Voice to Diabetes tool leverages vocal biomarker data to diagnose Type 2 diabetes based on the analysis of a person's voice. This innovation harnesses AI and machine learning techniques to detect subtle vocal changes that are imperceptible to the human ear. By incorporating features like pitch and intensity changes, the tool successfully distinguishes between individuals with and without Type 2 diabetes.

India is home to the second-largest number of diabetes patients in the world. Despite the staggering number of diabetes cases, more than half remain undiagnosed due to various factors including distance, cost, and lack of resources. India’s vast geographical expanse and uneven distribution of healthcare facilities make accessing diagnostic tests challenging for millions, especially in rural areas. The cost of healthcare, including blood tests, can be prohibitive for many Indians, leading to delayed or neglected health concerns. However, over a billion people living in India have mobile phones. With Voice to Diabetes, the widespread use of mobile phones across socioeconomic strata can now be leveraged to address healthcare disparities and save lives.

India ranks second globally in diabetes cases, with over half remaining undiagnosed due to barriers like distance, cost, and lack of resources. Leveraging the widespread use of mobile phones among over a billion Indian citizens, Voice to Diabetes addresses healthcare disparities and gender equality concerns. This innovative AI tool screens for Type 2 diabetes through voice analysis, offering a non-invasive alternative to traditional blood tests. The study, featured in a Mayo Clinic journal, involved 267 Indian participants. Using a smartphone app, participants recorded voice clips over two weeks, leading to the identification of 14 key acoustic characteristics. For women, pitch changes were vital, while for men, variations in voice strength mattered. Achieving up to 89% accuracy for women and 86% for men, this approach promises a cost-effective and accessible diabetes screening solution, potentially transforming healthcare access in India.

India ranks second globally in diabetes cases, with over half remaining undiagnosed due to barriers like distance, cost, and lack of resources. Leveraging the widespread use of mobile phones among over a billion Indians, Voice to Diabetes addresses healthcare disparities and gender equality concerns. This innovative AI tool screens for Type 2 diabetes through voice analysis, offering a non-invasive alternative to traditional blood tests. The study, featured in a Mayo Clinic journal, involved 267 Indian participants. Using a smartphone app, participants recorded voice clips over two weeks, leading to the identification of 14 key acoustic characteristics. For women, pitch changes were vital, while for men, variations in voice strength mattered. Achieving up to 89% accuracy for women and 86% for men, this approach promises a cost-effective and accessible diabetes screening solution, potentially transforming healthcare access in India.

Study design and idea came in fall 2022 and was completed Dec 12, 2020

Ethics approval for the study was granted on May 07, 2021

Data collection part of the study completed July 8 2022

Analysis completed and patent filed sept 11,2023. Systems, devices and methods for predicting diabetic status using voice Patent 18/244,400 filed Sept 11, 2023 Filed for US and PCT Foreign Filing License Granted: 09/18/2023 Projected Publication Date: 03/13/2025

Peer review publication published Oct 17, 2023 https://www.mcpdigitalhealth.org/article/S2949-7612(23)00073-1/fulltext

Voice to Diabetes is an innovative tool that leverages AI to diagnose Type 2 diabetes based on the analysis of a person's voice. The system uses an app designed to identify diabetes by speaking into the phone. It displays a sentence on screen for the user to read out loud. The sentence is “Hello how are you, what is my glucose level right now?” The app then runs an analysis of the acoustic properties of the voice sample using a predictive model that was pre-trained on thousands of voice recordings to detect features that are associated with diabetes. The system doesn't comprehend what people say or what accent they have; it listens to lower-level features, such as minute amplitude or frequency modulation of the fundamental frequency of the voice. This analysis takes a few seconds. Once completed, the system shows a risk score of this person having Type 2 diabetes. Voice to Diabetes offers higher accuracy and personalized results compared to existing costly solutions.

Voice to Diabetes is already changing the diagnostic landscape as we know it. The innovation has the potential to help the 1 in 2 adults, or 240 million adults, living with diabetes worldwide unaware they have the condition. This innovation would advance T2D diagnostic tests which include the glycated hemoglobin (A1C), along with the fasting blood glucose (FBG) test and the OGTT–all of which require a trip to a healthcare provider. Long-term outcomes include the potential to save $32.75 billion globally, hospital partnerships, and results in 7 other disease states coming soon.

Voice to Diabetes generated over 4 billion earned press impressions, including more than 200 broadcast segments. It was covered in more than 65 countries worldwide, including the U.S. Canada, the UK, Japan, China, Germany, France, and Brazil.