Artificial Intelligence for Geographic Atrophy: Pearls and Pitfalls

AI Generated Summary*

Framed as pearls and pitfalls, the review surveys deep learning for geographic atrophy (GA), a late-stage form of age-related macular degeneration, after approval of pegcetacoplan and avacincaptad pegol. Convolutional networks delineate GA on fundus photography, autofluorescence and OCT; in one 900-volume OCT dataset, B-scan-level and en-face-level algorithms detected ellipsoid zone (EZ) and RPE attenuation with accuracies of 91% and 82%, and other models forecast growth from a single baseline scan. FILLY, OAKS and DERBY analyses tied treatment to better preserved retinal layers, with AI-derived EZ attenuation now an FDA-accepted primary outcome and a higher EZ to RPE loss ratio marking likely fast progressors. Pitfalls: dependence on high resolution devices, thin validation in diverse populations, uneven regulation and privacy questions.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.