Background. The reflection of visible light off choroidal and retinal blood vessels causes the human pupil to appear red when examined by a direct ophthalmoscope or photographed with a camera flash. Testing children for this red pupillary reflex (red reflex testing) is a required part of a regular newborn and pediatric check-up. A white or yellow-orange reflex (i.e., leukocoria) can be a sign of childhood eye disorders that threaten vision and even life in the case of retinoblastoma, a malignancy of the retina that accounts for 6% of cancers in children under five years of age. Leukocoria is also observed in snap-shot photographs often retrospectively after clinical diagnosis of eye disease. Although red reflex testing is part of routine examination by the pediatrician, its effectiveness when performed as part of a general physical examination has been limited. A more effective way to detect leukocoria early in its development will lead to better outcomes from treatment. Digital photography, including that performed by parents, represents an alternative method for detection of leukocoria. The clinical utility of digital analysis of photographs for leukocoria is unreported. Methods. A smartphone application, CRADLE (ComputeR Assisted Detector of LEukocoria), was engineered to detect leukocoria in common digital photographs and was made freely
2 available to the public in 2014 under the name “White Eye Detector”. We trained a convolutional neural network to detect leukocoria in common photographs collected by parents, embedded the network into a smartphone application, and analyzed its effectiveness in early detection of photographic leukocoria and its properties as a screening tool for parents or clinicians. We determined the sensitivity, specificity, and accuracy of CRADLE in detecting pathologic leukocoria among 52,982 photographs (donated by parents) of 20 children with previously diagnosed eye diseases and of 20 children without eye disease. Results. Retrospectively, 18 children with eye disease exhibited photographic leukocoria before clinical diagnosis. For 16 of these children, CRADLE detected leukocoria in photographs taken before diagnosis by a mean of 1.32 years (95% confidence interval [CI], 0.45 to 2.20). The sensitivity, specificity, and accuracy of CRADLE for detecting pathologic leukocoria in photographs taken during the first year of life were 90%, 20%, and 55%. Conclusion. An application which analyzes digital photographs on a smartphone is a sensitive screening tool for eye disease in infants and young children and may result in earlier detection of leukocoria and better clinical outcomes. Shizuo Mukai, M.D. was born in Tokyo, Japan and grew up in Tokyo, London (England), Montreal (Canada), and Boston (USA). He attended Brown University where he studied sculpture and photography and received his degree (Bachelor of Art) in studio art in 1979. He also studied biology and received his degree (Bachelor of Science) in it in 1979. During 1978, he spent eight months studying Japanese language, theater, and history at Jochi Daigaku / Sophia University in Tokyo. He returned to Brown University as a medical student and a graduate student in biology and received his M.D. in 1982. He completed his medical internship at Mount Auburn Hospital in Cambridge, Massachusetts followed by a postdoctoral fellowship with Thaddeus Dryja at the Massachusetts Eye and Ear Infirmary and Harvard Medical School working on the molecular genetics of retinoblastoma and retinitis pigmentosa.
Dr. Mukai became a resident in ophthalmology then a vitreoretinal fellow at the Massachusetts Eye and Ear Infirmary and Harvard Medical School which he completed in 1990. During his vitreoretinal fellowship, he carried out research on molecular biology of retinoblastoma at the Whitehead Institute of Massachusetts Institute of Technology. He has been on the staff of Massachusetts Eye and Ear Infirmary and faculty of Harvard Medical School since. He is an Associate Professor of Ophthalmology at Harvard Medical School. He served as the Director of the Vitreoretinal Surgery Fellowship in the Retina Service for 25 years and has trained over ninety vitreoretinal surgeons. He was also the director of residency training during their retina rotations and is active also in teaching medical students and graduate students at the Harvard Medical School. In addition to Massachusetts Eye and Ear Infirmary, he has been on staff at Massachusetts General Hospital, Children’s Hospital of Boston, and Brigham and Women’s Hospital.
Dr. Mukai has clinical interest in pediatric retinal diseases, retinoblastoma, adult retinal surgery including complicated trauma, and genetic diseases that affect the retina. His research interests include molecular biology of retinoblastoma and other genetic diseases that affect the retina, retinal stem cells, new treatment modalities for retinoblastoma and pediatric retinal diseases, molecular biology of proliferative vitreoretinopathy, and new techniques for imaging of the posterior segment. He is a leader in the field of smartphone imaging of the retina as well as development of new inexpensive systems and devices for fundus photography, and he has collaborated extensively with the MIT Media Lab. He also participates in global surgery teaching in developing countries including Uganda, Madagascar, and Haiti. He has numerous publications in peer-reviewed journals including Nature, New England Journal of Medicine, Nature Genetics, Science Advances, Neuron, Genes and Development, PLOS ONE, and American Journal of Pathology in addition to ophthalmology journals such as Ophthalmology, Archives of Ophthalmology/JAMA Ophthalmology, American Journal of Ophthalmology, Progress in Retinal and Eye Research, and Investigative Ophthalmology and Visual Science. In addition, he has numerous chapters in books including Principles and Practice of Ophthalmology and Oncology of Infancy and Childhood.