WFDB Header Files

Each WFDB record includes one or more header files (.hea) that specify the associated signal files and their attributes. Header files are plain ASCII text and are critical for interpreting the associated signal and annotation files.

Structure of a Header File

A header file contains:

  • A record line (first non-comment line), specifying metadata about the entire record.
  • One or more signal specification lines, detailing each signal.
  • Optionally, comment lines (beginning with #).

Each line must be under 255 characters, and fields are separated by spaces or tabs (except where otherwise noted).

Detailed documentation on header files can be found at: https://physionet.org/physiotools/wag/header-5.htm

Record Line

The first non-comment line is the record line, which provides metadata about the overall record. It includes:

Field Description
Record name Identifier for the record (letters, digits, underscores only).
Number of segments (optional) If present, appended as /n. Indicates a multi-segment record.
Number of signals Number of signals described in the header.
Sampling frequency (optional) Samples per second per signal. Defaults to 250 if omitted.
Counter frequency (optional) Secondary clock frequency, separated from sampling frequency by a /.
Base counter value (optional) Offset value for counter, enclosed in parentheses.
Number of samples (optional) Total samples per signal when samps_per_frame=1; total frames otherwise.
Base time (optional) Start time of the recording (HH:MM:SS).
Base date (optional) Start date (DD/MM/YYYY).

Examples

Basic header (uniform sampling frequency): record 100 from the MIT-BIH database

100 2 360 650000
  • 100: Record name (must match the filename prefix).
  • 2: Number of signals in the record.
  • 360: Sampling frequency in Hz.
  • 650000: Number of samples for each signal.

Multi-frequency header (unique sampling rates per channel):

12345 3 62.5 625 12:00:00 30/01/1989
  • 12345: Record name (must match the filename prefix).
  • 3: Number of signals in the record.
  • 62.5: Sampling frequency in Hz.
  • 625: Number of samples. For multi-frequency records this is the total frames for each signal.
  • 12:00:00: Base time.
  • 30/01/1989: Base date.

Signal Specification Lines

Each signal has its own line immediately following the record line (for single-segment records). Fields are:

Field Description
File name Name of the file containing signal data.
Format Integer code describing how signal data is encoded (e.g., 16 for 16-bit integers).
Samples per frame (optional) Number of samples per frame (if not 1), separated by x.
Skew (optional) Signal-specific time offset (number of samples), separated by :.
Byte offset (optional) Byte position to start reading the signal file, separated by +.
ADC gain (optional) Scaling factor to convert integers to physical units (e.g., millivolts).
Baseline (optional) Integer value representing 0 physical units.
Units (optional) Label for signal units (e.g., mV).
ADC resolution (optional) Bits per sample (defaults to format’s resolution if omitted).
ADC zero (optional) Sample value corresponding to physical zero.
Initial value (optional) First sample value (used for checksums).
Checksum (optional) Sum of sample values (modulo 65536) for verification.
Block size (optional) Number of samples per block (for formats supporting block I/O).
Description (optional) Free-text description of the signal (e.g., lead name ECG Lead II).

Examples

Basic header (uniform sampling frequency): record 100 from the MIT-BIH database

100.dat 212 200 11 1024 995 -22131 0 MLII
100.dat 212 200 11 1024 1011 20052 0 V5
  • 100.dat: File name of the signal file.
  • 212: Format samples are stored in (12-bit two’s complement).
  • 200: ADC gain (i.e. number of digital values per physical unit).
  • 11: ADC resolution (bits).
  • 1024: ADC zero value.
  • 995, 1011: Initial value.
  • -22131, 20052: Checksum (sum of all signal samples modulo 2^16).
  • 0: Block size.
  • MLII, V5: Description (e.g., signal lead names).

Multi-frequency header (unique sampling rates per channel):

12345.dat 16x4 200/μV 12 0 0 2178 0 ECG
12345.dat 16x2 16/mmHg 12 0 0 3497 0 ICP
12345.dat 16x1 2500/Ohm 12 0 0 1366 0 RESP
  • 12345.dat: File name of the signal file.
  • 16: Format samples are stored in (16-bit integers).
  • x4,x2,x1: 4, 2, and 1 samples per frame, respectively. This indicates that the ECG signal has 157 * 4 = 628 samples while the ICP signal has 314 samples and the RESP signal has 157 samples.
  • 200/μV, 16/mmHg, 2500/Ohm: ADC gain (i.e., number of digital values per physical unit).
  • 12: ADC resolution (bits).
  • 0: ADC zero value.
  • 0: Initial value.
  • 2178 , 3497, 1366: Checksum (sum of all signal samples modulo 2^16).
  • 0: Block size.
  • ECG, ICP, RESP: Description (e.g., signal lead names).

Comments and Info Strings

  • Any line beginning with # is a comment.
  • Comments after the last signal line are preserved and may provide additional metadata (“info strings”).
  • Earlier comments are ignored by programs reading the header.

Notes

  • A header file may describe signals stored in multiple files or multiple signals in a single file.
  • Fields like sampling frequency, counter frequency, and base time/date improve time-aligned analysis but are optional.
  • Storage format options and details can be found at: https://physionet.org/physiotools/wag/signal-5.htm
  • Multi-segment records use a slightly different structure (described separately).