Overview

• University of Rochester’s online pangram screening test and demo went live July 24, offering rapid Parkinson’s risk assessments through any web browser.
• Participants recite two pangrams and the AI evaluates subtle vocal patterns linked to Parkinson’s disease in seconds.
• The underlying model leverages semi-supervised speech embeddings trained on millions of audio recordings and validated with data from over 1,300 participants to reach 85.7% accuracy.
• Researchers are exploring integration with Amazon Alexa and Google Home to embed the screening tool into popular voice assistant platforms for remote use.
• Developers emphasize that the AI test is a low-barrier preliminary flagging tool designed to prompt users to seek comprehensive clinical evaluations rather than replace medical diagnosis.