Researchers Built A Smartwatch That Could Help Millions Detect Dangerous Blood Pressure Spikes
When you have your blood pressure taken, a nurse wraps a cuff around your upper arm, pumps it full of air so it squeezes you tight, and waits to take a reading. It requires patients to sit still, can be uncomfortable, and only captures a single snapshot in time.
For the roughly half of American adults living with hypertension, that snapshot likely doesn't tell the whole story.
Now, researchers have developed a smartwatch that can monitor blood pressure continuously, without a cuff, and without the limitations that have plagued previous attempts at wearables for blood pressure. The technology, detailed in a new study published in Nature Communications1, uses electric sensing paired with a physics-informed AI model to deliver real-time readings directly from your wrist.
Why cuffless monitoring matters
Blood pressure isn't static. It fluctuates throughout the day in response to stress, exercise, sleep, and dozens of other factors. Traditional cuff-based monitors, whether at the doctor's office or at home, require you to sit still in a controlled environment to measure blood pressure. That means the dangerous spikes that can accompany acute cardiovascular events, or telling patterns that emerge during daily activity, often go undetected.
According to the study, existing ambulatory and at-home blood pressure devices are "obtrusive and impractical for preventive and long-term use due to discomfort." And because they must be used under near-rest conditions, critical blood pressure information during dynamic states, like during exercise or a cardiovascular episode, can be missed entirely.
Wearable devices have tried to fill this gap, but most existing approaches rely on methods that lack strong scientific foundations and are vulnerable to interference from the body's own signals. This new smartwatch takes a fundamentally different approach.
How the smartwatch works
The device is built around a technology called electrical bioimpedance, or BioZ, which is often used in wearables and smart scales. The technology allows small metal electrodes embedded in the watch band to send a tiny, harmless electrical current through the wrist and measure how that current changes as blood pulses with each heartbeat.
This works because blood conducts electricity, and its conductivity changes depending on how much blood is present, how fast it's moving, and how the red blood cells are positioned within the vessel. By continuously measuring tiny fluctuations in electrical resistance at the wrist, the device can detect changes in blood volume and conductivity that correspond with blood pressure.
The researchers built a detailed physical model of the human arm to establish exactly how blood pressure maps to these electrical signals, accounting for how blood flows through branching arteries, how arterial walls stretch and recoil, and even how red blood cells orient and shift under the stress of flowing blood. This physical foundation is what sets this smartwatch apart from prior wearable approaches to tracking blood pressure.
The AI that makes it work
Capturing the raw electrical signal is only half the challenge. Translating that signal into an accurate blood pressure reading requires a sophisticated model, and that's where the study's other major innovation comes in.
The team developed what they call a signal-tagged physics-informed neural network, or sPINN. Unlike conventional AI models that learn purely from data, the sPINN is guided by the same principles used to describe how liquids move through pipes and vessels. By embedding these physical laws directly into the AI, the model produces predictions that are both accurate and physiologically plausible.
In practical terms, this means the sPINN is less likely to produce readings that don't make biological sense. When compared against traditional AI models, the sPINN achieved the lowest gap between predicted and true blood pressure readings. Its blood pressure predictions were smoother and more stable than those from conventional AI, which matters clinically because a reading that jumps around unpredictably isn't useful for monitoring someone's health.
Who was tested & what the results showed
The researchers tested the system across three groups of participants. The first group consisted of 75 healthy individuals who wore the device before and after walking, running, and cycling, as well as during autonomic challenges designed to stress the cardiovascular system.
The second group included 85 patients seen in outpatient settings, with cohorts of patients with hypertension (both controlled and uncontrolled), cardiovascular disease, and other conditions. A third small group of three patients was evaluated in an intensive care unit setting.
For the population-wide model, the sPINN achieved strong accuracy scores for both the top number (systolic) and bottom number (diastolic) in a blood pressure reading, compared with continuous reference blood pressure measurements. Performance was even stronger within specific groups who need this technology most. Patients with hypertension and those with cardiovascular disease both showed higher accuracy scores than the general population model.
When the model was fine-tuned to individual subjects rather than trained on the population as a whole, performance improved even more across all patient groups.
What still needs to happen
While the results are promising, this is still early-stage technology, and several hurdles remain before you will spot this smartwatch on a consumer wrist.
First, a formal clinical study following established validation protocols is needed to evaluate real-world performance of the smartwatch at the bedside and in home settings. Second, the device needs to be tested in larger and more diverse populations, including people with uncontrolled, periodic, or positional hypertension (where blood pressure drops when standing up). For context, the study notes that Apple's Hypertension Notification Feature required data from more than 86,000 participants for its training process.
Third, the long-term stability of the AI model needs further investigation. In its current form, the sPINN required daily recalibration to maintain performance over time, reflecting the fact that factors like hydration status, ambient temperature, weight changes, and changes in blood vessel structure can shift the baseline electrical signal in ways the model hasn't yet learned to handle.
The takeaway
Cardiovascular disease remains the leading cause of death worldwide, and a significant portion of that burden is tied to undetected or poorly managed hypertension. A smartwatch that can continuously and accurately monitor blood pressure without a cuff, during exercise, stress, and daily life, would represent a huge shift in how people manage their heart health.
This study establishes the foundations for that possibility. The technology isn't ready for your wrist just yet, but the science behind it is more rigorous than anything that's come before it in this space. For anyone worried about hypertension, or who simply tracks their health closely, this development is worth paying attention to.