Sleep-EDF Database Expanded (PhysioNet sleep-edfx v1.0.0)

Description

TODO(confirm) — the author’s catalog does not hold this dataset (§10.1 rule 3); this page carries only the §10.2 facts until a registry stub is added (§10.11 item 8).

Curriculum role: Two nights per subject. SC: ‘Two PSGs of about 20 hours each were recorded during two subsequent day-night periods at the subjects homes. Subjects continued their normal activities but wore a modified Walkman-like cassette-tape recorder.’ ST: ‘The PSGs of about 9 hours were recorded in the hospital during two nights, one of which was after temazepam intake, and the other of which was after placebo intake.’ Every PSG has a matching *Hypnogram.edf: ‘All hypnograms were manually scored by well-trained technicians (identified by the eighth letter of the hypnogram filename) according to the 1968 Rechtschaffen and Kales manual, but based on Fpz-Cz/Pz-Oz EEGs instead of C4-A1/C3-A2 EEGs’. Stages are W, R, 1, 2, 3, 4, M (movement time) and ? (not scored) — R&K, not AASM, so N3/N4 are separate and must be merged to compare with an AASM stager.

Citation

• Paper DOI: 10.1109/10.867928
• Dataset DOI: 10.13026/C2X676
• Reference: Kemp, B., Zwinderman, A. H., Tuk, B., Kamphuisen, H. A. C., & Oberyé, J. J. L. (2000). Analysis of a sleep-dependent neuronal feedback loop: the slow-wave microcontinuity of the EEG. IEEE Transactions on Biomedical Engineering 47(9), 1185-1194. doi:10.1109/10.867928 (DOI verified against Crossref 2026-09-18) — the paper the dataset’s own landing page asks for: ‘When using this resource, please cite the original publication’. PhysioNet also asks for its own standard citation, and MNE-Python’s dataset documentation asks for Kemp et al. (2000) together with Goldberger et al. (2000). Dataset: Kemp, B. Sleep-EDF Database Expanded (v1.0.0). PhysioNet. doi:10.13026/C2X676

Download and access

Field	Value
Official source	https://physionet.org/content/sleep-edfx/1.0.0/
Access class	open
Size	8.1 GB total uncompressed (the landing page’s Files section: ‘Total uncompressed size: 8.1 GB’), 197 PSG + 197 hypnogram files. Individually addressable per file over HTTPS, so a notebook fetches only the nights it needs — one SC PSG is about 48 MB (SC4001E0-PSG.edf is 48,338,048 bytes) and its hypnogram is a few kB.
Format	edf, edf_plus
BIDS	false
Mirrors	Legacy PhysioBank path, the one MNE 1.10.2 actually fetches from (mne/datasets/sleep_physionet/age.py BASE_URL ‘/physiobank/database/sleep-edfx/sleep-cassette/’ and temazepam.py ‘/physiobank/database/sleep-edfx/sleep-telemetry/’). Verified 2026-09-18 to serve the same record: the legacy and the published path both return content-range …/48338048 for SC4001E0-PSG.edf.; S3 open bucket: ‘aws s3 sync —no-sign-request s3://physionet-open/sleep-edfx/1.0.0/ DESTINATION’ (the command PhysioNet’s own Files section prints)

