Physics-Informed AI Powers a Cuffless, Calibration-Free Blood Pressure Smartwatch

Hypertension is a leading risk factor for heart attack and stroke, yet most people only see their blood pressure as two static numbers captured under clinical conditions. A study published May 14, 2026, in Nature Communications describes a smartwatch that reads the full story—capturing every pulse wave and blood velocity in real time without a cuff or any individual calibration .

The device, built by mathematicians and engineers at the University of Utah and the University of Illinois, Chicago, combines real‑time electrical bioimpedance (BioZ) sensing with a physics-informed neural network (PINN) . Here’s why that matters for the future of wearable health monitoring.

How Bioimpedance Sensing Captures Blood Flow

The smartwatch applies a high-frequency, imperceptible electrical current through the wrist. Because blood is conductive, the impedance changes as blood pulses through the radial artery with each heartbeat . Those tiny fluctuations encode information about pressure, which is why the researchers call BioZ the raw signal.

To move from impedance variations to a physical blood pressure reading, the team built a multiscale analytical and computational model that maps the biophysical link between BioZ and blood pressure . This includes accounting for physiological factors, anatomical positioning, and experimental parameters that influence the BioZ signal at the wrist.

Embedding Physics Inside the Neural Network

The core machine-learning component is a “signal-tagged physics-informed neural network” that bakes the laws of fluid dynamics into its architecture . Unlike a conventional black‑box deep learning model that learns correlations from data, a PINN cannot produce physically impossible outcomes, which the researchers argue makes it more trustworthy for clinical decision‑making .

Because the model already understands the physics of pulsating flow and electromagnetism, it can reconstruct the entire pressure waveform from the electrical signal alone—without requiring a cuff to provide a baseline. This is what makes the system truly calibration‑free.

What the Watch Measures That a Cuff Can’t

A traditional cuff gives you systolic and diastolic pressure at one instant. The Utah team’s smartwatch outputs the full continuous blood pressure waveform over time . In addition to standard pressure, the device estimates radial blood velocity and axial blood velocity—the speed blood moves through the artery .

Co-author and mathematician Braxton Osting framed the advance plainly: “Blood pressure isn’t two numbers; it’s a function of time. The mathematical challenge was recovering that whole waveform from indirect electrical measurements at the wrist” .

The result is a rich hemodynamic picture that could reveal dangerous transient spikes, nocturnal patterns, and masked hypertension that periodic office readings miss.

Testing Across Healthy Subjects and Patients

The smartwatch was evaluated on 150 participants in total, including healthy individuals at rest and after physical activity (walking, running, stair climbing) . Crucially, the study also included patients with hypertension and cardiovascular disease in both outpatient and intensive care settings. This directly addresses whether BioZ sensing works in the populations that need it most.

While exact numerical accuracy metrics from the 2026 study were not included in the available summaries, earlier PINN‑based work by the same team reported strong correlations with reference measurements (systolic: 0.90, diastolic: 0.89). Those 2023 models achieved systolic error of 1.3 ± 7.6 mmHg and diastolic error of 0.6 ± 6.4 mmHg . The new device aims to match or exceed that performance in a real‑world wearable form factor.

Clinical Potential and Outstanding Questions

The promise of continuous, calibration‑free hemodynamic monitoring has significant clinical weight. The device could enable early detection of dangerous blood pressure instability in at‑risk patients, guide medication titration in real time, and remove the white‑coat effect that distorts single‑point readings .

Still, several caveats remain. The device has not yet received regulatory clearance, and the University of Utah—which holds the intellectual property—is in early‑stage licensing discussions . The study was funded by the NSF, NIH, the university, and B‑Secur, Ltd., a company in which lead author Benjamin Sanchez Terrones has an equity and leadership role . That tie represents a potential conflict readers and clinicians should weigh.

From a technical standpoint, the biggest advantage of the physics‑driven approach is also its biggest challenge: the quality of the reconstruction depends entirely on how faithfully the model captures real‑world bioimpedance variations. External factors such as motion artifacts, skin hydration, and contact pressure may still degrade signal quality. Ongoing work will need to prove the system is as robust during daily life as it is in controlled trials.

No wearable on the market today provides continuous, calibration‑free blood pressure monitoring at this hemodynamic depth. If the Utah team can navigate the path from lab bench to product, the familiar cuff may finally start to look like a relic of an older era of medicine.