Speech changes before, after diagnosis evident in celebrities

Years of vocal recordings made by select few could help public at large

Marisa Wexler, MS avatar

by Marisa Wexler, MS |

Share this article:

Share article via email
A woman uses a megaphone to make an announcement.

Analyzing vocal recordings could help in detecting Parkinson’s disease at early stages, according to a study that tested the idea using decades’ worth of recordings from celebrities who went on to develop the neurodegenerative condition.

“This study shows that early patterns of [speech changes] are detectable up to 10 years before diagnosis and highlight significant changes in speech features, supporting speech analysis for early [Parkinson’s] detection and personalized intervention,” the researchers wrote.

The study, “Unveiling early signs of Parkinson’s disease via a longitudinal analysis of celebrity speech recordings,” was published in npj Parkinson’s Disease.

Recommended Reading
An outstretched hand reaches toward another hand with the fingers folded onto the palm.

Poor hand grip strength ID’d as risk factor for Parkinson’s: UK study

Speech changes evident in people with decades of vocal recordings

Parkinson’s often affects the way people talk, leading to changes like difficulty articulating and a quieter voice. Because it’s relatively easy to collect and analyze recordings of people talking, several studies previously explored using vocal recordings as a tool to help diagnose Parkinson’s. Currently, the diagnostic process relies on a specialist looking for indicative disease symptoms, which is a subjective and often arduous process.

While studies have had some success using vocal recordings to detect Parkinson’s, most research in this field has a fundamental limitation. That is, most people don’t have years and years of vocal recordings available, making it impossible for scientists to look at how speech patterns change over time before and after Parkinson’s development.

A team of scientists in the U.S. set out to address this limitation using vocal recordings from celebrities, people with many years of recordings available for analysis.

The study’s main analysis focused on recordings from 15 public figures who disclosed their Parkinson’s diagnosis, with recordings dating up to a decade before and after the diagnosis. Recordings from 15 celebrities without Parkinson’s, matched by age and sex, were used for comparison. The researchers looked at vocal features like the length of pauses, rates of speech, and variations in pitch, using mathematical models to compare the two groups.

Possibility of developing tools to diagnose Parkinson’s early

To calculate how well speech analyses could detect Parkinson’s, the researchers used a statistical tool called the area under the receiver operating characteristic curve (AUC), which measures how well a test can distinguish between two groups (i.e., Parkinson’s or not). AUC values normally range from 0.5 to 1, with higher numbers reflecting better accuracy.

Results showed AUCs of 0.72 for 10 years before diagnosis and 0.75 for five years before diagnosis, indicating a moderate ability to detect the earliest stages of Parkinson’s. These findings “suggest that there exist early patterns of dysarthria [speech abnormalities due to loss of muscle strength and control] that can be leveraged to develop early detection/screening models.”

Data covering years after a diagnosis showed a much higher accuracy, with an AUC of 0.93. “This outcome was expected because manifestations of dysarthria tend to be more evident during the later stages of the disease,” the researchers wrote.

Overall, study findings support the idea that speech analyses could be exploited to help with early Parkinson’s detection, the scientists noted. They suggested insights gleaned from these long-term analyses of celebrity speech patterns could help in developing tools to screen for Parkinson’s based on vocal recordings in the general population.