A Google Research AI model identifies signs of high blood pressure using standard smartwatch sensors without requiring a calibration cuff. The system, detailed in a Nature study, combines data from a pulse sensor and a motion sensor to analyse heart rate patterns across multiple days. A neural network processed nearly 500,000 hours of sensor data to recognise typical pulse signatures for individual users. Google integrated this method into the blood pressure trends feature within the Google Health app, though the tool remains a research project and does not replace medical diagnosis.
The approach offers a potential alternative to traditional in-office measurements for spotting undiagnosed hypertension. In a trial involving 196 participants across five US sites, the watch identified 65.8 per cent of cases with high blood pressure, outperforming the 55.3 per cent detection rate of standard cuff measurements. Google states the goal is to encourage people to seek professional checks, noting that more than half of those with high blood pressure are unaware of their condition. The findings rely on existing hardware rather than new medical devices.
- The study involved 196 people across five US research sites.
- Detection rates reached 65.8 per cent compared to 55.3 per cent for cuffs.
- Results are currently available only within the Google Health app.
