PhysioBank Archives: Class 3

Class 3 databases include collections of data that may have been less thoroughly studied than those in class 1, but that may be of interest to the research community. These databases include works in progress, to which users are invited to contribute. In some cases, these databases may be archived on their creators' web sites.

The following class 3 databases are currently available:

• AF Termination Challenge Database. This database has been compiled for the PhysioNet/Computers in Cardiology Challenge 2004. It consists of a learning set of 30 records and two test sets of 30 and 20 records. Each record contains a one-minute excerpt of a two-lead long-term ECG recording exhibiting either self-terminating or sustained atrial fibrillation; the challenge is to identify which records in the test set show self-terminating AF.
• Apnea-ECG Database. This database has been assembled for Computers in Cardiology Challenge 2000. It consists of 70 ECG recordings, each typically 8 hours long, with accompanying sleep apnea annotations obtained from study of simultaneously recorded respiration signals, which are included for 8 of the recordings.
• Creighton University Ventricular Tachyarrhythmia Database. This database includes a preliminary set of beat annotations (all beats marked as normal) with additional annotations that indicate episodes of ventricular fibrillation/flutter.
• Fantasia Database. ECG and respiration recordings, with beat annotations from 20 young and 20 elderly subjects, all healthy, in sinus rhythm during a resting state (two hours each). Half of the recordings also include (uncalibrated) continuous noninvasive blood pressure signals. A subset of these recordings posted here previously is still available.
• Intracardiac Atrial Fibrillation Database. A collection of high-resolution recordings from eight subjects in atrial fibrillation or flutter; each recording includes three surface ECG signals and five intracardiac signals, all simultaneously recorded.
• MIMIC Database. The PhysioBank archives contain 72 complete records from this database, together with periodic measurements ("numerics") for all 121 records of the database, including multiple recordings of some of the 90 subjects. The lengths of these records vary, but average about 40 hours each.
• MIMIC II Database. The PhysioBank archives currently contain digitized signals (typically including ECG, ABP, and respiration), trends, and ICU monitor alarms (where available) from 2769 records of this database. These records are up to several weeks in duration.
• MIT-BIH Atrial Fibrillation Database (including signal files not previously released). Signal files for 23 of the 25 ten-hour records are available, along with reference rhythm annotations and unaudited beat annotations for all 25 records.
• MIT-BIH ECG Compression Test Database. This database is unannotated.
• MIT-BIH Long-Term Database. Six lengthy two-lead ECG recordings and one three-lead ECG recording.
• MIT-BIH Malignant Ventricular Arrhythmia Database. This database contains rhythm and signal quality annotations only (no beat annotations).
• MIT-BIH Normal Sinus Rhythm Database (including signal files not previously released). Also available: recordings excluded from the MIT-BIH Normal Sinus Rhythm Database (because of the presence of occasional ectopic beats).
• MIT-BIH Polysomnographic Database. Includes new annotation files with sleep stage and apnea annotations.
• MIT-BIH ST Change Database. This database includes beat annotations but currently no ST change annotations. The recordings are primarily from exercise stress tests and exhibit transient ST changes.
• MIT-BIH Supraventricular Arrhythmia Database. Seventy-eight half-hour ECG recordings chosen to supplement the examples of SV arrhythmias in the MIT-BIH Arrhythmia Database.
• Non-Invasive Fetal Electrocardiogram Database. Fifty-five recordings of maternal and maternal+fetal ECGs recorded over a 20-week period from a single subject, in EDF+ format.
• PAF Prediction Challenge Database. This database has been compiled for Computers in Cardiology Challenge 2001. It consists of 100 record sets, each including a pair of 30-minute excerpts from a long-term ECG recording. Approximately half of the subjects have PAF immediately following one of the two 30-minute excerpts; among the 50 record sets in the learning set, the PAF can be studied by referring to 5-minute ``continuation records'' that accompany each 30-minute record. In the 50 record sets belonging to the test set, the challenge is to identify which records immediately precede PAF.
• PTB Diagnostic ECG Database. This database of 549 high-resolution 15-lead ECGs (12 standard leads together with Frank XYZ leads) includes clinical summaries for each record. From one to five ECG records are available for each of the 294 subjects.
• Samples of MR Images (magnetic resonance angiography images)
• Sleep-EDF Database. This is a collection of sleep recordings from 8 healthy subjects. Each recording contains 2 EEG signals, EOG, and an event marker, and is accompanied by a manually-scored hypnogram. Four recordings also include submental EMG, and the other four recordings also include the submental EMG envelope, oro-nasal airflow, and body temperature.
• Sleep Heart Health Study Polysomnography Database. This database will eventually include 1000 overnight polysomnograms collected to study the relationship of sleep disordered breathing and cardiovascular disease. Recordings include EEG, EOG, EMG, ECG, nasal airflow and respiratory effort signals, periodic measurements of SaO2 and heart rate, annotations of sleep stages, respiratory events, EEG arousals, and more. At present, one sample polysomnogram is available.
• St. Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database. Seventy-five half-hour recordings extracted from 32 Holter records from patients undergoing tests for coronary artery disease, with reference annotation files containing over 175,000 beat annotations in all.
• St. Vincent's University Hospital / University College Dublin Sleep Apnea Database. This database contains 25 full overnight polysomnograms with simultaneous three-channel Holter ECG, from adult subjects with suspected sleep-disordered breathing.
• Sudden Cardiac Death Holter Database. This is a collection of long-term ECG recordings of patients who experienced sudden cardiac death during the recordings. Half-hour excerpts of these recordings are available as the MIT-BIH Malignant Ventricular Arrhythmia Database.

Your comments and suggestions are welcome. We encourage you to use our feedback form to comment on this page. If you would like to receive a reply, please send your comments by email to webmaster@physionet.org, or post them to: PhysioNet MIT Room E25-505A 77 Massachusetts Avenue Cambridge, MA 02139 USA

Updated Thursday, 01-May-2008 14:31:39 EDT