Researchers at the University of Rochester have developed an AI-driven speech-based screening tool designed to detect early signs of Parkinson’s disease, a rapidly growing neurological condition. Published in the journal npj Parkinson’s Disease, the web-based tool analyzes voice recordings from users reciting two pangrams, achieving nearly 86% accuracy in identifying potential symptoms. While traditional diagnoses rely on specialized neurologists and comprehensive assessments, this innovative tool offers a fast, accessible option for preliminary evaluation, especially in underserved areas.
The tool leverages advanced semi-supervised speech models trained on over 1,300 diverse participants, utilizing common speech interfaces like Amazon Alexa and Google Home to streamline the process. Recognizing that nearly 89% of individuals with Parkinson’s exhibit distinct vocal markers, this screening method aims to encourage further clinical evaluation. An interactive demo of the screening tests is available online. The initiative was supported by prominent funding sources, including the National Institutes of Health and Google.
