Leukocoria is an abnormal pupillary light reflection characterized by a white-eye reflex that usually results from an intraocular abnormality.1 Common causes of leukocoria in children are retinoblastoma, persistent hyperplastic primary vitreous, retinopathy of prematurity, Coats' disease, toxocariasis, congenital cataract, phakomatoses, Norrie disease, and retinal dysplasia and detachment.2 An asymmetric red reflex is frequently associated with less urgent conditions, such as anisometropia and strabismus, and may even be normal. Conversely, true leukocoria mandates prompt referral to an ophthalmologist because its causes may threaten vision and/or life. Leukocoria is the presenting sign in approximately 50% of patients with retinoblastoma.3 Leukocoria can be detected by pediatricians and ophthalmologists during screening visits by eliciting the red reflex, a non-invasive and child-friendly screening tool.4 However, parents or relatives often notice a white pupil on flash photography5 and this is the first reason for consultation.6 In addition, eye-related smartphone applications (apps) are currently available. In particular, the CRADLE (Computer Assisted Detection of Leukocoria) smartphone app for an iPhone operating system (iOS) (iPhone 7; Apple, Cupertino, CA) was developed to increase early diagnosis of retinoblastoma by detecting leukocoria.7
The aim of this study was to determine whether a white-eye detector smartphone app (CRADLE) can be used as a screening tool for the early detection of leukocoria in a clinical practice.
Patients and Methods
For this prospective single-visit study, children aged 1 to 6 years presenting to the University Eye Clinic of Genova, IRCCS Ospedale Policlinico San Martino, and Queen Sirikit National Institute of Child Health for a complete pediatric ophthalmological examination were enrolled. The study was conducted in accordance with the tenets of the Declaration of Helsinki and was approved by the local institutional review board. Exclusion criteria were any conditions affecting the ability to obtain an adequate image. None of the children had a preexisting diagnosis of any condition associated with leukocoria.
All children who met the enrollment criteria were screened using the CRADLE smartphone app on the iPhone 7 (Apple) by an orthoptist who was masked to the reason for referral for the ophthalmologic examination. The app was developed to detect the eyes on a face in a patient facing the camera with the face fully visible, providing a quadrate red or green colored template at the correct focal distance around each eye. The red template is associated with the legend “white,” indicating a leukokoric eye, whereas the green template is associated with the legend “normal,” indicating a normal eye. Following photoscreening, a comprehensive ophthalmic examination was performed according to the American Association for Pediatric Ophthalmology and Strabismus guidelines.8 Cycloplegic retinoscopy and fundus examination were performed 30 minutes after one to two drops of a pediatric combination drop, comprising tropicamide 1% and phenylephrine 2.5%, was instilled. The two pediatric ophthalmologists (AV, NW) who participated in the study were masked to photoscreener results.
Continuous variables are presented as means and standard deviations, whereas categorical variables are presented as the number and/or percentage of patients. A comparison of the two methods yielded sensitivity, specificity, and negative likelihood ratio values.
In total, 244 eyes of 122 children (64 [52%] male and 58 [48%] female) were enrolled for evaluation in the current study. The majority of patients were white (67%), followed by Asian (16%) and Hispanic (17%). The mean age was 19.7 ± 26.4 months (range: 2 to 6 months). Nine eyes of 244 (3.6%) patients had amblyogenic cataract, 1 (0.4%) patient had stage 5 retinopathy of prematurity, and 3 (1.2%) patients had retinoblastoma at the time of the eye examination. Leukocoria was evaluated in all of these eye conditions by penlight or ophthalmoscope. The smartphone app was able to detect only two leukokoric eyes of the same patient caused by bilateral retinoblastoma. None of the 9 eyes with amblyogenic cataract were detected by the smart-phone app (Figures 1–2). Therefore, there were 11 false-negative and 0 false-positive results. The sensitivity of the white-eye detector app was 15.38% (95% confidence interval [CI]: 1.92% to 45.45%), whereas the specificity was 100% (95% CI: 98.48% to 100.00%). The negative likelihood ratio was 0.85 (95% CI: 0.67 to 1.07).
A 2-year-old child with bilateral amblyogenic cataract who was evaluated as “normal” by the smartphone photoscreening application.
A 4-year-old child with bilateral amblyogenic cataract under general anesthesia who was evaluated as “normal” by the smartphone photoscreening application.
