Skip to content

Heart-rate variability (HRV)

Columns: hrv_sdnn_ms, hrv_rmssd_ms, hrv_pnn50 · Units: ms, ms, % · Measured on: lead II (default)

In one sentence: heart-rate variability describes how much the time between heartbeats changes from one beat to the next.

Everyday picture

A healthy heart is not a metronome. It speeds up a little every time you breathe in and slows down when you breathe out, and it keeps adjusting to what the body needs. A relaxed, healthy heart shows more of this wobble; stress, exercise, illness and age reduce it. A heart that beats with almost perfect regularity is not a sign of good health.

Try it

The top strip shows the first 10 seconds with each RR interval written above it. The bottom plot (a tachogram) shows every RR interval of the whole recording. Red links mark beat-to-beat jumps of more than 50 ms.

Interactive demo loading… (needs JavaScript)

What to notice:

  • Relaxed: the RR intervals swing slowly up and down with breathing.
  • Stressed / exercising: fast and very regular, so all three numbers drop.
  • Irregular rhythm: big random jumps between neighbouring beats. RMSSD and pNN50 shoot up. Very high values in a short ECG often point to an irregular rhythm such as atrial fibrillation, rather than to a "healthy" heart.
  • 10-second ECG: with only about 10 beats the numbers jump around when you change other settings. Standard 10 s ECGs give rough HRV estimates.

The three numbers

Column Name What it captures
hrv_sdnn_ms SDNN: standard deviation of the RR intervals the overall spread, slow swings included
hrv_rmssd_ms RMSSD: root mean square of successive differences the typical jump between neighbouring beats
hrv_pnn50 pNN50: percentage of successive differences above 50 ms how often neighbouring beats differ noticeably

There are no single "normal" values. HRV falls with age, rises with fitness, and depends heavily on the length of the recording, which is why HRV norms are usually given for 5-minute or 24-hour recordings.

Why it matters for synthetic ECGs

Many generators learn one "typical" beat and repeat it at a perfectly steady pace. Each beat looks right, but the rhythm is unnaturally regular, with HRV close to zero. Comparing the HRV distributions of real and synthetic recordings catches this. The hrv metric does that comparison automatically.

How ECGEN-Eval measures it

  1. R peaks are detected on the primary lead (the same peaks as heart rate).
  2. RR intervals outside 300–2000 ms are dropped as implausible.
  3. From the remaining RR series: SDNN (sample standard deviation), RMSSD and pNN50, with pNN50 in percent (0–100).

At least three R peaks are needed; otherwise the values are NaN.

pNN50 units differ

The hrv_pnn50 feature is a percentage (0–100). The hrv metric computes pNN50 as a fraction (0–1). Both are fine for comparing real with synthetic data, but don't mix them.

Learn more