Vocal Biomarkers to Diagnose Diabetes

Researchers at the Luxembourg Institute of Health have developed a voice-based algorithm with the help of artificial intelligence that can potentially detect Type 2 diabetes. The screening analyzes subtle changes in a person’s voice, which are not usually noticeable to the human ear. This could represent an early and noninvasive diagnostic tool. Diabetes can affect the vocal cords as a result of nerve damage, thereby affecting voice quality.

The technology was tested in a study of 600 U.S. participants published in PLOS Digital Health. Researchers deemed the accuracy of the algorithm to be comparable to traditional risk assessment tools recommended by the American Diabetes Association. Next steps include refining the algorithm to detect pre-diabetes and expanding its use in other languages.

The Luxembourg researchers estimate there are 400 million undiagnosed cases of Type 2 diabetes worldwide. Without treatment, this disease can lead to serious health issues such as cardiovascular disease and neuropathy, as well as higher healthcare costs and even mortality. Current screening relies on blood tests