In the red reflex screening test, a direct ophthalmoscope is used to compare the pupillary light reflex between the eyes to identify pathologies such as cataracts and intraocular tumors in preverbal children.9 The American Academy of Pediatrics10 recommends that all neonates, infants, and children undergo a red reflex ophthalmic examination before discharge from the neonatal nursery and at all subsequent routine health supervision visits. Leukocoria is an abnormal pupillary light reflection that usually results from an intraocular abnormality. Approximately half of the cases of childhood leukocoria are caused by retinoblastoma.11 Retinoblastoma is the most common intraocular tumor of childhood and can be life threatening. For these reasons, several analyses to detect retinoblastoma evaluating leukocoria by amateur photography have been developed.12,13 The CRADLE app was developed by a Baylor University chemistry professor whose son's life-threatening case of retinoblastoma was not diagnosed until he was 3 months old, even though photographs taken of him at age 12 days showed a white pupil. The CRADLE app converts a smartphone into a computer-assisted ophthalmoscope after activating the smartphone's LED and video camera. Signs of leukocoria are shown with red boxes around the examined eye. The CRADLE app has been considered as an ophthalmoscope alternative in underdeveloped countries.7 The researcher who developed this app stated that the purpose of the app is not to diagnose disease and it does not replace an eye visit with a physician.
The CRADLE app produces algorithms that can learn from examples to automatically detect when a photograph or video contains a white eye. Although the specificity of this app was 100%, we observed that the sensitivity was only 15.38%. The app was not able to recognize 11 eyes with leukocoria that were evaluable by penlight or ophthalmoscope. The reason for this was the lack of similar sample images in the experimental database. Additionally, cataract may produce a darker rather than a whiter fundus reflex.
A previous study evaluated the diagnostic performance of the CRADLE app in 23 patients with retinoblastoma and 4 controls. The app failed to detect leukocoria, with the exception of 4 eyes with late stage retinoblastoma.14 These results are consistent with those of our study in a larger cohort of patients. However, the detection of leukocoria with conventional red reflex testing suffers from a high degree of false-negative results.15 In particular, small and peripheral retinal lesions are harder to detect and may yield a completely normal red reflex on straight-on examination.16,17 Although oblique viewing may improve the detection rate,17 the technique is not feasible with the CRADLE app, which may reduce its sensitivity and limit its clinical utility.
The 100% specificity and absence of false-positive results obtained with the CRADLE app was an unexpected result. Photoscreening methods typically have a lower specificity, but leukocoria may be visable in off-axis photographs of healthy eyes due to the reflection of the optic disc.18 Although high specificity may reduce over-referral, it can also result in missing at-risk children due to low sensitivity.19 The imbalance in sensitivity and specificity constitutes the greatest limitation of the CRADLE app.
The worldwide diffusion of smartphone technology might have the potential to reduce barriers to accessibility of vision screening in children.20 GoCheckKids is a photoscreening app registered with the U.S. Food and Drug Administration that is designed to detect amblyopia in children. A recent study reported an overall sensitivity of 76% and specificity of 85% in detecting amblyopia risk factors in children aged 1 to 6 years.21 Using smart-phone photoscreening apps to detect leukocoria may be valuable support for children's parents and relatives, unless they cause decreased vigilance in the case of a normal result. However, we believe that the CRADLE app cannot be considered an alterative to the ophthalmoscope for children aged 1 to 6 years. Further improvements in the app, such as an expansion of the pathological samples database, are required before it can be recommended as an effective screening tool. Therefore, pediatricians and volunteers on community health projects in under-developed countries should be trained to perform the red reflex eye test. Further studies with larger simple sizes are needed to confirm the role of this tool in the detection of leukocoria.
