Can speech technology help detect Parkinson’s early?

Ninety percent of people with Parkinson’s develop speech problems. Their speech becomes monotonous, vowels become similar and breathing is less smooth. And because they are more difficult to understand, Parkinson’s patients are more often than others in social isolation, which in turn can lead to depression.

It is just one of the problems that Parkinson’s patients face, but with major consequences. Early detection of the disease would help to start speech therapy on time. But that early diagnosis is actually a problem, for which there is no simple test yet. According to Vass Verkhodanova, speech can play a key role in this. For her dissertation, which she will defend next week at the University of Groningen, she investigated how people recognize this speech. The ultimate goal is to use speech technology to detect Parkinson’s at an early stage.

Over the past 25 years, the number of Parkinson’s patients worldwide has grown from 3 million to more than 6 million.

“We know that changes in speech occur much earlier than known symptoms such as jerky movements. That is why scientists all over the world are now looking for features of that spoken language in order to detect Parkinson’s earlier.”

In Parkinson’s disease, a lack of dopamine in the brain causes a large part of the complaints. This concerns, for example, slower movements, stiffness of muscles and vibration of certain parts of the body. They all have to do with a loss of control over your muscles. Because many muscles are involved in speaking, most Parkinson’s patients have to deal with speech problems.

Unbiased Listeners

Verkhodanova herself conducted a large-scale study in which she showed audio fragments of both healthy Dutch people and Dutch people with Parkinson’s to about two hundred listeners all over the world, both regular listeners and speech therapists. “Remarkably enough, it was precisely the international group, so the people who did not speak Dutch, were best able to extract the ‘unhealthy’ speech fragments.” If you think about it, it’s not too bad, says Verkhodanova, because this group was not concerned with interpreting the message, but was purely focused on the speech signal.

Among the listeners were people from the Netherlands, Germany and England, but also from Russia and the Ukraine. The latter group mainly came from the personal circle of acquaintances of the researcher, who has a Slavic background. But the group that scored highest on recognizing ‘unhealthy’ speech consisted mainly of Germans and English. The researcher explains: “You can perhaps imagine that for Russian native speakers who hear Dutch for the first time, the hard Dutch g as in Groningen sounds like someone is choking on popcorn. Because they hear so many unfamiliar sounds, they are also a bit distracted.”

It was also striking that the ordinary listeners scored better than those who worked as speech therapists in daily life. This group also had the ability to listen without prejudice. “And that is actually positive,” says the researcher, “because it means that people from the nearby environment can play an important role in signaling disease.”

Language therapy helps

The fact that the group that was purely focused on the speech signal scored high is good news for speech technology. After all, it means that the computer software that is being developed to detect Parkinson’s through speech does not need to have knowledge of a specific language. During her PhD research, Verkhodanova also performed acoustic measurements on speech with a self-learning algorithm. Previous research has shown that articulation and prosody (rhythm, stress and intonation) are most characteristic of people with Parkinson’s. “They talk more slowly and allow sounds to flow into each other more quickly. Then, for example, you will not hear a difference between the ie and the oe sound. This has to do with the reduced functioning of the muscles in the vocal tract.”

Also from the acoustic measurements, articulation and prosody emerged as the most characteristic aspects of speech in people with Parkinson’s. People can eventually be screened for these parts for an early diagnosis of the disease. In her research, the PhD student made an important first step towards such a screening, by closely following a Parkinson’s patient for four years. Such long-term studies are scarce. “My test subject was bilingual Dutch-English. He had been diagnosed with Parkinson’s six years earlier, but had no symptoms of a speech disorder, according to his neurologists. In the first year that I followed him, I already witnessed the decline in his speech. At that time he also started speech therapy, and as a result his speech improved again, which was wonderful to see.”

For example, in speech therapy, the patient practiced with the volume of his voice and the pitch. Because these characteristics are not language specific, this had a positive effect on the pronunciation of his Dutch as well as his English. “My study is one of the few that has followed a bilingual patient, which is quite remarkable considering we live in a multilingual world.”

Minority Language App

Verkhodanova is working with other researchers on an app that allows people to train their own speech when it deteriorates. Such apps already exist, she explains, but they are not accessible to everyone because they are paid apps and also not made for speakers of minority languages. Verkhodanova and her colleagues are working on an app that is free and also suitable for native speakers of minority languages, such as Frisian or Groningen. “That is very important because many people with dementia or Parkinson’s revert to their native language – and that is not always Standard Dutch.”

The ultimate goal of the researchers is to develop an app in the future that can screen people for Parkinson’s at an early stage. Other countries are also working towards this with the help of speech technology. A major problem in this research is that the algorithms are trained on datasets from a fairly specific population. Moreover, it concerns people who are vulnerable, as a result of which data may not be exchanged between different countries.

Nevertheless, the researcher is hopeful that speech technology can one day be used to screen people at an early stage and to help Parkinson’s patients improve their speech. “Speech loss is just one of the problems these patients face, but it is a major determinant of their quality of life. You want to prevent people from becoming socially isolated by starting language therapy early.”


Source: Kennislink by www.nemokennislink.nl.

*The article has been translated based on the content of Kennislink by www.nemokennislink.nl. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!