Fighting Sugar Disease with Software
AI Recognises Diabetes by Voice in 10 Seconds
The voice is the key to recognising diabetes: a new AI tool can reliably diagnose type 2 diabetes based on 10 seconds of spoken text. The voice technology, which is easily accessible to patients, could enable diabetes to be recognised and treated much earlier.
Speech technology and artificial intelligence can revolutionise the diagnosis of diabetes and other indications.
A research paper by KlickLabs published in "Mayo Clinic Proceedings: Digital Health" shows that just six to ten seconds of voice recording and basic health data such as age, gender, height and weight are enough to determine whether a person has type 2 diabetes using an AI model. The accuracy of the model is 89 per cent for women and 86 per cent for men.
Voice Reveals Diabetes Diagnosis
»Our research shows significant vocal differences between people with and without type 2 diabetes," said scientist and first author Jaycee Kaufman. "Currently, diagnosing diabetes still requires a lot of time, money and frequent visits to the doctor. Voice technology has the potential to remove these barriers and redefine diabetes diagnosis.«
For the study, KlickLabs researchers asked 267 people who were either non-diabetic or had type 2 diabetes to record a sentence into their smartphone six times a day for a fortnight. From over 18,000 recordings, the scientists analysed 14 acoustic features to identify differences between non-diabetic and type 2 diabetic individuals - including changes in pitch and intensity that cannot be perceived by the human ear. Using signal processing, the scientists were able to detect changes in the voice due to type 2 diabetes. Surprisingly, these vocal changes manifested themselves differently in men and women.
Half of Diabetes Sufferers Don't Know
Almost half of the 240 million adults with diabetes worldwide do not know they have the disease, and almost 90 per cent of cases are type 2 diabetes. In most cases, the diagnosis of prediabetes and type 2 diabetes involves a haemoglobin A1C test, fasting blood glucose test and an oral glucose tolerance test (OGTT) - all of which require a visit to a doctor.
The KlickLabs team believes that the non-invasive and easily accessible voice test for detection has the potential to screen many people for diabetes and thus identify a large proportion of previously unrecognised patients with type 2 diabetes.
Voice for Simple and Affordable Diagnostics
»Our research emphasises the enormous potential of speech technology to identify type 2 diabetes and other health conditions,« says Yan Fossat, Vice President KlickLabs. »Speech technology could revolutionise diagnostics for many indications as an accessible and affordable digital screening tool.«
The next step is to replicate the study and expand the research to other areas such as prediabetes, women's health and hypertension. (uh)
