AI Model Detects Pediatric Eye Diseases Using Smartphone Images

Credit: Vitolda Klein/Unsplash

A detection model using artificial intelligence (AI) displayed a strong performance for early and accurate detection of pediatric eye diseases using smartphone images, including myopia, strabismus, and ptosis.¹

Examining more than 1400 facial photographs from nearly 500 individuals, the deep learning-based AI model achieved high accuracy in detecting all 3 pediatric eye diseases, with a comparable performance in both male and female children.

“These results suggest that AI prediction models utilizing smartphone photographs may identify eye diseases in children and adolescents, providing a handy and early diagnostic tool for families to use at home,” wrote the investigative team, led by Lin Li, MD, PhD, and Jie Xu, DHM, department of ophthalmology, Shanghai Jiao Tong University School of Medicine.

Myopia, strabismus, and ptosis are common pediatric eye problems that pose a significant risk to visual health, overall well-being, and childhood development.² Early screening and identification are critical for successful disease management, but delays are commonly noted due to the need for an in-hospital visit with an ophthalmologist.³

AI’s development across medicine, and within ophthalmology, has transformed the identification of eye diseases by limiting the need for in-person screening.⁴ In this study, Li and colleagues sought to invent a deep learning–based model that uses mobile smartphone images to predict myopia, strabismus, and ptosis in children and adolescents at home.

The cross-sectional study occurred at the investigator’s institution from October 2022 and September 2023. Participants were ≤18 years of age, had been diagnosed with one of the pediatric eye diseases, and cooperated during facial image acquisition.

Among the 476 patients matching the inclusion criteria, 225 (47.27%) were female and 299 (62.82%) were aged between 6 and 12 years. A total of 1419 images were obtained to build the model, of which 946 monocular images were used to identify myopia and ptosis and 473 binocular images were used to identify strabismus.

Overall, 251 patients had myopia, 180 had strabismus, and 171 had ptosis. Upon analysis, the AI-based model showed good sensitivity in detecting all 3 pediatric diseases: myopia (0.84 [95% CI, 0.82–0.87]), strabismus (0.73 [95% CI, 0.70–0.77]), and ptosis (0.85 [95% CI, 0.82–0.87]).

The model also demonstrated good specificity across all 3 diseases: myopia (0.76 [95% CI, 0.73–0.80]), strabismus (0.85 [95% CI, 0.84–0.86]), and ptosis (0.95 [95% CI, 0.93–0.97]). Accuracy was additionally strong for the model: myopia (0.80 [95% CI, 0.78–0.81]), strabismus (0.80 [95% CI, 0.79–0.82]), and ptosis (0.92 [95% CI, 0.91–0.93]).

In subgroup analysis, the model demonstrated a comparable performance in identifying these pediatric eye diseases across females and males within different sex-based cohorts. However, differences in sensitivity and specificity were noted for different age subgroups.

Specifically, the model demonstrated good sensitivity for myopia in those aged 13–18 and 6–12 years (0.85) but remained low for those aged 0–5 years (0.69). The model showed maximum sensitivity (0.78%) for strabismus in those aged 13–18 years and lowest (0.67) in those aged 0–5 years.

Overall, Li and colleagues noted the AI model achieved relatively high sensitivity in detecting all 3 pediatric eye diseases across different age and sex groups. These results suggested the model could assist families in screening children, reducing the risk of visual loss due to delays.

“Using such information can help achieve a more equitable allocation of limited medical resources. This is critical to the advancement of global health standards,” Li and colleagues wrote.

References

1. _{Shu Q, Pang J, Liu Z, et al. Artificial Intelligence for Early Detection of Pediatric Eye Diseases Using Mobile Photos. JAMA Netw Open. 2024;7(8):e2425124. doi:10.1001/jamanetworkopen.2024.25124}
2. _{Kandel H, Khadka J, Goggin M, Pesudovs K. Impact of refractive error on quality of life: a qualitative study. Clin Exp Ophthalmol. 2017;45(7):677-688. doi:10.1111/ceo.12954}
3. _{Garcia SSS, Santiago APD, Directo PMC. Evaluation of a Hirschberg Test-Based Application for Measuring Ocular Alignment and Detecting Strabismus. Curr Eye Res. 2021;46(11):1768-1776. doi:10.1080/02713683.2021.1916038}
4. _{Dong L, He W, Zhang R, et al. Artificial Intelligence for Screening of Multiple Retinal and Optic Nerve Diseases. JAMA Netw Open. 2022;5(5):e229960. doi:10.1001/jamanetworkopen.2022.9960}