Sleep-EDF Database

Expanded Database

A greatly expanded version of this database is now available (see the Sleep-EDF Database [Expanded]), and is recommended for all new studies. This version is obsolete but will remain available to support ongoing work.

Data Description

Files with extensions .rec and .hyp contain original recordings and their hypnograms, respectively, in European Data Format. EDF is specified in

B Kemp, A Värri, AC Rosa, KD Nielsen, J Gade. A simple format for exchange of digitized polygraphic recordings. Electroencephalography and Clinical Neurophysiology 82:391-393 (1992).

More about EDF is on www.edfplus.info. Each EDF file has a header specifying patient, recording and signals including their calibration. Converters to (and from) WFDB native format are included in the WFDB Software Package (see edf2mit(1) in the WFDB Applications Guide). Applications using WFDB library version 10.4.5 or later can read EDF files directly with no conversion required.

The recordings were obtained from Caucasian males and females (21–35 years old) without any medication; they contain horizontal EOG, FpzCz and PzOz EEG, each sampled at 100 Hz. The sc* recordings also contain the submental-EMG envelope, oro-nasal airflow, rectal body temperature and an event marker, all sampled at 1 Hz. The st* recordings contain submental EMG sampled at 100 Hz and an event marker sampled at 1 Hz.

The sc*.rec recordings contain the envelope of the EMG signal, at a 1Hz sampling frequency. The dimension of this envelope signal, "uV-mrs", means "microvolt mean-root-squared", in which 'mrs' is related to, but not synonymous with, the well-known abbreviation 'rms' (root-mean-square). In both rms and mrs, the letters, r, m, and s mean 'root', 'mean', and 'square', respectively. The order of these components differs between rms and mrs, however, making them essentially different variables. In this case, the electrical EMG signal is expressed in microvolts (uV), then rectified (root-squared: rs) and then smoothed (mean: m) over the 1-second time-intervals between two successive samples. So it is the mean (m) of the root-squared (rs) signal: mrs.

The st*.rec recordings came from body-worn bio-signal amplifiers that transmitted those signals digitally to a PC. All transmitters had the same FM radio frequency. So, when two patients were recorded next to each other the FM receivers might mix up those patients. That risk was very small due to the location of transmitter and receiver, but in order to be sure each transmitter had a different baseline of the ID+Sync+Error signal. For instance the st7132j0.rec and st7052j0.rec recordings were made by transmitters numbers 1 and 2, respectively, hence the baseline of the signal is 1 and 2, respectively in those two recordings. The transmitter also had a marker (M) button: pressing the button added 32 to the 'ID+Sync+Error' signal. The system also continuously monitored the quality of the received (digital) signal: any error subtracted 32 from the signal. So, the dimension "ID+M-E" indicates that the baseline of the signal identifies (ID) the transmitter, shows the marker button as a positive deflection (+M) and any error in the telemetry link as a negative deflection (-E). Nice examples are in all ST recordings because the telemetry link in this project was optical and it was disconnected between the hook-up in the afternoon and the start of the trial late in the evening. At the start (at 00:31:17 in st7132j0.rec and 23:23:15hr in st7052j0.rec) the ID+Sync+Error signal switches from negative (error: no connection) to 1 or 2 (ID of the transmitter), and a few seconds later the patient or nurse pressed the marker button which shows as a short (1 second) positive peak. In this project the baseline level proves that in the evening each patient was reconnected to the same file that he was hooked-up on in the afternoon. This is particularly important because the afternoon session included the creation of the patient-ID in the EDF header.

Hypnograms are manually scored according to Rechtschaffen & Kales based on Fpz-Cz / Pz-Oz EEG instead of C4-A1 / C3-A2 EEG: see

B van Sweden, B Kemp, HAC Kamphuisen, EA van der Velde. Alternative electrode placement in (automatic) sleep scoring (Fpz-Cz / Pz-Oz versus C4-A1 / C3-A2). Sleep 13(3):279-283 (1990).

The sleep stages W, 1, 2, 3, 4, R, M and 'unscored' are coded in the file as binaries 0, 1, 2, 3, 4, 5, 6 and 9 according to Q16.3 of the EDF-FAQ.

The 4 sc* recordings were obtained in 1989 from ambulatory healthy volunteers during 24 hours in their normal daily life, using a modified cassette tape recorder. Subjects, recordings and hypnogram scoring are described in

MS Mourtazaev, B Kemp, AH Zwinderman, HAC Kamphuisen. Age and gender affect different characteristics of slow waves in the sleep EEG. Sleep 18(7):557-564 (1995).

The 4 st* recordings were obtained in 1994 from subjects who had mild difficulty falling asleep but were otherwise healthy, during a night in the hospital, using a miniature telemetry system with very good signal quality. Subjects and recordings are more extensively described in the 2000 IEEE-BME paper cited above.

B Kemp, AH Zwinderman, B Tuk, HAC Kamphuisen, JJL Oberyé. Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG. IEEE-BME 47(9):1185-1194 (2000) [PDF].

Contributors

Bob Kemp, Sleep Centre, MCH-Westeinde Hospital, Den Haag, The Netherlands