Google Research Unveils Cuffless Smartwatch AI for Hypertension Screening in 196-Patient Study
Updated
Updated · Google Research · Nov 5
Google Research Unveils Cuffless Smartwatch AI for Hypertension Screening in 196-Patient Study
3 articles · Updated · Google Research · Nov 5
Summary
A Google Research system published in Nature used standard smartwatch PPG and accelerometer sensors to flag hypertension without any cuff calibration, aiming to detect persistent high blood pressure passively over several days.
In a five-site U.S. study of 196 participants, the model posted 65.8% sensitivity and 90.0% specificity against 24-hour ambulatory blood pressure monitoring, outperforming initial office screening on sensitivity and positive predictive value.
Among 48 people who first screened positive in clinic, the smartwatch reached 78.4% sensitivity, 90.9% specificity and 96.7% positive predictive value, with results competitive with multi-day home monitoring and stronger than repeat office readings.
The model was trained on 492,347 hours of unlabeled sensor data from 2,109 participants, aggregating 15-second waveform embeddings over days and combining them with height to estimate hypertension risk.
Google said the work is early feasibility research rather than a commercial diagnostic product, but it could support scalable screening for a condition affecting more than 1 billion people worldwide.