Voice’s Disease Detection Capabilities Expand
Numerous studies have already confirmed voice biomarker technology’s ability to detect certain respiratory, neurological, and mental health conditions, including asthma, depression, Parkinson’s disease, and Alzheimer’s disease. New research has expanded the use cases and the types of medical conditions that the technology can address. The most recent studies covered type 2 diabetes and high blood pressure. Another study found that clinical follow-up using virtual voice technology helped identify complications after transcatheter aortic valve implantation (TAVI) with a high degree of patient satisfaction.
The high blood pressure study, conducted by Klick Labs, asked 245 participants to record their voices up to six times daily for two weeks by speaking into a proprietary mobile app. The app detected high blood pressure with accuracies up to 84 percent for females and 77 percent for males. It uses machine learning to analyze hundreds of vocal biomarkers, including the variability in pitch (fundamental frequency), the patterns in speech energy distribution (Mel-frequency cepstral coefficients), and the sharpness of sound changes (spectral contrast).
“By leveraging various classifiers and establishing gender-based predictive models, we discovered a more accessible way to detect hypertension, which we hope will lead to earlier intervention for this widespread global health issue,” said Yan Fossat, senior vice president of Klick Labs and principal investigator of the study.
Previous Klick Labs research found that the same technology could be used to also detect type 2 diabetes, and a similar study just conducted in Europe confirmed those findings.
This latest study used on average 25 seconds of people’s voices along with basic health data, including age, sex, body mass index (BMI), and hypertension status, to develop an AI model to distinguish whether they have type 2 diabetes, with 66 percent accuracy in women and 71 percent accuracy in men. The model performed even better in females aged 60 years or older and in individuals with hypertension. Additionally, the voice-based analysis agreed with the questionnaire-based American Diabetes Association’s risk score 93 percent of the time.
To identify people with diabetes, the voice AI algorithm analyzed vocal features like changes in pitch, intensity, and tone, using an advanced technique that captured up to 6,000 detailed vocal characteristics and a more sophisticated deep learning approach that focused on a refined set of 1,024 key features.
“While our findings are promising, further research and validation are necessary before the approach has the potential to become a first-line diabetes screening strategy and help reduce the number of people with undiagnosed type 2 diabetes. Our next steps are to specifically target early-stage type 2 diabetes cases and pre-diabetes,” said co-author Guy Fagherazzi, M.D., from the Luxembourg Institute of Health.
The project for TAVI patients, meanwhile, centered on a virtual assistant named “Lola.” Based on artificial intelligence and natural language processing, the Lola assistant could make more than 40 phone calls in two hours, allowing healthcare providers to gather follow-up information and act accordingly.
The TeleTAVI study, conducted at the Dr. Balmis General University Hospital of Alicante, Spain, allowed patients undergoing TAVI via the femoral artery in 2023 to follow up with the virtual voice assistant. Lola called the patients in week 1, week 2, month 1, month 3, and month 12 after patient discharge. In these calls, a series of questions were asked, mainly related to vascular access and the patient’s cardiovascular situation. After finishing the call, all the information collected was uploaded to a web platform where the data were monitored by healthcare professionals who acted where necessary.
Patients generally responded favorably to the system, giving it a satisfaction score of 4.68 on a five-point scale; and 89 percent of patients reported good or very good satisfaction. In total, 86 percent of patients said they would recommend the use of Lola to others.
“Voice technology has the potential to exponentially transform healthcare, making it more accessible and affordable, especially for large, underserved populations,” said Jaycee Kaufman, Klick Labs research scientist. “Our ongoing research increasingly demonstrates the significant promise of vocal biomarkers in detecting hypertension, diabetes, and a growing list of other health conditions.”