Changes to the blood vessels in the brain are characteristic of both Parkinson’s and Alzheimer’s disease, as oxygen deprivation contributes to premature neuron death, and there’s strong evidence that blood vessels in the retina mirror those changes.
Fang noted that vascular changes are present in myriad other conditions including diabetic retinopathy — a vision condition caused when high blood sugar blocks blood vessels to the retina — and traumatic brain injury.
But it’s not about finding a single, silver-bullet biomarker, said Sharon Fekrat, a vitreoretinal surgeon and co-director of the Duke Neurodegenerative Disease Retinal Imaging Repository.To try to figure out which biomarkers are most effective forecasters of Alzheimer’s, Fekrat and her collaborators are feeding extensive patient profiles — including images of the retinal vasculature and measurements of the various layers of the retina in people who have known Alzheimer’s pathology — into a neural network, a form of machine learning that identifies patterns based on a set of training data.So far, according to Fekrat’s neural network, thinning of the retina’s ganglion cell layer has the most predictive power for an Alzheimer’s diagnosis — in other words, the neural network was able to use the thickness of a layer of the retina to distinguish between patients with clinically diagnosed Alzheimer’s and healthy controls of a similar age.Fekrat’s and the Duke team’s vision is for a cheap and accessible retinal scan that could identify warning signs for various neurodegenerative diseases at the same time.