Scientists have created an AI tool that uses voice analysis to identify Type 2 diabetes with remarkable precision.
Type 2 diabetes (T2D) affects millions worldwide and remains undiagnosed in many individuals. Traditional screening methods are not overly complex, but they involve invasive blood tests which can be a deterrent to some people. These tests are also difficult to scale in some parts of the world. Meanwhile, the prevalence of diabetes continues to soar, with estimates predicting 783 million cases globally by 2045 — or 1 in 7 adults. To avoid a massive global crisis, we need new, easier way to diagnose diabetes. And that’s exactly what researchers from Luxembourg have developed.
Researchers have discovered an innovative method for diagnosing diabetes through voice samples. The Colive Voice study, carried out by a global team, investigates how slight variations in vocal traits can accurately detect Type 2 Diabetes. If this approach can be widely implemented, it may revolutionize the field of medical diagnostics.
View pictures in App save up to 80% data.
Understanding diabetes
The link can be found in how diabetes impacts your respiratory system, neuromuscular function, and overall well-being. Factors such as inadequate glycemic management and peripheral neuropathy can negatively affect the small muscles and nerve pathways that play a role in voice production. Consequently, individuals with diabetes may experience nuanced vocal alterations, including hoarseness, diminished pitch consistency, or modified breathing patterns while speaking.
These alterations manifest in characteristics such as jitter, shimmer, and harmonic ratios. By analyzing these vocal metrics alongside data from healthy individuals, algorithms are able to identify patterns that may signify the presence of diabetes.
In a recent study, scientists examined voice recordings from 607 adults across the United States by employing a mix of conventional voice characteristics and advanced machine learning techniques. The participants were asked to read a standardized passage, which allowed the algorithm to detect nuanced variations in their vocal patterns.
How effective is it?
The algorithm predicted diabetes for 71% of male and 66% of female participants. It performed even better for people over 60 years old and suffering from hypertension.
“This research represents a major step in diabetes care. By combining AI with digital phenotyping, we are ushering in a more inclusive and cost-effective approach to early diagnosis and prevention. The ability to screen for diabetes using a simple voice recording could dramatically improve healthcare accessibility for millions of people around the world,” said Dr. Guy Fagherazzi
While the results are promising, there are hurdles to overcome before widespread adoption. The study’s reliance on English-speaking participants limits its generalizability to diverse populations. Expanding the dataset to include other languages and cultural contexts is crucial.
As diabetes continues to challenge public health systems worldwide, embracing such advancements could redefine the landscape of preventive care, ensuring timely treatment and better outcomes for millions.
The study “A voice-based algorithm can predict type 2 diabetes status in USA adults: Findings from the Colive Voice study” has been published in the journal PLoS Digital Health.
