NEW YORK– Researchers at Klick Labs have identified new vocal patterns associated with Type 2 diabetes using artificial intelligence and smartphone voice recordings, potentially advancing the development of noninvasive methods for detecting the disease.
The study, published in the IEEE Journal of Biomedical and Health Informatics, analyzed 14,759 voice recordings from 461 adults to identify acoustic patterns linked to Type 2 diabetes.
Researchers converted recordings of randomly selected short English sentences into visual representations of sound frequencies, known as spectrograms, which were analyzed using an artificial intelligence model called an Audio Spectrogram Transformer.
The approach allowed researchers to identify complex acoustic patterns without relying on participants’ age, body mass index or predetermined voice characteristics.
Using voice recordings and participant sex, the AI models achieved balanced accuracy rates of 71.3% for women and 72.4% for men. The researchers said the findings demonstrate the potential of vocal biomarkers as a tool for identifying patterns associated with diabetes.
“Now, we can literally see many more clues related to diabetes within vocal biomarkers,” said Yan Fossat, senior vice president of Digital Health R&D at Klick Applied Sciences and principal investigator of the study.
“By capturing relationships across frequencies and time, we expanded our analysis beyond shimmer, jitter, and other predefined voice features and found complex combinations of acoustic patterns and features that weren’t on anyone’s radar,” Fossat added.
The research builds on a 2023 Klick Labs study published in Mayo Clinic Proceedings: Digital Health, which identified 14 acoustic features associated with Type 2 diabetes, including pitch, intensity, shimmer and jitter.
That earlier study used fixed voice recordings along with basic health information, including participants’ age and body mass index. It reported conventional accuracy rates of 89% for women and 86% for men.
Researchers cautioned that the accuracy figures from the two studies cannot be directly compared because they used different methods, datasets and performance measures.
The latest findings expand Klick Labs’ research into vocal biomarkers, which examines how changes in voice characteristics may be associated with underlying health conditions.
In addition to diabetes, the company has investigated potential vocal biomarkers for hypertension and ovulation. Its previous diabetes research has also examined the effects of audio compression, blood glucose levels and voice recording frequency on biomarker detection.
In June 2026, Klick Labs announced a series of collaborative studies with Mayo Clinic in Florida to explore the use of vocal biomarkers in identifying several chronic diseases and serious health conditions.
The company said the latest findings support continued research into AI-powered voice analysis as a potential noninvasive approach to health screening, although further validation would be needed before such technology could be used clinically.
