Researchers are investigating whether a vocal analysis app can detect diabetes.
Studies show that people with diabetes have different vocal traits than those without; these traits vary between men and women.
In the next few years, doctors could diagnose type 2 diabetes with a voice analysis smartphone app. While the technology isn’t here yet, researchers have created computerized software for investigating unique vocal traits of diabetes, which differ for men and women.
A recent study of 267 participants, 75 of whom were diagnosed with type 2 diabetes, found that people with diabetes present certain vocal characteristics not present in those without diabetes. The study looked at data from age- and BMI-matched samples to weed out age and weight as confounding variables. Participants without diabetes were also confirmed not to have the disease through blood testing to eliminate risks of missed diagnosis.
“The biggest takeaway is that we can now identify diabetes through voice,” Yan Fossat, MSc, one of the study’s researchers, says.
Pending additional studies and approvals, the app’s algorithm could take on various implementations, including a website, backend diagnostic tool, or smartphone app, Fossat says. He adds that more studies and an official medical device approval are needed before next steps can occur—a process that he estimates will take a few years, but “not decades.”
Currently in the works is a study comparing vocal changes of people with diabetes and prediabetes to determine if and to what extent voices differ in the process of someone developing the condition.
Detecting diabetes through voice
The researchers evaluated participants based on 14 different voice features, including changes in pitch, variation in the voice, and changes in ‘jitter’ and ‘shimmer,’ two traits associated with increased perceived breathiness, hoarseness, and roughness in the voice. Fossat explains that shimmer is the change in intensity of the voice, while jitter is associated with the variation in voice frequency.
Trends in jitter and shimmer, along with a few other traits, were different for men and women with diabetes. For men, who accounted for 57 of the study’s diabetic participants, the strongest indicator was changes in shimmer. For women, who accounted for 18 diabetic participants, the strongest indicators were changes in vocal pitch, jitter, and intensity.
Pitch differences in women with diabetes tended to be lower or more varied than in women without diabetes, Fossat says. He adds that the voice changes are not detectable by the human ear and thus must be evaluated through software.
“They're very subtle,” Fossat explains. “The human voice in general is full of little variabilities. It just so happens that the level of [variability] is increased by diabetes.”
Sex-based differences are not altogether surprising in the realm of diabetes treatment, as men and women have been found to demonstrate different risk levels for symptoms of the disease. For example, women with diabetes can experience higher rates of edema than men, while men with diabetes can experience higher rates of muscle weakness of peripheral neuropathy than women, says. Men might also experience erectile dysfunction with diabetes.
“This isn’t a new concept of having [some] diabetes complications being more common in males or females,” says Jaycee M. Kaufman, MSc, the first author of the study. “It may be that one of the reasons that we observe the difference is because the presentation of diabetes itself is different.”
Noting that there are differences and what these differences are is crucial for creating an effective diagnostic tool, the researchers say.
Vocal changes in diabetes are not detectable by the human ear
Whether a phone app for patients or an algorithm for telehealth companies, a voice-based diabetes detector could have the most poignant impact on underserved areas in which diabetes cases often go undiagnosed and untreated, Fossat says.
“In a country or region where there [are] very few physicians, it would be great to have this available to the public,” Fossat says. “In places where there are physicians, but there are time or bandwidth issues, it could be something you do on the phone before you see a doctor.”
Fossat adds that, ultimately, the researchers want to help doctors connect to as many people with undiagnosed diabetes as possible.
“People don't necessarily have time, or budget, or access to a physician—and this is why there are tens of millions of people who don't have a diagnosis right now,” Fossat says. “That is the main barrier we're trying to eliminate by speaking into your phone to get a screening.”
What this means for you
Research shows that people with diabetes have specific vocal traits. In the future, researchers hope to use vocal detection apps to help diagnose the disease and connect people to treatment.