AI Smartwatch Detects Signs of Hypertension Without Cuff Calibration
A study published in Nature shows wearable wrist sensors can match the sensitivity of routine clinical checks.
Researchers have developed a way to screen for high blood pressure using wearable wrist sensors, matching the sensitivity of conventional clinic check-ups. The method, tested on a consumer device, applies an artificial intelligence model to continuous optical and motion data without requiring calibration against an inflatable arm cuff. The findings, published in the journal Nature, offer a practical path toward opportunistic screening for cardiovascular risk.
Optical sensors and motion tracking without calibration
More than half of adults with hypertension remain unaware of their condition, making silent cardiovascular damage a common clinical problem. While wearable gadgets often monitor heart rate, wrist sensors struggle to measure absolute arterial pressure directly because wrist anatomy varies widely. Lead researcher Ming-Zher Poh and colleagues addressed this limitation by designing an artificial intelligence algorithm that estimates the probability of hypertension rather than calculating exact blood pressure numbers, as reported by MedicalXpress.
This health monitoring innovation operates on a Google smartwatch and relies on photoplethysmography, an optical technique that measures blood volume changes in the microvascular tissue beneath the skin. The watch pairs these optical pulses with data from an integrated accelerometer to track wrist motion. By combining multiple sensor readings over time with the participant's height, the software estimates high blood pressure risk without needing periodic calibration against an arm cuff.
Tested against gold standard clinical monitoring
Developing the system required training on nearly 500,000 hours of recorded sensor data collected from 2,109 people, followed by evaluation across independent cohorts, according to MedicalXpress. To determine how well the algorithm performed in real clinical conditions, the authors validated the invention in a group of 196 participants.
Rather than relying on single spot checks, the study evaluated both the smartwatch and standard office cuff measurements against 24-hour ambulatory blood pressure monitoring, the gold standard in cardiology. In this comparison, the smartwatch achieved a sensitivity of 65.8%, correctly detecting slightly more cases of hypertension than standard office measurements, which recorded a sensitivity of 55.3%. Both approaches shared an identical specificity of 90%, meaning they avoided false positive readings at equal rates. The difference in detection rates between the watch and clinic cuffs remained within standard statistical margins.
A screening tool rather than a diagnostic replacement
The research team, led in print by first author Paolo Di Achille, emphasized that the wrist-based algorithm is not meant to replace conventional diagnostic cuffs or continuous ambulatory sleeves. Instead, the software functions as an early warning system designed to catch signs of hypertension opportunistically during everyday activities. Early flags could prompt patients to visit a physician for formal evaluation, helping doctors initiate treatment before arterial stiffening or cardiac strain worsens, as noted by MedicalXpress.
Future implementation will depend on how health authorities evaluate software acting as a specialized medical device feature on off-the-shelf hardware, alongside patent considerations for commercial rollout. The paper notes that obtaining reliable assessments requires collecting multiple readings over extended wear periods, ensuring that random movements do not distort risk assessments.
Read also on The Patent: Engineers Solve Key Physical Hurdle in Exhaled Breath Diagnostics
