From the source
Google Research presented a multimodal AI system that uses standard consumer smartwatch sensors to passively detect signs of hypertension without cuff calibration.
- The system, called AI-PPG-ACC-HTN, was validated against 24-hour ambulatory blood pressure monitoring, the clinical gold standard, in a prospective multi-center study across five US sites with 196 participants.
- Rather than estimating point-in-time systolic and diastolic values, it aggregates PPG and accelerometer waveform embeddings over multiple days into a person-level embedding, combined with height, to predict a persistent hypertensive state.
- In initial screening the smartwatch reached 65.8% sensitivity, 90.0% specificity and 80.6% positive predictive value, against 55.3% sensitivity and 77.8% PPV for office-based screening.
- Among 48 participants with elevated office blood pressure, confirmatory testing reached 78.4% sensitivity, 90.9% specificity and 96.7% PPV, while return office measurements showed 67.6% sensitivity, 63.6% specificity and 86.2% PPV.
- The findings helped inform the blood pressure trends Health Guardian feature in the Google Health app.