Hi, I'm Sam from the University of Edinburgh. The question behind this work is simple: can a photograph of the retina tell us, years in advance, that someone is on a path towards dementia?

The retina is the one place in the body where we can directly see blood vessels and nervous tissue, with a simple photograph. The retina is directly connected to the brain via the optic nerve —  sharing blood supply, embryological origins, and features — so changes in one are to some extent mirrored by changes in the other.

Researchers can train a model to predict a person's age from a retinal photo. When the retina looks older than it should, that "retinal age gap" has been linked to many diseases — including dementia. But that work used a single image per person. We wanted to track change over time, using repeat images from the same people.

We used SCONe dataset — over 600,000 routinely collected photographs from over 70,000 people across Scotland, linked to health records. We trained an age-prediction model and turned it into a simple 0-to-100 score: 50 is average for your age, above 50 means the retina looks older than expected.

We compared people who later developed dementia with matched controls. Two findings. First: in those who go on to develop dementia, the retina ages faster — the groups pull apart years before diagnosis. Second: that rise is faster in males.

The key point is this: these are ordinary eye-test photos, already being taken every day at optician practices. The infrastructure to detect this signal at scale may already exist in countries with a mature optometry sector. Thank you — please do explore the poster.