Google AI research suggests consumer smartwatches can detect hypertension
Google has published findings in Nature detailing an AI system that utilizes smartwatch sensors to detect hypertension without traditional cuff-based monitoring.
Google has developed a multimodal AI system, dubbed AI-PPG-ACC-HTN, designed to monitor for signs of hypertension using standard consumer-grade smartwatches. The research, published in the journal Nature, shifts focus from isolated, noise-prone blood pressure snapshots to the aggregation of longitudinal physiological data collected over several days.
To build the system, Google used a 1D-convolutional neural network pre-trained on 490,000 hours of data to identify cardiovascular patterns. By combining motion and optical heart rate data with user height to estimate arterial path length, the model calculates a probability score for persistent high blood pressure.
A prospective clinical trial involving 196 participants found that the system performed with higher sensitivity and specificity than traditional single-point clinic screenings. The research findings serve as the core technical support for blood pressure trend monitoring features currently integrated into Google's health applications.