AI Can Catch Alzheimer’s 10 Years Before Symptoms Show. Here’s HowSep 20, 2017
Doctors might soon have a new tool to help them catch Alzheimer’s disease years before tell-tale symptoms emerge. A research team in Italy has developed a new AI algorithm that detects changes to the brain which ultimately lead to Alzheimer’s. With this, the disease could be diagnosed almost a decade beforehand, giving patients a better shot at slowing down the symptoms.
In this new method, a machine-learning algorithm spots structural changes in how brain regions are wired. The research team trained the AI algorithm with 67 MRI scans of the brain. Once they selected brain scans of both healthy people and Alzheimer’s patients, they segmented each scan into smaller patches, and analyzed how they interconnected.
When they showed the AI a new set of brain scans, the AI was able to tell healthy brains apart from Alzheimer patients’ brains with 86 percent accuracy. It could also separate, with similar accuracy, brains with mild cognitive impairment; the catch is that those were brains that went on to develop Alzheimer’s 2.5 to nine years later. This means that the algorithm caught crucial alterations that triggered Alzheimer’s almost a decade before symptoms became apparent.
Whether or not this AI can detect the disease even earlier has yet to be tested – on this study, the researchers used scans from the University of Southern California’s Alzheimer’s Disease Neuroimaging Initiative, and they were limited to the data available there.
The cause of Alzheimer’s is still a cryptic mess. While the buildup of sticky beta-amyloid plaques is a likely culprit, a high-profile clinical trial thought to test that hypothesis once and for all recently failed. Even then, with Alzheimer’s affecting one in 10 people that are age 65 or over, early prediction would give that many people ample time to adjust. Certain treatments are also showing hope, and these likely work better the sooner they are given. Experts also say that the tool would be “immensely powerful” once preventative treatments are underway for the disease.
The study authors say that the new technique also gives more people access to early detection since it is simple, cheap, and non-invasive. “Nowadays, cerebrospinal fluid analyses and brain imaging using radioactive tracers can tell us to what extent the brain is covered with plaques and tangles, and are able to predict relatively accurately who is at high risk of developing Alzheimer’s 10 years later,” co-author Marianna La Rocca told New Scientist. “However, these methods are very invasive, expensive and only available at highly specialized centers.”
There’s more to do to make this AI more powerful – it needs to learn the difference between a person who has mild cognitive impairment but will age normally, or those that eventually suffer from other types of dementia. In the meantime, competitors might get ahead of the game. Blood tests that look for traces of Alzheimer’s are under development, and once it’s ready for the market, it could be even cheaper and simpler than anything else around. Others have invented a digital pen that detects the disease based on how you draw.
Even if AI doesn’t win this one, there’ll be jobs coming its way – since the technique itself is versatile, the research team plans to apply the same principles to predict other neurodegenerative diseases like Parkinson’s.