September 19, 2013
The number of Americans living with Alzheimer’s is expected to rise to 7 million by 2025. Alzheimer’s is hard to diagnose - there’s no way to know for sure who has it until after death.
With that challenge in mind, David Schaffer is looking for Alzheimer’s patients so he can record their voices. Schaffer believes if he can eventually get enough samples and enough funding, he could pinpoint how a deteriorating brain reveals itself in speech patterns, because so much of the brain is involved in speaking.
“We have to hear, we have to understand, we have to make decisions, we have to generate the motor controls that move the lips and the tongue, and so it just seems intuitive to me that if there should be some damage in the area of the cortex, either because of a stroke or disease mechanism like Alzheimer’s or brain injury, that it ought to leave a kind of identifiable fingerprint in the speech.”
Schaffer and his colleagues have a list of 100 indicators that they look for, things like pause length and frequency and the complexity of sentences. He uses a chain of software programs to analyze the audio.
“But the engineering challenge is can I change this into something that is so automatic that we could put in onto a computer at a doctor’s office so all they would have to do is collect a speech sample, then push a button.”
To collect the sample, a patient is shown a drawing and asked to describe it.
“One of the things you can hear is the sentences tend to be very short and clipped. Another feature is something they call perseverations, that is repeating the same story over and over again. That’s fairly characteristic of Alzheimer’s.”
Schaffer says a big challenge has been finding enough willing subjects with Alzheimer’s. But studying the connection between subtle language patterns and dementia is not new. A famous study began in the 80s known as the Nun Study established a link between Alzheimer’s affect on an aging brain and what’s called idea density.
Idea density is a ratio of how many new ideas a person includes in a sentence compared to how many words they use. In the Nun Study, researchers examined essays written by a group of 700 young Catholic nuns. They found that the higher the idea density in an essay, the less likely the subject was to show signs of dementia later in life.
Doctor Vineeta Chand is a sociolinguist at the University of Essex. She says it’s hard to take those findings and apply them to the population at large.
“So for example if you want to examine speech rate, if you want to examine lexical changes, if you want to examine pronunciation changes, that predisposes you to requiring a sample that has a similar linguistic and regional and cultural background. Because our language practices reflect where we were brought up and what we were institutionalized towards.”
Chand is working with researchers at the University of California looking for speech patterns that could help diagnose Alzheimer’s. She says the best way to get past the cultural differences in speech is to have patient samples from before and after a person develops dementia.
Schaffer started his research after retiring and losing his wife to Alzheimer’s. He says the changes in her speech were what gave him the idea.
“I noticed for instance she would frequently use a pronoun when she couldn’t remember the noun, hoping that I would fill it in since we were good friends for many, many years, that wasn’t an unlikely assumption…”
There is no treatment for Alzheimer’s, just ways to try and control the symptoms. Schaffer says using computer programs to analyze speech could help with earlier detection of the disease.
“So if we can develop a diagnostic test, even if there is no cure, it might find a role in recruiting people for clinical trials and play a small role in increasing the likelihood that we’ll get a cure in the near future.”
Schaffer is looking for 50 people with early stage Alzheimer’s to participate, and hopes to then move onto larger-scale research if they find patterns. So far, he’s only found three subjects.