Pioneering research conducted by Klick Labs in the United States has unveiled an innovative method for identifying Type 2 diabetes in individuals through their voice. The approach is remarkably simple, requiring participants to record a specific phrase on their smartphones lasting only six to ten seconds. These voice recordings are then paired with basic health information such as age, gender, height, and weight. This data was used to construct an artificial intelligence (AI) model capable of discerning the presence of Type 2 diabetes.
The results of this study, published in the Mayo Clinic Proceedings: Digital Health journal, are quite remarkable. The AI model demonstrated an impressive 89% accuracy for women and 86% for men. To achieve these outcomes, the researchers enlisted the help of 267 individuals, both diagnosed as non-diabetic and having Type 2 diabetes. Over a two-week period, these participants were asked to record the specific phrase multiple times daily, generating more than 18,000 voice samples. The analysis of these recordings unveiled 14 distinct vocal features that allowed for the differentiation of non-diabetic and Type 2 diabetic individuals.
Jaycee Kaufman, the lead author of the study and a research scientist at Klick Labs, emphasized the significant vocal variations associated with Type 2 diabetes, opening up the potential for an entirely new approach to diabetes screening. Unlike current detection methods, which can be time-consuming and costly, using voice technology could eliminate these barriers.
The research delved into various vocal characteristics, including subtle changes in pitch and intensity that are undetectable to the human ear. By applying advanced signal processing techniques, the researchers were able to identify these voice alterations linked to Type 2 diabetes. Interestingly, these vocal changes manifested differently in males and females.
Globally, nearly half of the 240 million adults living with diabetes are unaware of their condition, and approximately 90% of these cases are Type 2 diabetes, as reported by the International Diabetes Federation. Common diagnostic tests for prediabetes and Type 2 diabetes, such as the glycated hemoglobin (A1C), fasting blood glucose (FBG) test, and oral glucose tolerance test (OGTT), are frequently utilized.
Yan Fossat, Vice President of Klick Labs and the principal investigator of this study, highlighted the potential of this non-intrusive and accessible approach for screening a large number of people and identifying undiagnosed cases of Type 2 diabetes. Fossat underlined that the research showcases the immense possibilities of voice technology in the identification of Type 2 diabetes and other health conditions. This development could revolutionize healthcare practices by providing an accessible and affordable digital screening tool.
