Google Research has developed an AI model capable of identifying signs of hypertension using standard smartwatch sensors. The system utilizes data from built-in pulse and motion sensors, eliminating the need for traditional blood pressure cuffs or manual calibration.
The neural network was trained on nearly 500,000 hours of sensor data. It identifies individual pulse and heartbeat patterns without requiring prior medical diagnoses. The model calculates hypertension probability by averaging these patterns over several days.
In a clinical study involving 196 participants across five US research sites, the AI demonstrated high efficacy. It identified 65.8 percent of individuals with high blood pressure, while traditional in-office measurements detected 55.3 percent.
The findings are integrated into the blood pressure trends feature within the Google Health app. However, the company emphasizes that the technology remains a research project and does not serve as a medical diagnosis.
