A team of researchers from Luxembourg has developed a new AI algorithm trained to diagnose Type 2 diabetes by listening to a person’s voice—a novel, noninvasive, and possibly cheap screening methodology that could change the lives of many.
This breakthrough was possible with the help of the Luxembourg Institute of Health, guided by Abir Elbeji and Dr. Guy Fagherazzi. They used high-end machine learning, which identified a subtle change of voice— a biomarker-characteristic of this pathology, which makes it distinct from other technologies currently in use.
According to these researchers, their AI reached the same predictive level as the American Diabetes Association’s risk score.
“This study represents a major leap forward in diabetes care,” said Fagherazzi. This could be, one day, a game-changer in the early detection of diabetes, and millions of people in the world will have better access to health thanks to only their voice.
The results, published in the open-access journal PLOS Digital Health through the Colive Voice program, are based on speech recordings by over 600 participants residing in the US.
Based on their impressive result, the researchers will work to further refine the algorithm to detect prediabetes and undiagnosed cases of Type 2 diabetes. They also plan to expand the program to other populations and languages.
