AI model identifies type 2 diabetes risk through voice analysis

Researchers from Thymia and RMIT University have developed an artificial intelligence model capable of identifying individuals at risk for type 2 diabetes through the analysis of short voice recordings. The technology focuses on subtle vocal changes, such as increased hoarseness, roughness, and difficulties in breath control during speech.

The AI model was trained using over 63,000 voice samples from more than 21,000 individuals in the United Kingdom and the United States. In testing involving 7,319 adults in the UK, the model correctly assigned higher risk scores to those reporting type 2 diabetes in approximately 80% of comparisons.

Further validation compared the AI's predictions against glycated hemoglobin (HbA1c) blood tests in a subgroup of 801 participants. The model identified higher risk in roughly 75% of those cases. The goal of this method is to serve as an initial screening tool via phone or app to help identify people who may require follow-up clinical blood testing, rather than providing a formal diagnosis or replacing traditional medical exams.