Machine Learning Scans Retinal Images to Predict Alzheimer's Disease

Health IT Analytics reports that a study in the British Journal of Ophthalmology detailed how machine learning tools can analyze certain types of retinal images to detect Alzheimer's disease in symptomatic individuals. Duke University researchers designed a tool that reads combined retinal images to identify patients known to have Alzheimer's disease, based on earlier work that identified changes in retinal blood vessel density that correlated to changes in cognition. People with Alzheimer's disease had lower density within the capillary network around the center of the macula. The machine learning model was trained on four types of retinal scans as inputs, so it could discern relevant differences among images. The team fed scans from 159 study participants — 123 cognitively healthy and 36 with Alzheimer's disease — to the model. "We tested several different approaches, but our best-performing model combined retinal images with clinical patient data," said Duke's C. Ellis Wisely. "Our convolutional neural network differentiated patients with symptomatic Alzheimer's disease from cognitively healthy participants in an independent test group."