A new study reveals that using AI to analyze a person’s voice for a few seconds can determine whether that person has type 2 diabetes with up to 89% accuracy. This non-invasive method has the potential to revolutionize diabetes screening, eliminating current detection barriers such as time, cost, and travel.
Scientists at Klick Labs have identified voice technology as a potential breakthrough in the detection of type 2 diabetes.
Pioneering research from Klick Labs suggests identifying whether someone has diabetes could soon be as easy as saying a few phrases into your smartphone . This research combines voice recognition technology and artificial intelligence and represents a major advance in the field of diabetes identification.
New research published in Mayo Clinic Proceedings: Digital Health, an AI model that allows scientists to use 6 to 10 seconds of people’s audio and basic health data such as age, gender, height, and weight to distinguish whether an individual has type 2 diabetes. It outlines how it was created.The model has 89% Accuracy 86 percent for women and 86 percent for men.
For the study, Klick Labs researchers asked 267 people (non-diabetic or diagnosed with type 2 diabetes) to record phrases on their smartphones six times every day for two weeks. Scientists analyzed 14 acoustic features from more than 18,000 recordings for differences between a non-diabetic patient and her type 2 diabetic patient.
“Our study reveals significant differences in speech between individuals with and without type 2 diabetes, and could change the way the medical community screens for diabetes,” said lead author of the paper. said Jaycee Kaufman, a researcher at the Crick Institute. “Current detection methods can take a lot of time, travel, and cost. Voice technology has the potential to completely remove these barriers.”
A new clinical study from Klick Labs finds that AI and 10 seconds of audio could change the way people screen for diabetes, improving access and potentially lowering costs than current screening methods. The study, published in Mayo Clinic Proceedings: Digital Health, reported that acoustic voice features were 89 percent accurate in predicting her type 2 diabetes for women and 86 percent for men. Credit: Klick Labs
The team at Klick Labs looked at many audio characteristics, including changes in pitch and intensity, which are imperceptible to the human ear. Scientists were able to use signal processing to detect voice changes caused by type 2 diabetes. Surprisingly, these vocal changes manifested themselves differently in men and women, Kaufman said.
Potential new screening tool for undiagnosed diabetes
According to the International Diabetes Federation, nearly one in two adults living with diabetes worldwide, or 240 million people, are unaware that they have diabetes, and nearly 90% of people with diabetes have diabetes. He is said to have type diabetes. The most frequently used diagnostic tests for prediabetes and type 2 diabetes include glycated hemoglobin (A1C) in addition to the fasting blood glucose (FBG) test and OGTT. All of these tests include a visit to the patient’s health care provider.
Jan Fossat, vice president of Klick Labs and the study’s principal investigator, said Klick’s non-invasive and easy-to-access approach will test vast numbers of people and eliminate the vast majority of undiagnosed type 2 diabetes. He said it could help identify patients.
“Our study highlights the great potential of voice technology in identifying type 2 diabetes and other health conditions,” Fossat said. “Voice technology has the potential to revolutionize healthcare settings as an accessible and affordable digital testing tool.”
Fossat said the next step is to replicate this study and expand research using audio as a diagnostic in other areas such as prediabetes, women’s health, and hypertension.
Reference: “Acoustic Analysis and Prediction of Type 2 Diabetes Using Smartphone Recorded Audio Segments” by Jaycee M. Kaufman, Anirudh Thommandram, and Yan Fossat, October 17, 2023. Mayo Clinic Proceedings: Digital Health.
DOI: 10.1016/j.mcpdig.2023.08.005