Through Voice Analysis: New Technology Detects Diabetes in Seconds
A recent study has shown that an artificial intelligence-based tool could help detect indicators of type 2 diabetes by analyzing a voice recording no longer than 20 seconds.
Researchers from the technology company Themia, in collaboration with Australia’s RMIT University, developed a model to analyze voice characteristics associated with diabetes, including hoarseness, roughness, and difficulty controlling breathing during speech, according to the Daily Mail.
The model was trained using more than 63,000 voice samples from approximately 21,000 individuals in the UK and the US, before being tested on 20-second recordings of people reading stories.
The first phase of testing involved 7,319 participants in the UK, with the model assigning higher risk scores to individuals who reported having type 2 diabetes in 80% of cases.
In a second analysis involving 801 participants who underwent home blood tests within three months of recording their voices, the model assigned higher risk scores to those with diabetes in 75% of cases.
The model performed well across different age and gender groups, but its performance declined among other participants, particularly those of African descent, due to their underrepresentation in the sample.
Performance also decreased among individuals with heart disease, high blood pressure, or obesity, conditions that may be associated with similar vocal changes.
Diagnosis of type 2 diabetes currently relies on blood tests. Gedri Cibucaite, a researcher at Themia, stated that analyzing a voice sample via a phone call or app could provide access to individuals who do not undergo regular health screenings.
The researcher emphasized that the model is not a substitute for blood testing, noting that the next step involves testing it in clinical settings and evaluating its effectiveness across different populations.
Researchers concluded that voice analysis could offer a non-invasive and rapid method to help identify individuals at higher risk of developing diabetes, paving the way for necessary blood tests to confirm the diagnosis.
