We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
The retinal age gap-particulate hypothesis: environmental exposure as a latent contributor to deep learning-based retinal age prediction
Summary
AI can scan photos of your retina (the back of your eye) and guess your age surprisingly well—and when the guess is older than your real age, it's linked to higher disease and death risk. This paper proposes a new idea: some of what the AI is "seeing" might be tiny buildups of microplastics and other pollution particles in the eye, not just typical aging. It's an early, untested theory—not a proven finding—but if true, it could mean a simple eye scan might one day help track how much environmental pollution is affecting your body.
Deep learning models applied to retinal fundus photographs and optical coherence tomography (OCT) can predict chronological age with remarkable accuracy. The discrepancy between predicted and chronological age—the "retinal age gap”—is associated with mortality and multiple systemic diseases. However, the biological and physical correlates underlying these predictions remain incompletely understood. We propose that cumulative environmental particulate exposure—including microplastics, nanoplastics, and other airborne or ingested particulates—may contribute as a previously unrecognized component of the retinal age signal captured by deep learning models. Specifically, we hypothesize that particulate-associated retinal features, potentially manifesting as hyperreflective foci on OCT or tiny discrete foci on fundus imaging, may constitute one of several feature classes utilized by these models. We outline seven falsifiable predictions spanning epidemiology, imaging analysis, experimental manipulation, and post-mortem validation. Importantly, this framework does not posit that particulate burden is the primary determinant of retinal age, but rather that it may act as a latent, cumulative exposure signal embedded within a multifactorial aging phenotype. If supported, this hypothesis would expand the interpretability of retinal age models and position retinal imaging as a potential tool for studying long-term environmental exposure. This work is intended as a testable framework to guide empirical investigation rather than a definitive explanatory model.