Predicting Diabetes with Voice
An artificial intelligence (AI) developed by Canadian medical researchers has been programmed to accurately predict Type 2 diabetes based on six to ten seconds of the patient’s spoken voice.
This resulted from the model distinguishing fourteen acoustic characteristics that differentiate individuals without diabetes from those with Type 2 diabetes.
The AI analysed vocal characteristics and patient health data like age, gender, size, and mass. Such as minute fluctuations in pitch and intensity imperceptible to human hearing, including physicians.
Vocal Characteristics and Gender
Researchers showed that gender mattered: the AI diagnosed the condition 89% accurately for women and 86% for men.
The Potential of Remote Diagnosis
The AI model could significantly reduce the cost of in-person evaluation for average chronic health patients.
“Don’t miss out! Grab your free shares of Webull UK today!”
The potential benefits of remote, automated diagnosis are considerable, according to data gathered by the International Diabetes Federation. The organization estimates that 240 million adults, nearly half of those with diabetes, are unaware of their condition.
“Our research reveals significant vocal differences between individuals with and without Type 2 diabetes,” said Jaycee Kaufman, a research scientist at Klick Labs, which intends to market the software and the paper’s first author.
Kaufman aspires for the organization’s artificial intelligence to “radically revolutionize the way the medical community conducts diabetes screenings.”
Blood work and other expensive in-person diagnostic procedures were previously necessary for the detection of prediabetes and type 2 diabetes.
A1C, FBG, and OGTT are common diagnostic tests that require patients to visit their doctor.
Current methods of detection can be time-consuming, expensive, and travel-intensive,” Kaufman said in a statement that accompanied the publication of the new study in Mayo Clinic Proceedings: Digital Health on Tuesday.
She stated, “Voice technology has the capacity to completely eliminate these obstacles.”
The Study and Its Findings
Klick Labs scientists and Ontario Tech University faculty educated the AI using recordings from 267 Indian test volunteers.
The proportion of individuals with a prior diagnosis of nondiabetic status in the control group was approximately 72% (79 women and 113 men), whereas 18 women and 57 men had a prior diagnosis of Type 2 diabetes.
The Klick Labs and American Diabetes Association criteria guided the hiring process.
For two weeks, the 267 participants were instructed to record a phrase six times daily on their mobile phones.
Klick’s researchers identified 14 acoustic features from the 18,000 individual recordings that resulted from the study in an effort to identify repeatable, consistent differences between the groups with and without type 2 diabetes.
Only four auditory features accurately predicted diabetes or not.
The features ‘pitch’ and ‘standard deviation from pitch’ demonstrated utility in predicting the disease for both male and female patients. However, in the case of women, ‘relative average perturbation jitter’ proved to be a more effective predictor. Men were alerted to auditory characteristics referred to as “intensity” and “11-point amplitude perturbation quotient shimmer.”
Kaufman at Klick labs characterized the sex-based distinctions discovered through the AI’s signal processing as “surprising.”
In a peer-reviewed publication, the researchers found that incorporating the voice recorder’s age and BMI into their prediction model boosted the firm’s AI.
“Our research highlights the immense potential of voice technology in detecting health conditions such as Type 2 diabetes,” says Yan Fossat, principal investigator of the new study and vice president of Klick Labs.
Fossat is an adjunct professor in the Faculty of Science at Ontario Tech University. Where he specializes in computational science for digital health and mathematical modeling.
Future of Voice Technology in Healthcare
The professor expressed his optimism that Klick’s accessible and non-intrusive AI approach, which could lead to diagnoses performed via a simple phone app, will assist in identifying and aiding the millions of undiagnosed individuals who suffer silently from Type 2 diabetes.
Fossat continued, “As an affordable and easily accessible digital screening tool, voice technology could revolutionize healthcare practices.”
He stated that the subsequent actions will consist of attempting to replicate the new study.
Fossat conveyed optimism regarding the potential future expansion of Klick Labs’ voice-diagnosing research to encompass additional medical domains, including hypertension, prediabetes, and women’s health.
