2023-10-20 17:01:52
Published20. October 2023, 7:01 p.m
AI in medicine: AI recognizes from your voice whether you have diabetes
Advances in artificial intelligence are driving breakthroughs in medical diagnostics. Ten seconds of audio recordings of a voice are enough to detect diabetes in a patient, according to a new study.
von
Can you still do sports with diabetes? Jasmin (31) was diagnosed at a young age and talks about her experiences with the disease.
20Min
That’s what it’s about
Researchers are finding that the voice provides evidence of diabetes.
In one study, they examined the voices of 267 people for differences.
They found acoustic characteristics that differ between people with and without diabetes.
The artificial intelligence was able to predict diabetes in the participants in 85 to 90 percent of cases.
This type of diagnostics could be launched as a smartphone app in the future.
A short conversation on the phone could soon be enough to give people the first reliable indications of diabetes. A team at research company Klick Applied Sciences in Toronto found that just ten seconds of audio recordings of a voice are enough to detect signs of diabetes in patients. This is made possible thanks to artificial intelligence. The AI program analyzes the voice for subtle clues that we humans hardly notice.
How did the team go about it?
The Klick Labs department conducted a study with 267 participants who were currently experiencing symptoms and were being examined for type 2 diabetes. For two weeks, the participants had to record a sentence on their smartphone six times a day. The researchers analyzed over 18,000 recordings and looked for differences between people who tested positive and negative.
The artificial intelligence identified 14 acoustic characteristics that differed between people with and without diabetes. Using these features, the team was able to predict diabetes in women with 89 percent accuracy. For men it was 86 percent – and this without any blood test, which would be common in diagnostics.
What is diabetes and what types are there?
What are these characteristics?
An important clue in the study was the pitch and how it changed between recordings. “What was interesting was that these changes were different for both sexes,” said the researchers. Micro-variations in the frequency of the voice were also crucial. Humans cannot reliably identify such subtle changes, but programs with artificial intelligence can.
«The voice can be influenced by the respiratory system, nervous system and the larynx. Anything that affects these systems can also affect the voice,” the researchers write. Previous studies have shown that high blood sugar causes long-term damage to peripheral nerves and muscle fibers, which could lead to voice disorders.
Check quickly with a voice message in the app whether there is a risk of diabetes? Probably possible in the future.
Imago
How useful are the finds?
Jaycee Kaufman, who was involved in the study, explained that this discovery could significantly change the way diabetes screenings are done. With this new voice technology, a large number of people could be examined using a smartphone app, for example – without them having to go to a doctor’s office immediately. This could be particularly useful for diagnosing previously unrecognized cases of diabetes. According to the International Diabetes Federation, approximately 240 million adults worldwide are unaware that they have diabetes.
Klick Labs now wants to repeat the study with more people to improve the accuracy. This type of diagnostics should also be expanded to other diseases such as prediabetes or high blood pressure.
The results of the study were published in the scientific journal Mayo Clinic Proceedings: Digital Health published.
Don’t miss any more news
With the daily update you stay informed about your favorite topics and don’t miss any news about current world events.
Receive the most important things, briefly and concisely, directly to your inbox every day.
1697835509
#detect #diabetes #voice