Thanks to smartphones, diabetes can now be diagnosed easily and quickly. Differences in voice recordings on a smartphone app have been shown to help identify which people have type 2 diabetes.
A recent study involving experts from Ontario Tech University in Canada analyzed the potential of speech analysis as a pre-screening or monitoring tool for type 2 diabetes mellitus. The results can be read in the English-language journal “Mayo Clinic Proceedings”.
What does the voice say about the risk of diabetes?
In the future, recordings of one’s own speech on the telephone could be sufficient to determine which people have diabetes. People with type 2 diabetes have been shown to experience vocal changes compared to people who are not affected by the disease.
An AI is able to identify these differences when evaluating the voice recordings. Voice analysis could serve as a pre-screening or monitoring tool for type 2 diabetes, especially in combination with other risk factors associated with the disease, the experts report.
Speech was recorded up to six times daily
The team examined a total of 267 participants, including 79 women and 113 men without diabetes, and 18 women and 57 men who suffered from type 2 diabetes. They were instructed to use a smartphone app to record a fixed phrase they spoke up to six times a day for two weeks. A total of 18,465 images were created.
With the help of an AI, the experts extracted fourteen different acoustic features from these recordings, which indicate whether a person has diabetes or not. The team wanted to develop a new method for predicting type 2 diabetes.
Significant differences in the voice recordings identified
It was found that there are actually significant differences between the voice recordings of healthy men and women and people suffering from diabetes. The highest prediction accuracy in women was achieved by pitch. Whereas for men, the highest prediction accuracy was related to the intensity of the voice.
The results obtained by the team highlight the potential of speech analysis as an accessible and cost-effective screening tool. According to the researchers, incorporating voice analysis into the diagnosis of diabetes could make it possible to improve early intervention and treatment of type 2 diabetes in the future. (as)
