Detecting Alzheimer’s through speech
By Joel Wittnebel/Active Senior’s Digest
The way someone talks can tell you a lot about them. One company believes it could even tell you whether they have Alzheimer’s disease.
New technology, developed by Toronto-based Winterlight Labs, can detect changes in the way Alzheimer’s patients speak in order to diagnose the disease with an 85 per cent accuracy rate.
“One of the promises of this type of technology is that it kind of opens the door to be able to test people earlier, quicker, and for much less costs than some of the other technologies out there,” says Liam Kaufman, CEO and co-founder of Winterlight Labs.
Working with Dr. Frank Rudzicz, Maria Yancheva, and Katie Fraser from the University of Toronto, Winterlight developed the software that uses artificial intelligence to analyze approximately 400 different variables in the way a person speaks. Through an exercise of having a patient describe a photograph, the technology is able to look at many different factors coming out of two main components of a person’s speech, the words they use and how they use them. Because according to Kaufman, people with Alzheimer’s disease speak very differently.
“You tend to have word-finding difficulties,” he says. “Those pauses and hesitations can actually produce acoustic difference and changes the way people speak.”
People with Alzheimer’s will also use more general terms when they can’t quite find the word they’re looking for.
“Say they’re trying to think of the word SUV, and they can’t, they may just default to car,” he says.
They also tend to speak slower and use more pronouns when referring to things among other slight differences.
“There’s sort of like a whole constellation of changes that when you put them together allows to reliably detect if this person has Alzheimer’s disease or not,” Kaufman says.
Thanks in part to a massive data set of recordings made up of 400 people, 200 with the disease and 200 without describing the same picture, in laboratory tests, the software can detect the differences between Alzheimer’s, ephasia, and Parkinson’s with between 85 and 100 per cent accuracy.
And the test is quite simple, a one to five minute exercise in describing a photo, something many seniors already do, and the software can provide a clear idea about the person’s cognitive state. Compared to other tests that currently exist, this one could be a game-changer.
“Most assessments right now, they’re stressful,” Kaufman says. “You’re asking somebody with Alzheimer’s disease to remember three words then five minutes later you’re asking them what those three words are.”
Not only can they be stressful for the patient, but they can be time consuming, some pen and paper tests taking hours to complete, while other more technologically advanced methods are too expensive for general use.
“We’re not asking them to remember anything, we’re not asking them to do some complex task, we’re simply asking them to describe a picture. So it’s a much more natural assessment.”
And while Winterlight currently makes no claims that the software can detect early-onset Alzheimer’s, as it has only been tested on individuals known to have the disease, it is something Kaufman is hopeful will be coming down the road following further testing, some of which will be occurring in the coming months.
“From previous research that shows that peoples’ speech changes very early on…we think that there is a chance that we could detect it before someone is diagnosed, which could be quite helpful,” he says.
Helpful indeed, as new statistics show that currently over 47.5 million individuals in the world have dementia. That number is expected to triple by 2050.
“This technology really promises to help,” says Bridgette Murphy, manager director with AGEWELL, an organization dedicated to research in the field of technology and aging. “The need is ever increasing.
This past July, Winterlight won AGEWELL’s Pitch Competition: Technology to Support People with Dementia, earning the honour among a field of 10 other groundbreaking technologies.
“What they’ve developed is able to very quickly and accurately monitor cognitive health,” Murphy says. “So it’s really leaps and bounds ahead of what’s available now in terms of monitoring and detection.”
Thanks to funding from AGEWELL, Winterlight will begin field tests with the software in the hope of obtaining regulatory approval in Canada and the United States to make it available for family doctors and speech-language pathologists.