License

• Name: ODC-By-1.0
• Note (verbatim from the directory): SETTLED, and it qualifies. Four statements, all read from primary material on 2026-09-18, none of them in conflict — one grant plus three silences. (1) The record’s own landing page states, in its Access block, ‘License (for files): Open Data Commons Attribution License v1.0’ and ‘Access Policy: Anyone can access the files, as long as they conform to the terms of the specified license.’ (2) The full licence text is served by PhysioNet for this record at /content/sleep-edfx/view-license/1.0.0/ and was read there: it is the ODC-By v1.0 text, whose section 3.1 grants ‘a worldwide, royalty-free, non-exclusive, terminable (but only under Section 9) license to Use the Database’, that ‘These rights explicitly include commercial use, and do not exclude any field of endeavour’, and that the rights granted cover ‘Creation of Derivative Databases’ and ‘Distribution, communication, display, lending, making available, or performance to the public by any means and in any form, in whole or in part, including of any Derivative Database’ — subject only to the attribution requirement in section 4. So derivative snippets are expressly permitted with attribution. (3) The licence is a property of THIS record, not of PhysioNet: the same Access block on PhysioNet’s mimiciii/1.4 reads ‘Only credentialed users who sign the DUA can access the files’, ‘License (for files): PhysioNet Credentialed Health Data License 1.5.0’, ‘Data Use Agreement: PhysioNet Credentialed Health Data Use Agreement 1.5.0’ and ‘Required training: CITI Data or Specimens Only Research’ (read 2026-09-18 for exactly this comparison). sleep-edfx carries none of those: its view-dua page serves the licence page with no agreement text, and there is no required-training clause. (4) DISAGREEING STATEMENTS: none found. Three silences, all recorded as silences and not as grants — the published files carry no LICENSE file (the record’s root holds only sleep-cassette/, sleep-telemetry/, RECORDS, RECORDS-v1, SC-subjects.xls, SHA256SUMS.txt and ST-subjects.xls; LICENSE, LICENSE.txt, license.txt and LICENSE.md all return 404); MNE-Python’s doc/documentation/datasets.rst names no licence for this dataset and asks only ‘If you use these data please cite KempEtAl2000 and GoldbergerEtAl2000’; and the legacy PhysioBank path MNE fetches from serves the bytes with no licence page attached. Per §10.7 the most restrictive reading governs, and the most restrictive reading available is still ODC-By 1.0, because nothing states anything narrower. Same licence and same §10.7 class A as ds-eegbci, which is also ODC-By 1.0 on PhysioNet.
• Snippets on this site: allowed — the site may ship short derived snippets and precomputed products from this dataset, labelled as derivatives (re-referenced, filtered, cropped) with citation, license and DOI.

Acquisition

Field	Value
Device	Ambulatory polysomnography, two different recorders. Sleep cassette study (SC): ‘a modified Walkman-like cassette-tape recorder described in chapter VI.4 (page 92) of Bob’s 1987 thesis’. Sleep telemetry study (ST): ‘a miniature telemetry system with very good signal quality’. Both quotes from the PhysioNet landing page, read 2026-09-18.
Device class	TODO(confirm)
Channels	2 (2 EEG derivations (Fpz-Cz and Pz-Oz) plus horizontal EOG, submental chin EMG and an event marker; SC files often also carry oro-nasal respiration and rectal body temperature. This is a polysomnography montage, not an EEG array — the channels scalar counts only the two EEG derivations.)
Sampling rate (Hz)	100 (EEG and EOG at 100 Hz. The submental EMG is not a raw signal: ‘The submental-EMG signal was electronically highpass filtered, rectified and low-pass filtered after which the resulting EMG envelope expressed in uV rms (root-mean-square) was sampled at 1Hz.’ Oro-nasal airflow, rectal body temperature and the event marker are also sampled at 1 Hz. The ST study samples EOG, EMG and EEG at 100 Hz and the event marker at 1 Hz.)
Online filters	TODO(confirm)
Reference	Bipolar derivations (Fpz-Cz, Pz-Oz) — no common reference channel is recorded or needed; the landing page states the derivations and nothing about a reference.
Mains frequency (Hz)	TODO(confirm)
Paradigms	overnight-psg
Participants	100 (197 whole-night recordings from 100 subjects: 153 SC files from 78 subjects and 44 ST files from 22 subjects (RECORDS, read 2026-09-18). Two nights per subject in both studies; ‘The first nights of subjects 36 and 52, and the second night of subject 13, were lost due to a failing cassette or laserdisk’, which is why SC is 153 files and not 156.)
Population	100 subjects across two studies (counted from the dataset’s own RECORDS file, 2026-09-18: 78 distinct SC subject numbers + 22 distinct ST subject numbers). SC (153 files): ‘obtained in a 1987-1991 study of age effects on sleep in healthy Caucasians aged 25-101, without any sleep-related medication’. ST (44 files): ‘obtained in a 1994 study of temazepam effects on sleep in 22 Caucasian males and females without other medication. Subjects had mild difficulty falling asleep but were otherwise healthy.’ Quotes from the PhysioNet landing page; per-subject age and sex are in the file headers and in SC-subjects.xls / ST-subjects.xls, which were not read here.
Clinical groups	none
Sessions	2
Durations	Two nights per subject. SC: ‘Two PSGs of about 20 hours each were recorded during two subsequent day-night periods at the subjects homes. Subjects continued their normal activities but wore a modified Walkman-like cassette-tape recorder.’ ST: ‘The PSGs of about 9 hours were recorded in the hospital during two nights, one of which was after temazepam intake, and the other of which was after placebo intake.’ Every PSG has a matching *Hypnogram.edf: ‘All hypnograms were manually scored by well-trained technicians (identified by the eighth letter of the hypnogram filename) according to the 1968 Rechtschaffen and Kales manual, but based on Fpz-Cz/Pz-Oz EEGs instead of C4-A1/C3-A2 EEGs’. Stages are W, R, 1, 2, 3, 4, M (movement time) and ? (not scored) — R&K, not AASM, so N3/N4 are separate and must be merged to compare with an AASM stager.

