This database contains 135 pairs of RR interval time series, recorded by implanted cardioverter defibrillators (Medtronic Jewel PlusTM ICD 7218) in 78 subjects. Each series contains between 986 and 1022 RR intervals. One series of each pair includes a spontaneous episode of ventricular tachycardia (VT) or ventricular fibrillation (VF), and the other is a sample of the intrinsic (usually sinus) rhythm.
The ICD continuously digitizes and analyzes the patient's electrogram (EGM). Four EGM signals are digitized:
- HVA - HVB (Can - Coil)
- P-/S-P+/S (Tip - Ring)
- HVA-P+/S (Can - Ring)
- P-/S - HVB (Tip - Coil)
R-waves are detected in the EGM with an auto-adjusting sense amplifier. The automatic threshold adjustment can affect measured cycle lengths if the sensed R wave crosses the decaying threshold at different levels. RR interval measurements are always based on the P-/S-P+/S (Tip-Ring) EGM. Each interval is classified depending on the programmed tachycardia and fibrillation detection intervals (TDI and FDI, respectively).
The ICD maintains a buffer containing the 1024 most recently measured RR intervals. When a tachyarrhythmia is detected, the ICD delivers programmed pacing, cardioversion, or defibrillation therapy and records a snapshot of the RR interval buffer, thus capturing the sequence of 1024 RR intervals that immediately precede the detected event (ending at the time when the ICD first detected the event, not at the termination of the event). The stored intervals are truncated (not rounded) to a multiple of 10 ms (e.g., an interval of 457 ms is recorded in the snapshot as 450 ms). The most recent snapshot is stored in the ICD's memory until the patient returns for a checkup, at which time the ICD is interrogated using a Medtronic 9790 ICD programmer, the stored snapshot is downloaded, and the ICD may be reprogrammed as appropriate.
In addition to the stored snapshot of the most recently detected event, a snapshot of the RR interval buffer is also obtained when the ICD is interrogated; such snapshots thus contain the most recent RR intervals, which can be used as baseline (control) data to compare with the detected events. Varying numbers of the least recent intervals in these snapshots are lost during device interrogation, so these baseline sequences contain between 986 and 1022 intervals each.
The snapshots downloaded from the ICDs also include a few seconds of the high-resolution EGM, as illustrated below; EGM segments are not available for the records in this database, however.
Each RR interval sequence included in this database has a three-part file name of the form ssss.ttn, where ssss is a four-digit patient number, tt is either vf, vt, or mr, and n is a sequence number for each type of event. The files as provided by Medtronic can be found in the RRdata1 and mr directories. They are in plain text format, and each contains a column of numbers, which are the RR intervals in milliseconds. In addition, a set of PhysioBank-compatible annotation and header files has been created from the text files; these are in the data directory. These have record names of the form ssss_ttn, where the components are as described above.
|intervals prior to detection of spontaneous VF
|intervals prior to detection of spontaneous VT
|intervals prior to ICD interrogation ("mr" = "most recent")
Each vf and vt sequence has a matching mr (baseline) sequence designated by the same numerals in the sequence names. For example, 8076.mr2 is the baseline sequence matching 8076.vt2. These matched pairs of RR interval sequences are referred to as sequence pairs below.
Several patients had multiple VT or VF episodes. The database was subdivided into 4 datasets:
|the most recent VT sequence pair from each patient who experienced at least one detected VT
|the most recent VF sequence pair from each patient who experienced at least one detected VF
|all earlier VT sequence pairs
|all earlier VF sequence pairs
Researchers should use care in combining these datasets to avoid bias due to any one patient with many sequence pairs.
Six patients had both VT and VF episodes. In five of these cases, the baseline sequence associated with the first VF has the suffix mra, to avoid conflicts with the baseline sequences associated with the first VT (with the standard suffix mr1). The names of the files in the sequence pairs for these six patients are given in the table below:
|VT sequence pairs
|VF sequence pairs
The episodes included in this database were recorded by Medtronic ICD 7218 defibrillators, from patients without pacemakers in the US and Canada on or before 3 January 1997. Episodes were retrieved from the implanted devices during the patients' scheduled follow-ups and during subsequent unscheduled visits. Each episode was classified manually using the criteria listed in the next section. The final database included all recorded episodes except for:
- spontaneous episodes that occurred immediately post-implant, or between the initial and final interrogation during a follow-up
- episodes for which the VT or VF EGM could not be compared with a sinus rhythm EGM from the same source
- VTs and VFs for which fewer than 1000 R-R intervals were recorded (e.g., an episode occurring within a few minutes of another)
- SVTs that were inappropriately detected as VT or VF.
