Spotting Alzheimer’s Early

A machine learning algorithm has been developed that can identify structural changes in the brain that are characteristic of Alzheimer’s Disease (AD) nearly a decade before symptoms appear. Such a tool would be incredibly useful in slowing the pace of the disease, as most of the drugs currently in development work best the earlier they are administered.

The algorithm was initially developed using magnetic resonance imaging (MRI) brain scans from 67 patients—38 of whom had been diagnosed with AD and 29 of whom who had not. The scans were divided into small regions, and the neuronal connections between them were analyzed.

The system was then tested on an additional set of scans from 148 subjects—52 of whom were healthy, 48 of whom had AD and 48 of whom had mild cognitive impairment (MCI) but had subsequently developed AD up to nine years later. The algorithm correctly distinguished between healthy brains and those from AD patients 86 percent of the time. More importantly, it was able to correctly differentiate MCI with an accuracy of 84 percent.

In the future, the researchers hope to apply the same technique to diagnosing other conditions, including Parkinson’s Disease.