API
< Back to all clusters
[TECHNOLOGY] · United States · 3 sources

started · updated

Google Research develops AI to detect hypertension via smartwatch

Google researchers have developed an AI system, named AI-PPG-ACC-HTN, capable of detecting signs of hypertension using standard consumer smartwatch sensors. Unlike many existing cuffless wearables that require regular calibration with a traditional blood pressure cuff, this system operates without cuff calibration by utilizing photoplethysmography (PPG) and accelerometry (ACC) sensors.

The system functions in two stages: a deep convolutional network trained on nearly 500,000 hours of unlabeled sensor data, followed by a regularized logistic regression that incorporates participant height to estimate the probability of hypertension. The model does not provide specific systolic or diastolic blood pressure numbers, but rather identifies whether a user is persistently hypertensive over several days.

In a prospective multicenter study involving 196 participants in the United States, the AI demonstrated a sensitivity of 65.8 percent and a specificity of 90 percent for primary screening. This performance exceeded the sensitivity of a single office measurement, which was 55.3 percent. In a confirmatory test of a subset of 48 participants with elevated baseline pressure, the system reached 78.4 percent sensitivity and a 96.7 percent positive predictive value.

Entities

Google Research · Ming-Zher Poh · Nature · Paolo Di Achille