AI Scan Flags Dementia Years Early

Your next routine eye exam might reveal something your brain hasn’t yet had the nerve to tell you — early warning signs of Alzheimer’s disease hiding in plain sight at the back of your eye.

Quick Take

Researchers have found measurable changes in retinal blood vessels and tissue layers that appear linked to Alzheimer’s risk, sometimes years before memory symptoms show up.
An artificial intelligence model trained on retinal scans correctly identified early-onset Alzheimer’s with 93.55% accuracy in a published study.
A large study of 30,573 people found that a thinner retinal layer called the ganglion cell-inner plexiform layer was tied to higher Alzheimer’s risk.
Experts caution the science is promising but not yet ready for routine screening — no major health organization has endorsed it as a standard test.

Why the Eye Is a Window to the Brain

The retina at the back of your eye is actually brain tissue. It shares the same blood vessel structure and nerve cell makeup as the brain itself. When Alzheimer’s damages the brain, it appears to leave fingerprints on the retina too. Postmortem studies confirm that retinal and brain changes happen at the same time in people with neurodegenerative diseases, giving researchers a rare chance to look at brain health without opening the skull.^[2]

Indiana University School of Medicine researchers published a 2025 study showing that mice carrying the APOE4 gene — the gene variant most strongly tied to human Alzheimer’s risk — developed twisted retinal vessels, narrowed arteries, and reduced branching as early as six months of age.^[1] That is striking. It means the eye may signal genetic Alzheimer’s risk before any cognitive decline is detectable. The retina, it turns out, is not just a passive observer.

What the AI Studies Found in Human Retinal Scans

Duke University researchers built an artificial intelligence model that could tell Alzheimer’s patients apart from healthy people using retinal scans. The key finding was decreased blood vessel density in the macula — the central part of the retina.^[3] Separately, a team at the National University of Singapore developed a model called Eye-AD, which analyzed a type of retinal scan called optical coherence tomography angiography, or OCTA. Eye-AD hit 93.55% accuracy detecting early-onset Alzheimer’s.^[5]

The UK Biobank study added the most weight to the argument. Researchers tracked 30,573 people and found that a thinner retinal layer — the ganglion cell-inner plexiform layer — was consistently linked to higher Alzheimer’s risk.^[8] That is a large, prospective dataset. It is harder to dismiss than a small lab experiment. The direction of the evidence across multiple independent research teams points the same way.

The Honest Limitations Researchers Admit

A 2025 systematic review of 22 studies threw cold water on some of the excitement. It found conflicting results — some studies showed retinal thinning, others showed thickening, and several found no significant difference at all between early Alzheimer’s patients and healthy controls.^[2] That kind of inconsistency matters. It means scientists have not yet agreed on exactly which retinal change to look for, which makes building a reliable screening tool much harder.

Eye-AD also performed noticeably worse when researchers tested it on people with mild cognitive impairment — a pre-Alzheimer’s stage — scoring an accuracy of about 80% on external test data.^[4] That gap is important. The whole promise of retinal screening is catching the disease before symptoms appear. If the tool struggles most at that earliest stage, it has not yet cleared the most critical hurdle.

Where This Research Stands Right Now

No major medical body — not the Food and Drug Administration (FDA), not the National Institutes of Health (NIH), not the Alzheimer’s Association — has approved retinal imaging as a standard Alzheimer’s screening tool.^[2] That is not a dismissal of the science. It is the normal, responsible pace of medical validation. What exists today is a compelling body of early evidence that needs larger, longer, more diverse studies to confirm.

The practical upside here is real and worth watching. Retinal scans are cheap, painless, and already happening in eye doctors’ offices across the country. If researchers can lock down a reliable, specific biomarker, Alzheimer’s screening could become as routine as checking your blood pressure. The science is not there yet — but the trajectory is pointed in a direction that should make anyone over 50 pay close attention to their next eye appointment.