016;8:147–151.
39. Day S, Menke AM, Abbott RL. Retinopathy of prematurity malpractice claims: the Ophthalmic Mutual Insurance Company experience. Arch Ophthalmol. 2009;127:794–798.
40. Morrison D, Bothun ED, Ying G-S, et al. Impact of number and quality of retinal images in a telemedicine screening program for ROP: results from the e-ROP study. J AAPOS. 2016;20:481–485.
41. Kemper AR, Freedman SF, Wallace DK. Retinopathy of prematurity care: patterns of care and workforce analysis. J AAPOS. 2008;12:344–348.
42. Omer MT, Abbas E. OptiCard: an inexpensive and portable method of bedside direct fundoscopy. J Coll Physicians Surg Pak. 2017;27:719–721.
43. Russo A, Morescalchi F, Costagliola C, et al. A novel device to exploit the smartphone camera for fundus photography. J Ophthalmol. 2015;2015:823139.
44. Sharma A, Subramaniam SD, Ramachandran KI, et al. Smartphone-based fundus camera device (MII Ret Cam) and technique with ability to image peripheral retina. Eur J Ophthalmol. 2016;26:142–144.
45. Gelman R, Martinez-Perez ME, Vanderveen DK, et al. Diagnosis of plus disease in retinopathy of prematurity using Retinal Image multiScale Analysis. Invest Ophthalmol Vis Sci. 2005;46:4734–4738.
46. Woo R, Chan RV, Vinekar A, et al. Aggressive posterior retinopathy of prematurity: a pilot study of quantitative analysis of vascular features. Graefes Arch Clin Exp Ophthalmol. 2015;253:181–187.
47. Wittenberg LA, Jonsson NJ, Chan RV, et al. Computer-based image analysis for plus disease diagnosis in retinopathy of prematurity. J Pediatr Ophthalmol Strabismus. 2012;49:11–19; quiz 10, 20.
48. Kemp PS, VanderVeen DK. Computer-assisted digital image analysis of plus disease in retinopathy of prematurity. Semin Ophthalmol. 2016;31:159–162.
49. Ting DSW, Wu W-C, Toth C. Deep learning for retinopathy of prematurity screening. Br J Ophthalmol. 2019;103:577–579.
50. Abbey AM, Besirli CG, Musch DC, et al. Evaluation of screening for retinopathy of prematurity by ROPtool or a lay reader. Ophthalmology. 2016;123:385–390.
51. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521:436–444.
52. Brown JM, Campbell JP, Beers A, et al. Automated diagnosis of plus disease in retinopathy of prematurity using deep convolutional neural networks. JAMA Ophthalmol. 2018;136:803–810.
Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.