Artificial intelligence in pediatric eye care: Assessing the necessity and impact

Pediatric eye diseases causing visual impairment are of significant public health concern. Ocular diseases such as myopia, cataract, amblyopia, and keratoconus (KC) are growing burdens among children. If these conditions are not diagnosed early or are missed due to inadequate screening, it can result in delayed treatment and long-term complications. In India, the prevalence of myopia among urban children has gradually increased from 4.44% in 1999 to 21.15% in 2019 and is expected to rise to 48.14% in 2050. Furthermore, the global incidence of childhood cataracts is reported to be 1–15/10,000 live births; India contributes 10% to 12%, and the prevalence of KC is 1.04% in children versus 0.55% in adults.[1] Strengthening the mass screening initiatives in schools and communities is a critical step toward successful comprehensive pediatric eye care. Pediatric ophthalmology is challenging due to the uncooperative behavior of the children, their short attention spans for diagnostic imaging, poorly dilating pupils, high risk of amblyopia, and noncompliance with eye drops.[2] The combination of artificial intelligence (AI) and eye care delivered by pediatric ophthalmologists improves the quality of eye care and early detection of ocular diseases.[3] Several AI-based smartphone applications have been developed to improve pediatric eye care. Wang et al.[4] incorporated two vision screening apps, refraction and strabismus apps, among rural school children. The smartphone-based Logy AI cataract screening solution, which works through Whatsapp to detect cataracts at the home/community level, has been reported to have an overall accuracy of 90.08% for the adult population.[5] Pueyo et al.[6] employed a Track AI project with a device for an integral visual examination (DIVE) for children aged 6 months to 14 years. SI algorithms operating on a smartphone are used to automatically analyze visual data collected by DIVE. A deep learning home screening model has been developed for three common eye diseases – strabismus, myopia and ptosis. Facial photos and smartphone apps had an overall accuracy of 0.80, 0.80 and 0.92, respectively. Strabismus can be detected using CNN based on fixation deviations from eye-tracking data. Lin et al.[7] developed an AI-based cloud-based platform called CC-Cruiser, which automatically detects and grades cataracts from slit-lamp images and recommends treatment. Although the author did a multicenter randomized controlled trial, it observed lower performance in diagnosing cataracts (87.4%0% compared to experts at 99.1%). However, it achieved high patient satisfaction through its rapid evaluation and treatment recommendations. AI-based analysis algorithms, such as random forest and adaptive boosting, have also been used to detect childhood and congenital cataracts. Liang et al.[8] recently trained an AI model based on 101.6 million data points from 1,362,559 children. Although many smartphone apps such as Strabismus tools Strab, e-Paarvai, Strabis Pixi, Peek Acuity, GoCheck Kids, Tre C Orculistica, Myopia, Kay Amb Tracker, and Kay I sight Pro for vision testing are available for adult eye care, designing custom models for pediatric eye care would be of immense potential.[4,9–13] Large datasets are required to train algorithms for pediatric age groups. The rising prevalence of myopia among school children highlights the need for AI to assist in detection and treatment of this condition through mass screening programs. Barraza-Bernal et al.[14] obtained the best results with an algorithm that combines linear support vector and Matern Gaussian process regression to predict Myopia in a study involving 12,780 children. They also compared their findings with existing resources like Myopia calculator (Brien Holden Vision Institute, Sydney, Australia) and MyAppia (Myopia Care by Eyetific, Renens, Switzerland) for Myopia prediction, yielding positive results. Additionally, Du et al.[15] utilized ML algorithms to predict no differences in Spherical equivalent before and after cycloplegic refraction among 2467 school children, demonstrating another effective tool for mass screening in schools. The transformative potential of AI-enhanced screening protocols would aid mass school screening protocols [Fig. 1] in addressing the unique challenges faced in pediatric eye care.Figure 1: Flow chart showing AI-Integrated School Screening Workflow for Ocular Health

Abstract licence: CC BY-NC-SA

1. 4.

   Prediction of myopia development among Chinese school-aged children using refraction data from electronic medical records: A retrospective, multicentre machine learning study · 2018

14 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.