A physics-informed peripheral hemodynamic digital twin for personalized estimation of arterial stiffness

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Arterial stiffness / Vascular aging (n=118). Intervention: Physics-informed peripheral hemodynamic digital twin model (PBM + O4) vs. Baseline physiologically based model (PBM) using age and blood pressure. Primary outcome: Estimation of central-brachial pulse wave velocity (PWV) root mean square error (RMSE, m/s) in the blind test cohort (r = 0.67 vs r = 0.22 (RMSE 1.12 vs 1.90 m/s)).

Q: What does this research mean for the field?

Incorporating parameters from a physics-informed peripheral hemodynamic digital twin improves central-brachial pulse wave velocity estimation compared with models relying solely on age and blood pressure. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Arian Aghilinejad; Alessio Tamborini; Faisal Amlani; Morteza Gharib

doi:10.1038/s41746-026-03431-6

October 28, 2026npj Digital MedicineOpen Access

A physics-informed peripheral hemodynamic digital twin for personalized estimation of arterial stiffness

View Full Paper

Ask AI

Bookmark

Share

Key result

Physics-informed digital twins improve central-brachial PWV estimation accuracy, lowering estimation error to ~1.1 m/s.

r = 0.67 vs r = 0.22 (RMSE 1.12 vs 1.90 m/s)
n=118

Why the study?

Conventional pulse wave velocity measurements for arterial stiffness require dual-site recordings, which limits their scalability in clinical settings.

Does a physics-informed peripheral hemodynamic digital twin framework using a single brachial pressure waveform improve the estimation of central-brachial pulse wave velocity compared with a baseline model using only age and blood pressure in adult subjects?

Key Points

To develop and evaluate a physics-informed peripheral hemodynamic digital twin framework capable of estimating central-brachial arterial stiffness from a single brachial pressure waveform.
Acquired high-resolution brachial pulse waveforms from a cohort of 118 subjects (mean age 65 years, 40.7% women) using a laboratory-developed cuff device.
Calibrated a reduced-order wave propagation model via PDE-constrained optimization of four parameters: maximal ventricular elastance, left ventricular volume, aortic wave speed, and peripheral resistance.
Evaluated machine learning models predicting central-brachial pulse wave velocity (PWV) using a split-sample approach with a training cohort (n = 83) and a blind testing cohort (n = 35).
Incorporating all four digital-twin-derived parameters achieved r = 0.67 and an RMSE of 1.12 m/s on the blind testing cohort, compared with r = 0.22 and an RMSE of 1.90 m/s for a baseline model using only age and blood pressure.
Bland–Altman analysis for the full digital twin model on the blind testing cohort demonstrated a mean bias of 0.21 m/s with limits of agreement between −1.98 and 2.41 m/s.
Permutation feature importance identified peripheral resistance as having the largest effect on model performance, followed by diastolic blood pressure, age, and left ventricular volume.

Study Design

Type

Observational (n=118)

Multicenter

Yes

Structured PICO

Does a physics-informed peripheral hemodynamic digital twin framework using a single brachial pressure waveform improve the estimation of central-brachial pulse wave velocity compared with a baseline model using only age and blood pressure in adult subjects?

P

Population

118 adults (mean age 65 years, 40.7% women) referred for non-emergent left heart catheterization across multiple centers in the United States.

E

Exposure

Physics-informed peripheral hemodynamic digital twin framework using a single brachial pressure waveform acquired by a laboratory-developed cuff device, personalizing a reduced-order left ventricle–aortic wave propagation model through PDE-constrained optimization of four physiological parameters (maximal ventricular elastance, left ventricular volume, aortic wave speed, and peripheral resistance) combined with conventional clinical variables (age, SBP, DBP) in a machine learning model (PBM + O4).

C

Comparator

Baseline physiologically based model (PBM) using only conventional predictors (age, systolic blood pressure, and diastolic blood pressure) and reference central-brachial pulse wave velocity derived from pulse transit time.

O

Outcome

Estimation accuracy of central-brachial pulse wave velocity (PWV), assessed by Pearson correlation coefficient (r), root mean square error (RMSE), bias, and limits of agreement against reference PWV in a blind testing cohort.surrogate

Main Result

Effect estimate: r = 0.67 vs r = 0.22 (RMSE 1.12 vs 1.90 m/s)

Absolute Event Rate: 1.12% vs 1.9%

A physics-informed digital twin framework calibrated to a single brachial cuff waveform significantly improves noninvasive estimation of central-brachial arterial stiffness compared to demographic and blood pressure predictors alone.

Limitations

Targets central-brachial PWV rather than the guideline reference standard carotid-femoral PWV
Simplified single aortic transmission tube model does not resolve regional differences along the arterial tree
Central-brachial path length was estimated from subject height rather than measured directly
Modest sample size, particularly in the held-out blind test cohort (n = 35)
Older cohort with higher prevalence of cardiovascular risk factors limits direct extrapolation to younger or healthier populations
Limited number of subjects in each cardiovascular risk-factor stratum within the testing cohort
Inferred parameters are only partially identifiable from a single distal waveform (particularly aortic wave speed ca)

Implication

May enhance single-site PWV estimation in research; leaves open prospective validation before clinical adoption.

Abstract

Arterial stiffness, commonly assessed by pulse wave velocity (PWV), is an important biomarker of cardiovascular risk, yet conventional PWV measurements require dual-site recordings that limit scalability in clinical settings. Here we introduce a physics-informed peripheral hemodynamic digital twin framework for estimating central-brachial arterial stiffness from a single brachial pressure waveform. Using our laboratory-developed cuff device, high-resolution brachial pulse waveforms were collected in a cohort of 118 subjects. The digital twin then personalizes a reduced-order wave propagation model through PDE-constrained optimization, generating a compact parameter space that encodes ventricular–vascular characteristics relevant to stiffness. Incorporating digital-twin-derived parameters improved PWV estimation on a blind test cohort ( r = 0.67, RMSE = 1.12 m/s) compared with a baseline model using only age and blood pressure ( r = 0.22, RMSE = 1.90 m/s). These results indicate that physics-informed digital twins can enable scalable, noninvasive assessment of vascular aging from accessible peripheral waveforms.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Aghilinejad et al. (2026) conducted an observational in Arterial stiffness / Vascular aging (n=118). Physics-informed peripheral hemodynamic digital twin model (PBM + O4) vs. Baseline physiologically based model (PBM) using age and blood pressure was evaluated on Estimation of central-brachial pulse wave velocity (PWV) root mean square error (RMSE, m/s) in the blind test cohort (r = 0.67 vs r = 0.22 (RMSE 1.12 vs 1.90 m/s)). Incorporating physics-informed digital-twin parameters improved central-brachial PWV estimation (r = 0.67, RMSE = 1.12 m/s) compared to a baseline age and blood pressure model (r = 0.22, RMSE = 1.90 m/s).

synapsesocial.com/papers/6ae433d2d9418cd7c8f4701b https://doi.org/10.1038/s41746-026-03431-6

View Full Paper

Ask AI

Bookmark

Share

A physics-informed peripheral hemodynamic digital twin for personalized estimation of arterial stiffness

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

A physics-informed peripheral hemodynamic digital twin for personalized estimation of arterial stiffness

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading