...
By Hannah Millington
Health Reporter
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
A new AI-powered, speech-based screening tool could help people assess whether they are showing signs of Parkinson's disease at home.
Developed as part of a study by University of Rochester computer scientists, the web-based test asks users to recite two pangrams, short sentences using every letter of the alphabet.
In just seconds, the AI can analyze the voice recordings to look for subtle patterns linked to Parkinson's—with nearly 86 percent accuracy.
Read More
Parkinson’s Breakthrough Could Help Prevent Brain Cell Loss5 min read
Parkinson’s Disease Reversed by Brain-Stimulating Nanoparticles3 min read
This New Test Could Diagnose Parkinson’s With AI4 min read
The researchers believe such a tool could be implemented as a feature in popular speech-based interfaces like Amazon Alexa or Google Assistant (with users consent).
"While we haven't established partnerships with these companies yet, our model is very lightweight and well-suited for such integration," paper author and computer scientist Tariq Adnan told Newsweek.
"With appropriate privacy safeguards and user consent, it could operate passively or on-demand through such widely used home devices.
"Once integrated, [in theory] a user could say something like 'screen me for Parkinson's symptoms,' and the system would prompt a short sentence to be read aloud.
"A risk score would then be generated, along with a recommendation to consult a clinician if warranted."
...
Parkinson's is typically diagnosed by specially trained neurologists through a combination of examinations, brain imaging and family history.
While the new tool is not intended to be a replacement for a clinical diagnosis, the researchers see it as a fast, low-barrier and accessible way to flag people who might have the disease and encourage them to seek thorough clinical evaluation.
Around 1.1 million people in the U.S. are living with Parkinson's disease, expected to rise to 1.2 million by 2030, according to Parkinson's Foundation.
Nearly an estimated 90,000 people are diagnosed across the country each year, making it the second-most common neurodegenerative disease after Alzheimer's. While incidence increases with age, some four percent of people are diagnosed before 50.
"There are huge swaths of the U.S. and across the globe where access to specialized neurological care is limited," said Rochester computer science professor Ehsan Hoque in a statement.
"With users' consent, widely used speech-based interfaces like Amazon Alexa or Google Home could potentially help people identify if they need to seek further care."
Most people living with Parkinson's in the U.S. seek their care from neurologists, with approximately 50 percent seeing a community neurologist and 9 percent seeing a movement disorders specialist, a previous Parkinson's Foundation study found.
However, nearly 40 percent of people with the disease do not see a neurologist for their care but see a primary care provider (29 percent) or no provider (11 percent.)
...
To train and validate the new tool the researchers collected data from more than 1,300 participants—with and without Parkinson's—across home care settings, clinical settings at the University of Rochester Medical Center and the InMotion Parkinson's disease care center in Ohio.
Using their computer's microphone, participants were asked to read aloud the following sentences: "The quick brown fox jumps over the lazy dog. The dog wakes up and follows the fox into the forest, but again the quick brown fox jumps over the lazy dog."
By utilizing advanced semi-supervised audio models trained on millions of digital recordings to understand the characteristics of speech, the tool can obtain enough vocal cues from the short sentences to flag warning signs.
"For example, the way someone with Parkinson's would utter sounds, pause, breathe, and inadvertently add features of unintelligibility is different in someone without Parkinson's," explained paper author and computer science master's degree student Abdelrahman Abdelkader in a statement.
The tool was 85.7 percent accurate when tested, providing a strong indication of whether someone may have Parkinson's, according to the study authors.
"We envision a gentle, user-centered experience. The tool would never deliver a diagnosis, rather would only flag whether a follow-up with a clinician might be worthwhile," Adnan explained.
Clear disclaimers would emphasize that the result is not definitive, especially since Parkinson's may not always affect speech.
"To support users emotionally, the interface could provide information about trusted resources, helplines, and suggestions on what to do next."
...
As well as speech, some people can display signs of Parkinson's through motor movement tasks or facial expressions. Over the past decade, this has led Hoque's lab to pursue algorithms to combine multiple indicators with state-of-the-art results.
"Research shows that nearly 89 percent of people with Parkinson's have a deformity in their voice that can be indicative of the disease, making speech a strong starting point for digital screening," said Adnan in a statement.
"By combining this method with assessments of other symptoms, we aim to cover the majority of people through our accessible screening process."
On how such an AI tool would ensure data protection, Adnan said: "Any practical deployment would extract speech features locally and discard raw audio immediately. No voice recordings would be stored or transmitted without consent. On-device processing is feasible to enhance user privacy."
An interactive, browser-based demonstration of the lab's three screening tests, including this new speech test, is available to try out at home here.
Do you have a health story to share with Newsweek? Do you have a question about Parkinson's? Let us know via health@newsweek.com.
Update 07/28/25, 05:26 a.m. ET: This article was updated to include more information from Tariq Adnan.
References
Adnan, T., Abdelkader, A., Liu, Z., Hossain, E., Park, S., Islam, M. S., & Hoque, E. (2025). A novel fusion architecture for detecting Parkinson's Disease using semi-supervised speech embeddings. Npj Parkinson's Disease, 11(1), 176. https://doi.org/10.1038/s41531-025-00956-7
Pearson, C., Hartzman, A., Munevar, D., Feeney, M., Dolhun, R., Todaro, V., Rosenfeld, S., Willis, A., & Beck, J. C. (2023). Care access and utilization among medicare beneficiaries living with Parkinson's disease. Npj Parkinson's Disease, 9(1), 108. https://doi.org/10.1038/s41531-023-00523-y
View 0 Comments
Submit Correction
View Editorial & AI Guidelines