The AI Eye Scan That Flags Alzheimer’s

Your next routine eye exam might do something no brain scan, blood test, or memory quiz has ever done cheaply and easily — flag Alzheimer’s disease before you forget a single thing.

Quick Take

Researchers have found that changes in the tiny blood vessels and nerve layers of the eye mirror what happens in the brains of Alzheimer’s patients — sometimes years before symptoms appear.
An artificial intelligence model built at the National University of Singapore detected early-onset Alzheimer’s from retinal scans with 93.55% accuracy in a 2024 study.
A large UK study of 30,573 people found that a thinner retinal nerve layer was linked to higher Alzheimer’s risk, giving scientists one of their first large-scale longitudinal clues.
Experts caution the science is promising but not yet ready for your doctor’s office — no major health body has approved retinal screening for Alzheimer’s risk.

Why the Eye Is a Window Into the Brain

The retina is brain tissue. It grows from the same cells, shares the same blood supply patterns, and responds to the same disease pressures as the brain itself. When Alzheimer’s starts damaging the brain, the retina often shows the same damage. Postmortem studies confirm that retinal and brain changes happen at the same time in people with neurodegenerative disease. That biological connection is what makes this research so exciting — and so credible. ^[2]

Indiana University School of Medicine researchers found that mice carrying the APOE4 gene — the strongest known genetic risk factor for Alzheimer’s — showed twisted retinal blood vessels, narrowed arteries, and reduced vessel branching as early as six months of age. ^[1] That is not a subtle finding. It means the genetic risk shows up in the eye before the brain shows obvious damage. The retina, in this case, is sounding the alarm first. ^[7]

What AI Is Finding That Human Eyes Cannot

Standard eye exams use a tool called optical coherence tomography, or OCT, to photograph the layers of the retina in fine detail. A newer version called OCT angiography, or OCTA, also maps blood flow through those tiny vessels. Researchers at the National University of Singapore fed thousands of these OCTA images into an artificial intelligence model called Eye-AD. The result was striking — 93.55% accuracy in detecting early-onset Alzheimer’s by reading changes in the deep vascular layers of the retina. ^[5]

Duke University researchers built a separate artificial intelligence model that could tell the difference between Alzheimer’s patients and healthy people by looking at both retinal structure and capillary density — the thickness and density of the smallest blood vessels in the eye. Reduced capillary density emerged as a key signal. ^[3] Two independent research teams, two different approaches, arriving at the same conclusion: the eye is leaking information about the brain.

The Large-Scale Human Study That Changes the Conversation

Mouse studies and small clinical trials are a start, but they rarely move medicine. The UK Biobank study is different. Researchers tracked 30,573 people and found that a thinner ganglion cell-inner plexiform layer — a specific nerve layer in the retina — was linked to a measurably higher risk of developing Alzheimer’s. ^[6] That is not a small sample in a controlled lab. That is real-world data from tens of thousands of people, and it points in the same direction as every other major study in this field. ^[8]

Where the Science Still Falls Short

A systematic review of 22 studies found conflicting results. Some studies showed retinal thinning in Alzheimer’s patients. Others showed thickening — possibly an early inflammatory response. Researchers have not yet agreed on which specific change is the most reliable warning sign. Most studies were also cross-sectional, meaning they took a snapshot in time rather than following people for years to see who actually developed the disease. ^[2] That gap matters enormously. Knowing someone has a thinner retinal layer today does not yet prove that thinner layer predicted their Alzheimer’s a decade ago.

The Eye-AD artificial intelligence model also performed noticeably worse when detecting mild cognitive impairment — the stage just before Alzheimer’s — compared to full early-onset disease. Its accuracy dropped to an area under the curve of 0.8037 for that earlier stage. ^[4] That is still a solid result, but it exposes the harder problem: catching the disease before even mild symptoms appear is the goal, and that goal remains out of reach for now.

Promising Science That Deserves Honest Expectations

The Food and Drug Administration, the National Institutes of Health, and the Alzheimer’s Association have not endorsed retinal imaging as a standard Alzheimer’s screening tool. That institutional caution is reasonable given the current evidence. The science is genuinely exciting and the biological logic is sound, but medicine demands prospective proof — meaning researchers must show that retinal changes today predict Alzheimer’s ten or twenty years from now in people who are currently healthy. That proof does not yet exist at scale.

What does exist is a compelling and growing body of evidence pointing in a consistent direction. The eye shares biology with the brain. Alzheimer’s changes both. Artificial intelligence can read those changes with impressive accuracy. If researchers can confirm through long-term studies that specific retinal signals reliably predict who will develop Alzheimer’s, the implications are enormous — early intervention, clinical trials targeting the right people, and a screening tool that costs far less than a brain scan. That future is worth watching closely, starting with your next eye appointment.