Artificial intelligence and health: diabetes screening by voice in Luxembourg
Mesut çiçen, Unsplash
An innovative project from the Luxembourg Institute of Health (LIH) proposes the use of voice analysis for the early detection of type 2 diabetes. This technology, developed by a team led by Dr Guy Fagherazzi, makes it possible to identify changes in the voice characteristic of patients with the disease.
According to Dr Fagherazzi, the voice of people with diabetes can become more hoarse due to the effects of hyperglycaemia, fatigue and nerve damage. These changes are imperceptible to human hearing, but artificial intelligence can recognise them.
The project uses data from the international Colive Voice study, which collected over 8,000 voice samples in different languages. A sample of 600 recordings in English was chosen to analyse type 2 diabetes. Participants read a text lasting 20-30 seconds, after which the algorithm assessed the presence of characteristic signs of the disease.
As reported by colleagues at RTL, the AI system has demonstrated the ability to distinguish the voices of diabetic patients from healthy people with an accuracy of about 75 per cent. While this is only a preliminary result, the researchers aim to improve the accuracy of the analysis by increasing the database of voice recordings.
Dr Fagherazzi stressed that the technology will not replace traditional diagnostic methods such as blood tests, but will be used as a pre-screening to identify at-risk groups.
To improve the algorithm, the LIH team continues to collect voice samples and invites volunteers to participate through the www.colivevoice.org platform. With around 800 million people worldwide with diabetes, half of whom are unaware of their condition, such tools could be a meaningful step towards more accessible diagnosis.
