A form of artificial intelligence is using eye scans to identify patients who have Alzheimer’s disease, said Science Daily. This is expected to aid in diagnosing the disease and predicting it in potential patients.
An interdisciplinary study from Duke University published in the British Journal of Ophthalmology showed that the new AI model peeks at the retinal structure and blood vessels in the eye through images. It looks for characteristics that are known to be related to cognitive changes.
The study’s conclusion serves as a proof-of-concept that machine learning can potentially be used as a non-invasive way of diagnosing Alzheimer’s disease in individuals who are showing signs.
Senior author Sharon Fekrat, M.D. said, “Diagnosing Alzheimer’s disease often relies on symptoms and cognitive testing. Additional tests to confirm the diagnosis are invasive, expensive, and carry some risk.”
The AI model offers a more accessible way of identifying Alzheimer’s in symptomatic patients, with improvements in precision. It can also allow patients to enter clinical trials earlier and plan lifestyle adjustments.
The team is composed of experts in neurology, electrical and computer engineering, and biostatistics, and bioinformatics. Their work is based on a project they worked on earlier, which looks at the changes in retinal blood vessels of people who exhibit cognition changes.
According to their earlier research, patients with Alzheimer’s disease have a decreased density of the retinal blood vessel network around the macula. With this, they built a machine learning model, which they called the convolutional neural network (CNN).
The CNN is trained using four types of retinal images which will train the AI to see relevant distinctions in each image.
To test the AI, the researchers input the retinal scans of 159 participants, 123 of which were healthy and 36 have Alzheimer’s disease.
Duke comprehensive ophthalmologist and lead author C. Ellis Wisely M.D. said, “We tested several different approaches, but our best-performing model combined retinal images with clinical patient data.”
Regarding CNN’s performance, Wisely added, “Our CNN differentiated patients with symptomatic Alzheimer’s disease from cognitively healthy participants in an independent test group.”
Aside from the machine learning developed by the Duke University researchers, organizations such as Neuroglee Therapeutics have been developing artificial intelligence to tackle Alzheimer’s disease.
The company has developed AI-powered digital solutions that gamifies therapeutic activities to rehabilitated mid to moderation symptoms in patients, which can be combined with other therapies.