- Damasco VC, Dire DJ. A child with leukocoria. Pediatr Emerg Care. 2011;27:1170–1174. doi:10.1097/PEC.0b013e31823b0316 [CrossRef]
- Kembhavi SA, Sable N, Vora T, Arora B. Leukokoria: all that's white is not retinoblastoma. J Clin Oncol. 2011;29:e586–e587. doi:10.1200/JCO.2011.34.8334 [CrossRef]
- Abramson DH, Frank CM, Susman M, Whalen MP, Dunkel IJ, Boyd NW. Presenting signs of retinoblastoma. J Pediatr. 1998;132:505–508. doi:10.1016/S0022-3476(98)70028-9 [CrossRef]
- McLaughlin C, Levin AV. The red reflex. Pediatr Emerg Care. 2006;22:137–140. doi:10.1097/01.pec.0000199567.87134.81 [CrossRef]
- Christian LW, Bobier WR. Case 2: a three-year-old boy with photographic leukocoria. Paediatr Child Health. 2015;20:345–346. doi:10.1093/pch/20.7.345a [CrossRef]
- Asensio-Sánchez VM, Díaz-Cabanas L, Martín-Prieto A. Photoleukocoria with smartphone photographs. Int Med Case Rep J. 2018;11:117–119. doi:10.2147/IMCRJ.S163735 [CrossRef]
- Gray S. UMMS student testing smartphone app to detect childhood cancer in Guatemala. UMass Medical School Communications. June16, 2016. https://www.umassmed.edu/news/news-archives/2016/06/umms-student-testing-smartphone-app-to-detect-childhood-cancer-in-guatemala
- Donahue SP, Arthur B, Neely DE, Arnold RW, Silbert D, Ruben JB.AAPOS Vision Screening Committee. Guidelines for automated preschool vision screening: a 10-year, evidence-based update. J AAPOS. 2013;17:4–8. doi:10.1016/j.jaapos.2012.09.012 [CrossRef]
- Balmer A, Munier F. Differential diagnosis of leukocoria and strabismus, first presenting signs of retinoblastoma. Clin Ophthalmol. 2007;1:431–439.
- American Academy of PediatricsSection on OphthalmologyAmerican Association for Pediatric Ophthalmology and StrabismusAmerican Academy of OphthalmologyAmerican Association of Certified Orthoptists. Red reflex examination in neonates, infants, and children. Pediatrics. 2008;122:1401–1404. doi:10.1542/peds.2008-2624 [CrossRef]
- Smirniotopoulos JG, Bargallo N, Mafee MF. Differential diagnosis of leukokoria: radiologic-pathologic correlation. Radiographics. 1994;14:1059–1079. doi:10.1148/radiographics.14.5.7991814 [CrossRef]
- Abdolvahabi A, Taylor BW, Holden RL, et al. Colorimetric and longitudinal analysis of leukocoria in recreational photographs of children with retinoblastoma. PLoS One. 2013;8:e76677. doi:10.1371/journal.pone.0076677 [CrossRef]
- Rivas-Perea P, Baker E, Hamerly G, Shaw BF. Detection of leukocoria using a soft fusion of expert classifiers under non-clinical settings. BMC Ophthalmol. 2014;14:110. doi:10.1186/1471-2415-14-110 [CrossRef]
- Khedekar A, Devarajan B, Ramasamy K, Muthukkaruppan V, Kim U. Smartphone-based application improves the detection of retinoblastoma [published online ahead of print January 11, 2019]. Eye (Lond). doi:10.1038/s41433-018-0333-7 [CrossRef].
- Sun M, Ma A, Li F, et al. Sensitivity and specificity of red reflex test in newborn eye screening. J Pediatr. 2016;179:192–196.e4. doi:10.1016/j.jpeds.2016.08.048 [CrossRef]
- Khan AO, Al-Mesfer S. Lack of efficacy of dilated screening for retinoblastoma. J Pediatr Ophthalmol Strabismus. 2005;42:205–210.
- Li J, Coats DK, Fung D, Smith EO, Paysse E. The detection of simulated retinoblastoma by using red-reflex testing. Pediatrics. 2010;126:e202–e207. doi:10.1542/peds.2009-0882 [CrossRef]
- Marshall J, Gole GA. Unilateral leukocoria in off axis flash photographs of normal eyes. Am J Ophthalmol. 2003;135:709–711. doi:10.1016/S0002-9394(02)02079-2 [CrossRef]
- Wallace DK, Morse CL, Melia M, et al. Pediatric Eye Evaluations Preferred Practice Pattern: I. Vision screening in the primary care and community setting; II. Comprehensive ophthalmic examination. Ophthalmology. 2018;125:P184–P187. doi:10.1016/j.ophtha.2017.09.032 [CrossRef]
- Arnold RW, Arnold AW, Hunt-Smith TT, Grendahl RL, Winkle RK. The positive predictive value of smartphone photoscreening in pediatric practices. J Pediatr Ophthalmol Strabismus. 2018;55:393–396. doi:10.3928/01913913-20180710-01 [CrossRef]
- Arnold RW, O'Neil JW, Cooper KL, Silbert DI, Donahue SP. Evaluation of a smartphone photoscreening app to detect refractive amblyopia risk factors in children aged 1–6 years. Clin Ophthalmol. 2018;12:1533–1537. doi:10.2147/OPTH.S171935 [CrossRef]