Metrology Atlas

Describing clinical time-series data as a measurement system

A recorded time and a recorded value are both measurements, and neither is exact. The Metrology Atlas describes a dataset in the terms of the International Vocabulary of Metrology (VIM) and the Guide to the Expression of Uncertainty in Measurement (GUM), so that what each record tells you, and how sure you can be, is written down, checkable and machine-readable.

Browse the atlases Read the paper

What an atlas states, for every variable

The interval it describes A heart rate averages several seconds of ECG; a breath-by-breath value describes one breath. The atlas gives that interval and where the time stamp sits in it.
Bias and uncertainty The bias and the standard uncertainty of its time and its value, as a budget of named components: which errors are shared with other signals, and for how long they persist.
The evidence Every number is marked as measured, derived, estimated or assumed, and linked to the study and the code that produced it.

An atlas is a living document. Studies are added and numbers revised as evidence comes in, and each release is archived with a DOI, so that a paper can cite a fixed version while the website moves on.

Atlases

VitalDB metrology atlas

Open for review Version 0.3.0 · VitalDB 1.0.0 (PhysioNet)

Every recorded value in the VitalDB open dataset (6,388 surgical cases, PhysioNet release 1.0.0) described as a measurement: the interval it describes, the bias and standard uncertainty of its time and value, and the studies behind each number.

143 variables · 12 clocks · 93 studies · 136 parameters

vitaldb.metrology-atlas.org · Atlas file (JSON) · doi:10.5281/zenodo.23494768

The approach will be applied to further datasets. Each atlas gets its own site under metrology-atlas.org and its own DOI.

Using an atlas

Read it Each variable's page states its timing forms with their uncertainty budgets, its value model and the chain of clocks behind its stamp; each study links to its code.
Load it The whole atlas is one JSON file. Analysis code can read the forms, biases and budgets directly, and the planner combines signals into the uncertainty of their difference.
Ask about it The file is a compact way to give an AI assistant quantitative context: attach it, describe the signals and assumptions of a study, and check each answer against the website.

Software

metrology-atlas 0.3.0. The engine. Schema, checker, planner, and the builder of each atlas's website and machine-readable file.
physiosync 0.3.4. Timing corrections for multi-signal physiological recordings. Works on times and samples you give it; needs only numpy and scipy.
vitaldb-reader 0.1.0. Reads VitalDB .vital files at native resolution and applies the VitalDB atlas's corrections as separate, versioned steps. Needs only numpy.

Papers

The Metrology Atlas: describing clinical time-series data as a measurement system. Andrew J. Goodwin, Hyung-Chul Lee, Danny Eytan. arXiv preprint, arXiv:2610.29001. Version 2 accompanies VitalDB atlas v0.3.0.

News

21 Oct 2026 VitalDB atlas v0.1.0 and the first preprint The first atlas, released for review with the engine and a preprint.

All news · RSS