A Brief Vocal Sample: Novel Technique for Diabetes Detection Unveiled by Scientists

A recent scientific investigation shows that a short vocal recording of an individual, when analyzed by artificial intelligence, can accurately identify the presence of Type 2 diabetes with an accuracy rate of up to 89%. This groundbreaking approach has the capacity to dramatically change the landscape of diabetes screening, mitigating existing challenges such as time, expense, and geographical constraints.

Researchers at Klick Labs have pointed to voice technology as a promising avenue for the accurate identification of Type 2 diabetes.

The innovative study from Klick Labs proposes that diagnosing someone with diabetes could soon be as straightforward as the individual speaking a few sentences into their smartphone. The research integrates voice recognition technologies with machine learning algorithms, representing a notable progression in the domain of diabetes detection.

Published in Mayo Clinic Proceedings: Digital Health, the study details the methodology employed by scientists, who used voice samples lasting between six to 10 seconds in conjunction with fundamental health metrics like age, gender, height, and weight to develop a machine learning model. This model successfully differentiated between individuals with and without Type 2 diabetes, with an accuracy of 89% for females and 86% for males.

In the scope of the study, Klick Labs researchers instructed a sample of 267 individuals, classified as either non-diabetic or Type 2 diabetic, to record a sentence on their smartphones six times a day over a two-week period. An extensive data set of over 18,000 vocal recordings was then scrutinized to identify 14 acoustic features that varied between the two groups.

Jaycee Kaufman, the lead author of the paper and a research scientist at Klick Labs, stated, “The discrepancies in vocal characteristics between individuals with and without Type 2 diabetes are substantial and have the capacity to completely transform the manner in which the medical community conducts diabetes screenings. Current diagnostic methods often necessitate considerable investment of time, money, and travel; voice technology could eliminate these barriers.”

Klick Labs’ latest clinical study demonstrates that the fusion of AI and a mere 10-second vocal sample could significantly alter how diabetes screenings are carried out, offering more accessibility and reducing costs compared to existing methods. This research, also appearing in Mayo Clinic Proceedings: Digital Health, reported a predictive accuracy of 89% for women and 86% for men based on the acoustic attributes of the voice.

Klick Labs’ team delved into multiple vocal features such as alterations in pitch and volume that are imperceptible to the human auditory system. Through advanced signal processing techniques, the researchers were able to discern voice changes attributable to Type 2 diabetes. Interestingly, Kaufman noted that these vocal changes appeared differently in men and women.

A Promising New Instrument for Identifying Undiagnosed Diabetes Cases

According to the International Diabetes Federation, nearly half of the 240 million adults worldwide who have diabetes are not aware they suffer from the condition, and approximately 90% of these cases are of Type 2 diabetes. Conventional diagnostic methods such as the glycated hemoglobin (A1C), fasting blood glucose (FBG) tests, and the oral glucose tolerance test (OGTT) all involve visits to healthcare providers.

Yan Fossat, the Vice President of Klick Labs and the chief investigator of this study, emphasized that Klick’s unobtrusive and accessible method holds the potential to facilitate large-scale screenings and assist in identifying the sizable proportion of people who are unaware that they have Type 2 diabetes.

Fossat stated that the next course of action will be to reproduce the study and broaden the scope of the research to include voice-based diagnostics in other health conditions, such as prediabetes, women’s health issues, and hypertension.

Reference: “Acoustic Analysis and Prediction of Type 2 Diabetes Mellitus Using Smartphone-Recorded Voice Segments” by Jaycee M. Kaufman, Anirudh Thommandram, and Yan Fossat, published on 17 October 2023 in Mayo Clinic Proceedings: Digital Health.
DOI: 10.1016/j.mcpdig.2023.08.005

Frequently Asked Questions (FAQs) about Type 2 diabetes detection using AI and voice analysis

More about Type 2 diabetes detection using AI and voice analysis

• Klick Labs Official Website
• Mayo Clinic Proceedings: Digital Health
• International Diabetes Federation Statistics
• Glycated Hemoglobin (A1C) Test Information
• Fasting Blood Glucose Test Information
• Oral Glucose Tolerance Test (OGTT) Information
• Machine Learning in Healthcare
• Signal Processing Techniques
• Acoustic Features in Speech