Speech test may help diagnose Parkinson's

Copied from The Northwest Parkinson’s Foundation Weekly News Update

Using professional recording equipment and computer algorithms, researchers have had success detecting tiny changes in speech patterns that may point to the disease in its early stages.
Robin Erb

www.usa.today.com - - DETROIT -- Identifying Parkinson's disease one day might be as easy as a speech test -- a non-invasive, inexpensive tool that in early experiments has shown surprisingly accurate results, a Michigan State University researcher said Wednesday.

The key: Tiny changes in speech that occur early in Parkinson's development and can be detected by professional recording equipment and computer algorithms -- perhaps even before loved ones notice slurred or slow speech or other symptoms, said Rahul Shrivastav, professor and chairman of Michigan State University's Department of Communicative Sciences and Disorders.

The method detected those with Parkinson's and those without nine out of 10 times.

Changes in speech patterns are noticeable in late-stage Parkinson's, Shrivastav said. But in early stages, "the changes are small … they are not necessary big enough to notice."

Parkinson's affects nerve cells that use the brain chemical dopamine to help control muscle movement. Over time, the dopamine levels fall and nerve cells no longer properly send messages, leading to loss of muscle function, tremors and even dementia, Shrivastav said.

That can affect speech because muscle movements in the jaw and tongue become slow and have a reduced range, he said.

Over several years, Shrivastav and colleagues at the University of Florida's Department of Speech, Language and Hearing Sciences tested 76 men and women ranging in age from their 40s to their 80s, half of whom had been diagnosed with Parkinson's.

They asked the participants to recite 10 sentences that covered different sounds, including: "The beer drinkers raised their mugs" and "The boat sailed along the coast."

Special software developed by the team then dissected the speech sentence by sentence and in pieces that were just 1/50th of a second long.

Using thousands of readings, the software this year correctly differentiated the speech between patients with Parkinson's and those without Parkinson's most of the time. It was so sensitive, in fact, that it could make a determination with just two seconds of speech.

"We anticipated good results, but getting two seconds and 90 percent -- that was a surprise," Shrivastav said.

Parkinson's is one of the most common nervous system disorders of the elderly, and about 50,000 new cases are reported annually, according to the National Institutes of Health.

Researchers have tried for years to understand the disease, but the cause and cure remain evasive, said Maureen Gartner, information and referral nurse for the Cincinnati-based Tri-State Parkinson's Wellness Chapter, a chapter of the American Parkinson Disease Association that covers Michigan.

"We don't know where it starts, so any new twist that could help us diagnose it earlier would be great," she said.