Next: The Records Up: A Database for Evaluation of Algorithms for Previous: Introduction

The QT Database

The QT Database includes ECGs which were chosen to represent a wide variety of QRS and ST-T morphologies, in order to challenge QT detection algorithms with real-world variability. The records were chosen primarily from among existing ECG databases, including the MIT-BIH Arrhythmia Database[3], the European Society of Cardiology ST-T Database[4] (courtesy of Prof. Carlo Marchesi), and several other ECG databases collected at Boston's Beth Israel Deaconess Medical Center. The existing databases are an excellent source of varied and well characterized data, to which we have added reference annotations marking the location of waveform boundaries. The additional recordings were chosen to represent extremes of cardiac (patho)physiology. We gathered new data from Holter recordings of patients who experienced sudden cardiac death during the recordings, and age-and-gender matched patients without diagnosed cardiac disease.

The QT Database contains a total of 105 fifteen-minute excerpts of two channel ECGs, selected to avoid significant baseline wander or other artifacts. (We note that a consequence of this selection procedure is that heart rates during these excerpts tend to be relatively low, probably since higher rates are frequently associated with noisy signals that would have failed to satisfy our selection criteria.) Table 1 shows the sources of the data.

Table 1: Distribution of the 105 records according to the original Database.
Arrhyt. ST DB Sup. Vent. Long Term STT NSR DB Death
15 6 13 4 33 10 24

Within each record, between 30 and 100 representative beats were manually annotated by cardiologists, who identified the beginning, peak and end of the P-wave, the beginning and end of the QRS-complex (the QRS fiducial mark, typically at the R-wave peak, was given by an automated QRS detector), the peak and end of the T-wave, and (if present) the peak and end of the U-wave. In order to permit the study of beat-to-beat variations such as alternans, 30 consecutive beats of the dominant morphology were annotated in each case if possible. In records with significant QRS morphology variation, up to 20 beats of each non-dominant morphology were also annotated. Only ``normal'' beats were annotated. The criteria for beat selection were: 1) all are classified as ``normal'' by ARISTOTLE[5], and 2) the preceding and following beats are also normal. Beats were only annotated during the final 5 minutes of the excerpts in order to allow analysis algorithms a minimum of 10 minutes for learning. In all, 3622 beats have been annotated by cardiologists. These annotations have been carefully audited to eliminate gross errors, although the precise placement of each annotation was left to the judgment of the expert annotators. The current edition of the QT Database includes two independently derived sets of annotations for 11 records (to permit study of inter-observer variability). The remaining 94 records contain only a single set of expert annotations.

All records were sampled at 250 Hz. Those which were not originally sampled at that rate were converted using xform, an application from the MIT Waveform Database Software Package [3].

Next: The Records Up: A Database for Evaluation of Algorithms for Previous: Introduction