A machine learning model developed by researchers at Duke Health can use retinal images from the eye to distinguish between normal cognition and mild cognitive impairment.
The model analyzes retinal images and related data and recognizes specific features to identify individuals with mild cognitive impairment.Publication in magazines ophthalmic scienceThis model demonstrates the potential for a non-invasive, inexpensive method to identify early signs of cognitive impairment that may progress to Alzheimer’s disease.
“This is a particularly exciting study because previous models have been unable to distinguish between mild cognitive impairment and normal cognition,” said Duke University professor of ophthalmology and neurology and associate professor of surgery. said lead author Sharon Fecklatt, M.D., Ph.D. . “This study brings us one step closer to detecting cognitive impairment early, before it progresses to Alzheimer’s disease.”
Fekrat and colleagues previously developed a model that successfully identifies patients diagnosed with Alzheimer’s disease using retinal scans and other data. Optical coherence tomography (OCT) and OCT angiography (OCTA)-based scans have detected structural changes in the neurosensory retina and its microvasculature in patients with Alzheimer’s disease.
The current study extends that work to use machine learning techniques to detect mild cognitive impairment, which is often a precursor to Alzheimer’s disease. The new model identifies specific features in OCT and OCTA images that indicate the presence of cognitive impairment, along with patient data such as age, gender, visual acuity, and years of education, as well as quantitative data in the images themselves.
The researchers reported that the model was able to analyze retinal photographs and images along with quantitative data and was able to distinguish between people with normal cognitive abilities and those diagnosed with mild cognitive impairment. The sensitivity was 79% and the specificity was 83%.
“This is the first study to use retinal OCT and OCTA imaging to distinguish people with mild cognitive impairment from those with normal cognitive function,” said co-author C. Ellis Weisley, M.D., Ophthalmology. assistant professor) said.
“Especially as new treatments for Alzheimer’s disease may become available, it’s becoming increasingly important to have non-invasive and inexpensive means to reliably identify these patients,” said Weisley. said Mr.
“The retina is the window to the brain, and machine learning algorithms that utilize non-invasive, cost-effective retinal imaging to assess neurological health status are a powerful tool for screening patients at scale.” It could be,” co-lead author Alexander Richardson said. Student in the Ophthalmic Multimodal Imaging Lab in Neurodegenerative Diseases at Duke University.
For more information:
C. Ellis Wisely et al, Convolutional Neural Networks Using Multimodal Retinal Imaging to Differentiate Mild Cognitive Impairment from Normal Cognition, ophthalmic science (2023). DOI: 10.1016/j.xops.2023.100355
