Royal Holloway and Bedford New College and DDM Health Limited KTP 24_25 R4

A smartphone app could soon detect Type 2 diabetes and prediabetes from a few seconds of a person's speech, without needles or blood tests. This matters because millions of people with early-stage Type 2 diabetes or prediabetes remain undiagnosed until complications develop. Current diagnostic methods require blood draws and lab visits, which many people avoid or cannot easily access. Voice analysis offers a non-invasive alternative that could be deployed at scale through devices people already carry. If this research succeeds, the system could transform how diabetes is caught early. A person might speak into their phone during a routine check-up or at a pharmacy kiosk, and receive an immediate risk assessment. This would shift screening from clinical settings into everyday environments, potentially catching cases years earlier than current practice allows. The technology could also be integrated into telehealth platforms, making regular monitoring feasible for people in remote areas or those who struggle to attend appointments. The project is applied research with a clear practical goal: building and validating an AI model that links specific vocal biomarkers—changes in pitch, tremor, or breathiness—to blood glucose levels. It does not explore fundamental mechanisms of how diabetes affects the voice, but focuses on creating a reliable diagnostic tool ready for real-world testing.