Used on this site

L7.2

Loader

Fetches only the stated subset. Recorded loader: mne.datasets.sleep_physionet.age.fetch_data(subjects, recording) for the SC study and mne.datasets.sleep_physionet.temazepam.fetch_data(subjects) for the ST study; each returns the [PSG, Hypnogram] pair per requested night.

# TODO(confirm): loader recorded from the MNE API, not from the catalog.
from mne.datasets.sleep_physionet.age import fetch_data

files = fetch_data(subjects=[0], recording=[1])  # one subject, one night

Caveats worth knowing

• Not in the author’s catalog: every fact here was read from PhysioNet’s own record and from MNE-Python’s source on 2026-09-18, not from data/catalog (§10.1 rule 3). The author should add a registry stub (§10.11 item 8) carrying license.name ODC-By-1.0.
• Hypnograms are scored to the 1968 Rechtschaffen & Kales manual, not to AASM: stages 3 and 4 are separate and there is a Movement-time stage. Any comparison against a modern automatic stager has to merge 3+4 into N3 and decide what to do with M and ’?’ before a kappa means anything — that mapping is a choice the lesson must state, not a detail.
• Scoring was done on Fpz-Cz/Pz-Oz derivations rather than the clinical C4-A1/C3-A2, and there are only two EEG derivations. Spindle and K-complex detectors tuned to central derivations are not being asked the same question here.
• The submental EMG channel is a rectified, filtered envelope sampled at 1 Hz, not a raw EMG signal, and respiration, temperature and the event marker are also at 1 Hz. Only EEG and EOG are at 100 Hz.
• device_class is TODO(confirm) because the §5.2 vocabulary (research-cap, research-dry, consumer, intracranial, microarray) has no class for an ambulatory two-derivation polysomnography montage; none of the five is true of it and guessing one would be a fabricated fact.
• mains_hz is TODO(confirm): the recordings are Dutch (the contributor is at MCH Westeinde Hospital, The Hague) but no primary source read here states a mains frequency, and online_filters is TODO(confirm) for EEG — the landing page documents the EMG processing chain and says nothing about EEG filtering.

Related datasets

• ERP CORE (Compendium of Open Resources and Experiments)
• EEG Motor Movement/Imagery Dataset (EEGMMIDB)
• LEMON (MPI-Leipzig Mind-Brain-Body)
• Dortmund Vital Study resting EEG
• SRM Resting-state EEG

# Loader recorded in the catalog: mne.datasets.sleep_physionet.age.fetch_data(subjects, recording) for the SC study and mne.datasets.sleep_physionet.temazepam.fetch_data(subjects) for the ST study; each returns the [PSG, Hypnogram] pair per requested night
# Fetch only the subset you need; see the loader's docstring for subject/run arguments.
import mne

files = mne.datasets.sleep_physionet.age.fetch_data(subjects, recording) for the SC study and mne.datasets.sleep_physionet.temazepam.fetch_data(subjects) for the ST study; each returns the [PSG, Hypnogram] pair per requested night(...)  # TODO(confirm) subset arguments