A new study shows that using artificial intelligence to analyze a few seconds of a person's voice can determine whether they have type 2 diabetes with an accuracy of 89%. This non-invasive method promises to revolutionize diabetes screening by eliminating current barriers to testing such as time, cost and travel.
Scientists at KlickLabs see voice technology as a potential breakthrough in detecting type 2 diabetes. Identifying whether a person has diabetes may soon become as simple as speaking a few phrases into a smartphone, a groundbreaking study from a laboratory suggests. The research, which combines speech recognition technology and artificial intelligence, marks a major advance in diabetes identification.
The new study, published in the journal Mayo Clinic Proceedings: Digital Health, outlines how scientists used six to 10 seconds of a person's voice and basic health data, including age, gender, height and weight, to create an artificial intelligence model that could tell whether the person had type 2 diabetes. The model was 89% accurate for women and 86% accurate for men.
For the study, researchers in the Klick lab asked 267 people (diagnosed as non-diabetic or type 2 diabetic) to record a sentence on their smartphones six times a day for two weeks. From more than 18,000 recordings, scientists analyzed 14 vocal signature differences between non-diabetics and people with type 2 diabetes.
"Our study highlights significant vocal differences between people with Type 2 diabetes and those without Type 2 diabetes, which could change the way the medical community screens for diabetes," said first author Jaycee Kaufman, a research scientist in the Klick Laboratory. "Current detection methods require significant time, travel and cost. Voice technology has the potential to eliminate these barriers entirely."
A new clinical study from Klick Labs finds that artificial intelligence and 10 seconds of speech can change the way people are screened for diabetes, providing better service and lower cost than current screening methods. The findings, published in Mayo Clinic Proceedings: Digital Health, showed that voice signatures were 89 percent accurate in predicting type 2 diabetes for women and 86 percent for men. Source: Klick Labs
The team at Klick Lab studied some characteristics of sound that are imperceptible to the human ear, such as changes in pitch and intensity. Using signal processing techniques, scientists were able to detect changes in sound caused by type 2 diabetes. What's surprising, Kaufman says, is that these vocal changes manifest themselves differently in men and women.
According to the International Diabetes Federation, almost one in every two diabetics worldwide, or 240 million adults, do not know they have diabetes, and nearly 90% of diabetes cases are type 2 diabetes. The most commonly used diagnostic tests for prediabetes and type 2 diabetes include glycosylated hemoglobin (A1C), fasting blood glucose (FBG) test, and OGTT, all of which require patients to visit a medical facility.
Yan Fossat, vice president of KlickLabs and the principal investigator of this study, said that Klick's non-invasive and barrier-free method provides the possibility to screen a large number of people and help discover a large number of undiagnosed type 2 diabetes patients.
"Our study highlights the huge potential of voice technology for identifying type 2 diabetes and other health conditions," Fossat said. "Voice technology has the potential to revolutionize healthcare practice as an easy-to-use, affordable digital screening tool."
Fossatt said the next steps will be to replicate the study and expand their research into using speech as a diagnostic tool in other areas such as prediabetes, women's health and hypertension.