# PulseDB Blood Pressure and Synthetic Modeled Quantities
## Description
This package contains 100,000 single-period blood pressure waveforms
derived from PulseDB together with synthetically generated physiological
time-series produced by a biophysical forward model. The dataset was
created to support the training and evaluation of physics-informed
machine learning models for cuffless blood pressure estimation from
wrist bioimpedance signals.
Each record represents one cardiac cycle (diastole to diastole),
represented as 100 uniformly spaced time points. Waveforms were made
periodic via truncation to the first 10 Fourier modes.
## Files
- BP_min2min.csv — Blood pressure waveforms with subject
metadata (SubjectID, Age, Gender, Period,
100 temporal columns in mmHg)
- a_min2min.csv — Modeled radial artery inner radius (mm)
- SBl_min2min.csv — Modeled bulk blood conductivity (S/m)
- ReZ_min2min.csv — Modeled wrist surface bioimpedance,
real part (Ω)
- Biolvar_<param>/ — One folder per biophysical parameter
metaparams.csv Parameter values used for each record
ReZ_min2min.csv Corresponding bioimpedance waveforms
(parameter varied; others held constant)
- Biolvar_all/ — All parameters co-varied simultaneously
metaparams_<1-5>.csv
ReZ_<1-5>_min2min.csv
## Associated Publication
Crandall, H., Schuessler, T., Bělík, F. et al.,
"Cuffless hemodynamic monitoring with physics-informed
machine learning models," Nature Communications, 2026.
https://www.nature.com/articles/s41467-026-72693-1
## Source Data
Blood pressure waveforms are sourced from PulseDB:
Wang et al., Frontiers in Physiology, 2023.
https://physionet.org/content/pulsedb/
## License
[As specified on the PhysioNet project page]