Other Databases of Physiologic Signals

The new PhysioNet website is available at: https://physionet.org. We welcome your feedback.

The databases described here are likely to be of interest to many readers, although they are not currently available in PhysioBank. Please use the contact information below to obtain further information on these resources, since they are independent of PhysioNet; in general, we have no further information about them.

If you are the curator of one of the databases listed below, we encourage you to contribute one or more samples to be posted here. We also welcome corrections and suggestions for this page; please use the link at the bottom of the page to write to us.

AHA Database

The American Heart Association (AHA), with funding from the National Heart, Lung, and Blood Institute (NHLBI), sponsored the development of the AHA Database for Evaluation of Ventricular Arrhythmia Detectors during the late 1970s and early 1980s at Washington University (St. Louis). The first portions of the AHA Database were released in 1982, and it was completed in 1985.

Until recently, the only available portion of the AHA database consisted of 80 two-channel excerpts of analog ambulatory ECG recordings, digitized at 250 Hz per channel with 12-bit resolution over a 10 mV range. These recordings, designated as the development set, are divided into eight classes of ten recordings each, according to the highest level of ventricular ectopy present:

  1. no ventricular ectopy (records 1001 through 1010)
  2. isolated unifocal PVCs (records 2001 through 2010)
  3. isolated multifocal PVCs (records 3001 through 3010)
  4. ventricular bi- and trigeminy (records 4001 through 4010)
  5. R-on-T PVCs (records 5001 through 5010)
  6. ventricular couplets (records 6001 through 6010)
  7. ventricular tachycardia (records 7001 through 7010)
  8. ventricular flutter/fibrillation (records 8001 through 8010)

The final thirty minutes of each recording are annotated beat-by-beat, although supraventricular ectopic beats are not distinguished from normal sinus beats. Two versions of the database are available: the short version includes five minutes of unannotated ECG signals prior to the thirty-minute annotated segment of each recording, and the long version includes 2.5 hours of unannotated ECG signals prior to each annotated segment. The names of the short-version records are the same as the corresponding long-version records, except that the second digit is 2 (thus record 4207 is the short version of record 4007, etc.).

At the time the AHA Database was created, a second set of 75 recordings (designated as the test set) was constructed according to the same criteria as the development set. (Only 5 recordings in the R-on-T PVC class were included in the test set). The test set was intended for evaluations without any possibility that the detectors might have been tuned (optimized) for the test data; for this reason, the test set was unavailable until recently. The names of the records in the test set are similar to those in the development set; the first digit indicates the class, and the second digit is 1 (for the long version) or 3 (for the short version). For example, the test set's record 6102 is a long-version record containing ventricular couplets (class 6).

Both sets of the AHA Database are now available on a DVD from:

ECRI
5200 Butler Pike
Plymouth Meeting, PA 19462 USA

http://www.ecri.org/
Contact:
Kate O'Rourke <korourke@ecri.org>

Additional information is available on ECRI's AHA ECG Database DVD page.

The WFDB Software Package includes a set of converters that can translate between the formats used for the AHA database and the standard formats used for PhysioBank databases.

A record that was excluded from the AHA Database is available here.

Ann Arbor Electrogram Libraries

These annotated recordings of intracardiac electrograms were made and annotated by Dr. Janice Jenkins and her colleagues at the University of Michigan over a period of more than fifteen years. This collection of over 800 recordings, made during electrophysiology studies, is a uniquely valuable resource, particularly for developers and evaluators of implantable cardiac devices. Since intracardiac electrograms can reveal atrial activity much more clearly than typical surface ECGs, these recordings are also useful for basic research in cardiac electrophysiology. The recordings are available from:

Ann Arbor Electrogram Libraries
http://electrogram.com/

CSE Database

The Common Standards for Electrocardiography (CSE) Database is a collection of approximately 1000 short (12- or 15-lead) ECG recordings, designed for evaluating diagnostic ECG analyzers. It is available on a CD-ROM (not in PhysioBank format) from:

Prof. Paul Rubel
INSERM ERM107 - MTIC
Hôpital Cardiologique
28, avenue du Doyen Lépine
69677 BRON Cedex
France

e-mail: prubel.lyon@gmail.com

Improve Data Library

IMPROVE is a European Union concerted action project with the goal of improving on-line assessment and management of patient state in critical care and operating theater environments. The project has developed a large collection of recordings (more than 1300 hours, from 50 subjects) of multiple physiologic signals made in critical care units, extensively annotated and supplemented by simultaneously acquired clinical information. The database is available on nine CD-ROMs (in European Data Format). For further information, see these articles: Building the IMPROVE Data Library, A clinical description of the IMPROVE Data Library, and Collecting EEG signals in the IMPROVE Data Library, all published in IEEE Eng Med Biol Mag, 1997 Nov-Dec; 16(6). Contact:

Dr Mark van Gils
VTT Information Technology
PO Box 1206
FIN-33101 Tampere
Finland

e-mail: mark.vangils@vtt.fi

A sample record was formerly available from the IMPROVE project's web site, http://www.vtt.fi/tte/samba/projects/improve/improve.htm, but this URL is no longer functional.

Sleep Heart Health Study PSG Database

The National Sleep Research Resource (NSRR) offers free web access to large collections of de-identified physiological signals and clinical data elements collected in well-characterized research cohorts and clinical trials. Users can query and search across thousands of data elements, identify those data of most relevance for given needs, explore the statistical distributions of each, and download the data. Physiologic signals from overnight sleep studies are available as downloadable EDF (European data format) polysomnograms (PSGs). Users can also download standard (Rechtschaffen and Kales or AASM) annotations of these PSGs, and summary measures derived from them. The first data set available includes over 8,000 studies from exams 1 and 2 from the Sleep Heart Health Study (SHHS), with scheduled new data releases every quarter from additional data sets. NSRR also provides open-source software for viewing and analyzing these data.

A sample record from the SHHS PSG Database has been available on PhysioBank since 2003.

Telemetric and Holter ECG Warehouse (THEW)

The THEW project at the University of Rochester Medical Center provides its members with access to 3- and 12-lead continuous ECG data for the design and validation of analytic methods in quantitative electrocardiography with a focus on cardiac safety. As of May 2012, the THEW project has 14 ECG databases, most containing several hundred recordings each.

12-lead ECG Library

This resource is a well-organized collection of 12-lead ECGs, available via the Web at http://www.ecglibrary.com/ecghome.html. The ECGs are available as images only. Also available at this site is an extensive set of links to other ECG resources on the Web.

Questions and Comments

If you would like help understanding, using, or downloading content, please see our Frequently Asked Questions.

If you have any comments, feedback, or particular questions regarding this page, please send them to the webmaster.

Comments and issues can also be raised on PhysioNet's GitHub page.

Updated Thursday, 20-Oct-2016 22:52:33 CEST

PhysioNet is supported by the National Institute of General Medical Sciences (NIGMS) and the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under NIH grant number 2R01GM104987-09.