Physics-informed models enable cuffless continuous blood pressure

Researchers at the University of Utah and the University of Illinois, Chicago published a manuscript in Nature Communications (published 14 May 2026) and an arXiv preprint (arXiv:2601.00081, submitted 31 Dec 2025) describing a smartwatch prototype that uses real-time electrical bioimpedance (BioZ) sensing and a physics-informed neural network to estimate continuous blood pressure and radial/axial blood velocity. The Nature Communications manuscript and the arXiv preprint report a multiscale analytical and computational modelling framework that links pulsatile BioZ signals at the wrist to hemodynamic variables, and describe a "signal-tagged" physics-informed model that incorporates fluid dynamics principles to produce calibration-free BP estimates. News-Medical and a university press release note that the University of Utah holds the intellectual property and that its Technology Licensing Office is exploring licensing opportunities. Benjamin Sanchez Terrones, a University of Utah researcher, is quoted in the News-Medical coverage as calling elevated blood pressure "the silent killer."

Per the arXiv preprint and Nature Communications manuscript, the team combines: a multiscale biophysical model that identifies physiological, anatomical, and experimental parameters influencing wrist BioZ; an experimental smartwatch that records BioZ waveforms; and a physics-informed learning architecture that constrains neural-network predictions with hemodynamic equations. The authors report validation datasets spanning healthy subjects at rest and after physical activity, autonomic challenges, and patients with hypertension and cardiovascular disease in outpatient and intensive care settings, as described in the preprint. The papers emphasise the method is calibration-free and yields continuous arterial pressure waveforms and blood velocity estimates rather than intermittent cuff readings.

Physics-informed approaches embed known differential equations or conservation laws into learning models to reduce reliance on large labelled datasets and to improve generalization under distribution shifts. Industry and academic work on PINNs and physics-informed temporal networks (see PITN and related literature) has shown gains in robustness for physiological time series, especially when ground-truth cuff measurements are sparse or obtrusive. For practitioners, this means sensor fusion with principled biophysical priors can reduce per-subject calibration needs and make ambulatory hemodynamics more tractable, provided the underlying forward models capture dominant physiology and sensor confounders.

Industry reporting and the peer-reviewed manuscript frame this result as addressing limitations of existing cuffless BP methods that rely on pulse wave analysis or pulse arrival time and often behave like black boxes. Editorial analysis: If reproduced and clinically validated at scale, continuous cuffless BP from BioZ plus physics-informed models could shift monitoring from episodic clinic checks to persistent ambulatory assessment, changing how hypertension is tracked in trials and care pathways. However, broad clinical adoption will require prospective regulatory-grade validation, interoperability with clinical workflows, and demonstration of measurement performance across diverse populations and motion conditions.

• •Independent replication studies and larger multicenter clinical validation cohorts demonstrating regulatory-standard accuracy and repeatability.
• •Detailed performance metrics under motion, different skin types, arrhythmias, and during vasoactive drug use, as these are common failure modes for cuffless devices.
• •Licensing announcements or spinouts from the University of Utah and any commercial partners that publish device-level specifications and regulatory filings.