An artificial-intelligence system running on signals from a consumer smartwatch may be able to flag high blood pressure without first being calibrated against a conventional cuff reading. In testing described in a paper in Nature, the approach detected hypertension with sensitivity comparable to traditional office-based blood-pressure screening.
The result does not mean a smartwatch can directly measure blood pressure or replace clinical assessment. Instead, it points to a different use for wearables: identifying people whose sensor patterns suggest they should receive follow-up testing. That distinction matters because hypertension is a major cardiovascular risk factor, yet more than half of adults affected by it are unaware of the condition.
Screening from everyday signals
The research team, led by Ming-Zher Poh, used data collected by a Google smartwatch. The device combines photoplethysmography, an optical sensing technique that tracks changes in blood volume in tissue beneath the skin, with an accelerometer that measures wrist movement. Along with a participant's height, those inputs were processed by an AI model to produce an estimated probability that the person has hypertension.
The system is designed to work without cuff calibration. Existing wearable sensors are generally not capable of directly measuring blood pressure to the standard required for a conventional blood-pressure reading, and many approaches depend on comparison with cuff measurements. The new work instead estimates hypertension risk from patterns in the wearable data and requires multiple readings.
That makes it a screening approach rather than a miniature sphygmomanometer on the wrist. A person flagged by the system would still need a recognized blood-pressure assessment to establish whether hypertension is present and whether treatment is appropriate.
A large training set and ambulatory reference
The model was trained using nearly 500,000 hours of smartwatch data from 2,109 people, with further training and testing across several cohorts. For validation, the researchers compared the watch's estimate of hypertension in 196 participants with standard office blood-pressure measurements.
Crucially, the comparison was evaluated against 24-hour ambulatory blood-pressure monitoring. This method records blood pressure over the course of a day and night and was used in the study as the reference standard. The use of ambulatory monitoring addresses a limitation in earlier work, which had not always used gold-standard measurements as reference points or compared wearable assessments directly with established hypertension-screening methods.
According to the report, the smartwatch system's sensitivity was similar to that of office-based cuff screening. Sensitivity describes how effectively a test identifies people who have a condition. It is an important metric for a screening tool, where missing people who may need follow-up can undermine the value of a convenient first check.
Why convenience could matter
Blood pressure can vary with time, activity and setting. Office screening remains an established route to detection, but it depends on people having access to a measurement and attending an appointment. A wearable-based screen could potentially create many more opportunities to identify risk, particularly when a device is already being worn in daily life.
That potential is especially relevant for hypertension because it is often silent. People may feel well while elevated blood pressure contributes to their cardiovascular risk. A tool that prompts someone to seek a conventional measurement could therefore help uncover cases that might otherwise remain unrecognized.
The study also illustrates a broader shift in wearable health technology. The useful signal may not always be a direct physiological measurement displayed as a single number. In some cases, a combination of imperfect sensors, repeated observations and a trained model may provide a risk estimate that is useful enough to guide the next clinical step.
Important limits remain
The researchers' findings should be read as evidence for a specific AI-assisted screening method, not as proof that any smartwatch can diagnose hypertension. The reported system relied on particular optical and motion signals, participant information and an algorithm trained on a substantial dataset. Its output is an estimated probability of hypertension, not a substitute for the clinical measurements used to diagnose and manage the condition.
Multiple readings are also part of the approach. That requirement reflects the fact that the system is looking for a pattern across data rather than treating a single momentary sensor reading as a definitive answer.
Before such a capability could become a routine health feature, questions would remain around validation in additional populations, how users are informed about a positive screen and how follow-up care is handled. The study's validation against 24-hour ambulatory monitoring is a significant element, but real-world deployment would need to preserve the distinction between an alert and a diagnosis.
A possible role alongside, not instead of, care
The most plausible near-term role for the technology is as an additional doorway into established care. A smartwatch could identify a person who may benefit from formal testing; a cuff and clinical assessment would remain responsible for confirming the result. Used that way, the device's value would come from reach and repeatability rather than from claiming to replace medical equipment.
The work is also a reminder that consumer devices can become clinically relevant through analysis as much as through new hardware. The watch's optical sensor and accelerometer are familiar components. What changes their potential is the attempt to combine their signals with AI and validate the result against an accepted reference method.
For now, the study offers a promising screening result rather than a new standard of diagnosis. But if the findings hold across broader use, a consumer wearable could become one more way to spot a common, consequential condition before it goes unnoticed.
This article is based on reporting by Medical Xpress. Read the original article.
Originally published on medicalxpress.com