In all, 142 patients were studied, and this database contains 135 episodes from 78 patients who experienced at least one VT or VF episode. Five patients had both a VT and a VF episode included here.
An accompanying Excel spreadsheet (phys.xls) contains the following information for each sequence pair in the database:
- file name for the VT or VF sequence
- file name for the MR sequence
- date and time of the VT or VF episode
- ICD's classification of the episode
- manual classification of the episode
- prestorage (on/off) [?]
- EGM source
- cardiac history
- cardiac surgery (if applicable)
- arrhythmia history
- date of implant for the ICD 7218
- when a previous brady or tachy device was explanted (if applicable)
- ejection fraction
- antiarrhythmic medication taken at the time of each episode
- NYHA classification
The letter "A" was used in the cells of the spreadsheet to indicate missing data.
Rhythm Classification Criteria
For each sequence pair, the detected episode was classified as VT, VF or SVT by comparing the its associated electrogram with that of the (presumed) sinus rhythm (SR) electrogram. Five morphologic characteristics (shape, components crossing the baseline, initial slope, largest component, and notching) were used to classify the detected episodes; if any of these differed between the detected episode's EGM and the paired SR EGM, the episode was classified as a VT or VF. If the two EGMs were similar in all five characteristics, but the detected rhythm slowed after anti-tachycardia pacing (ATP) therapy, with an increase in R-R intervals of at least 60 ms, the detection was considered appropriate. If all characteristics were similar, but the ICD delivered a shock, the episode was classified as an SVT.
The morphologic characteristics are illustrated below.
Shape: Successive beats in the EGM may be similar, differing, or in multiple sets of similar beats.
|Monomorphic, same basic shape from beat-to-beat
|Polymorphic, the episode has many different shapes
|Multimorphic, the episode has at least two distinct morphologies that may or may not be similar to SR
Number of components that cross the baseline: Each component must be at least 10% of the peak-to-peak amplitude.
The Initial slope may be positive or negative.
|First component positive
|First component negative
Position of the largest component: The largest component may be first, second, third, etc. In the example below, at left the second component is largest; at right, the first is largest.
Notching: A component (defined as a portion of the signal that does not cross the baseline) may contain a notch. In the example below, the waveforms have a notch in the second component.
Description of the Excel Files
In Excel (version 5.0), all VT / VF episodes and the Most Recent interval files were saved in a text format and are located in column A. Each VT/VF episode file contain a total of 1024 cells of the R-R intervals. As you look at the episode files in Excel, the last cell in the spreadsheet for the detected episodes, are the intervals just prior to detection. For example, in episode file 0003.vt5, the last interval prior to detection is 390 ms which would be found in column A, cell number 1024. The detected episodes are stored only until the initial detection by the ICD (not episode termination). The Most Recent interval files have less than 1024 intervals (range from 986-1022). The least recent intervals are lost during the interrogation process and the number of intervals lost is variable.
This database was contributed to Physionet by Medtronic, Inc., for use by researchers to study precursors and possible predictors of spontaneous ventricular tachyarrhythmias. The original description was prepared by Ann M. Pearson and Walter H. Olson of Medtronic; it has been edited for presentation here.
To facilitate comparisons in the literature, please do not alter this database or the rhythm classifications.
When referencing this database, please refer to it as the "Spontaneous Ventricular Tachyarrhythmia Database Version 1.0 from Medtronic, Inc." and cite its URL (http://physionet.org/physiobank/database/mvtdb/).
If errors are discovered, please inform us so we can investigate and communicate them to other users of this database. We have made our best effort to provide accurate data, but neither we nor Medtronic, Inc. are liable for the consequences of possible errors or omissions in these data. The user assumes all such responsibility for the use of these data.
- Swarna US, Callans DJ, Schwartzman D, Gottlieb CD, Marchlinski, FE. The value of right ventricular endocardial electrogram recordings in discriminating different morphologies of ventricular tachycardia. Address for copies: David Schwatzman, MD, Philadelphia Heart Institute, Philadelphia, PA 19104.
Please include the standard citation for PhysioNet: (show more options)
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.