AI can spot type 2 diabetes from subtle vocal clues in seconds

Type 2 diabetes is a chronic condition that affects millions of people worldwide. It occurs when the body does not produce enough insulin or use it effectively, leading to high blood sugar levels. If left untreated, type 2 diabetes can cause serious complications such as heart disease, kidney failure, nerve damage, and vision loss.

Diagnosing type 2 diabetes usually requires blood tests that measure the amount of glucose in the blood. However, these tests can be costly, time-consuming, and inconvenient for some people, especially those who live in remote areas or have limited access to health care facilities.

But what if there was a simpler and faster way to screen for type 2 diabetes using just your voice?

That is the idea behind a new study by Canadian researchers from Klick Applied Sciences and Ontario Tech University, who have developed an artificial intelligence (AI) model that can diagnose type 2 diabetes from subtle vocal clues in just seconds.

How does the AI model work?

The AI model works by analyzing six to 10 seconds of voice recordings from people who either have or do not have type 2 diabetes. The recordings are made using a smartphone app that prompts the users to say a fixed phrase.

The AI model then uses signal processing techniques to detect changes in pitch and intensity of the voice that are caused by type 2 diabetes. These changes are too subtle for human ears to perceive, but they can reveal important information about the health status of the speaker.

The researchers found that there were 14 vocal features that differed significantly between people with and without type 2 diabetes. However, four of these features were more relevant for diagnosing the condition. For women, these features were related to pitch, while for men, they were related to intensity or amplitude.

The AI model also takes into account other basic health data from the users, such as age, sex, height, and weight. By combining these factors, the AI model can achieve an accuracy of up to 89% for women and 86% for men in diagnosing type 2 diabetes.

Why is this study important?

The study is important because it shows the potential of using voice as a biomarker for type 2 diabetes. Voice is a convenient and non-invasive way to screen for the condition, as it does not require any blood samples or specialized equipment. Voice analysis can also be done remotely and automatically, making it more accessible and affordable for people who need it.

The researchers hope that their AI model will eventually lead to a simple, in-home test for type 2 diabetes that can help people monitor their health and seek medical attention if needed. They also believe that their AI model can be applied to other diseases that affect the voice, such as Parkinson’s disease, Alzheimer’s disease, and COVID-19.

The study is published in Mayo Clinic Proceedings: Digital Health.