Canadian tech firm Klick Labs has introduced an artificial intelligence tool that can help indicate the presence of type 2 diabetes through a person’s voice. The study detailing this advancement appeared in the scientific journal Proceedings of the Mayo Clinic: Digital Health.
According to Klick Labs, the diagnosis can be formed from a six to ten second voice recording along with basic health details such as gender, age, height and weight. The model reported accuracy rates of 89% for female voices and 86% for male voices.
In the study design, 267 volunteers were enlisted, including individuals with type 2 diabetes and healthy controls. They were asked to record sentences on their smartphones six times daily over a two week period. Researchers then processed more than 18,000 audio samples and identified 14 acoustic features that correlated with the disease’s presence.
One Klick Labs researcher and lead author, Jacy Kaufman, commented that the findings show meaningful voice differences between those with and without type 2 diabetes and may influence future clinical approaches to diagnosis.
Global health data from the International Diabetes Federation show that nearly half of the 480 million people living with diabetes worldwide remain undiagnosed. Roughly 90% of diabetes cases are type 2 diabetes, underscoring the importance of accessible screening methods. (Klick Labs, 2024)
Previously, artificial intelligence methods were applied to the diagnosis of polycystic ovary syndrome, illustrating a broader trend of AI-assisted health screening across conditions.