Offline AI-assisted fundus imaging for preliminary screening of childhood glaucoma.

PURPOSE: To evaluate the feasibility and performance of AI-assisted portable fundus photography in children with glaucoma.

METHODS

This case series describes the use of smartphone-based fundus imaging with integrated offline glaucoma artificial intelligence (AI) in children (≤18 years) with various types of glaucoma who were evaluated at a tertiary eye care center, and comparing AI-derived referral recommendations with presence or absence of structural optic nerve changes determined on clinical evaluation.

RESULTS

AI-assisted fundus photography and clinical evaluation were completed for 21 children. For vertical cup:disk ratio, the 95% limits of agreement between AI image analysis and clinical assessment on Bland-Altman analysis were -0.15 to +0.23. The AI output recommended referral for 7 of 8 children who had structural optic nerve changes on clinical examination and did not recommend referral for any of the children without such optic nerve changes.

CONCLUSIONS

AI-assisted smartphone-based fundus imaging may help identify cases of pediatric glaucoma exhibiting structural optic nerve changes.