Next: Manual Annotations Up: A Database for Evaluation of Algorithms for Previous: The QT Database

The Records

The records are provided in the MIT-BIH Database format. QT Database records have record names of the form selnnnn, where nnnn is the name of the original record in the source database. Each record includes a signal file (record.dat, where record is the record name), a header file (record.hea, describing the format of the signal file), and several annotation files: record.atr contains the original annotations from the source database; record.ari contains QRS annotations obtained automatically by ARISTOTLE[5]; record.pu0 and record.pu1 contain the automatic waveform onsets and ends in signals 0 and 1 respectively, as detected using the differentiated threshold method by ecgpuwave[6]; and record.man, record.qtn, and record.qnc contain the manual annotations, as described in the next section, for annotators n=1,2. The record.atr files do not exist for the 24 sudden death records, since no reference annotations were available in this case. ARISTOTLE classifies the detected beats as normal (N) or belonging to other beat types[5]. Those beats considered normal are clustered into different groups specified by the num fields of the ari annotations. In the num fields of the pu* annotations, ecgpuwave classifies the T waves as normal (0), inverted (1), only upwards (2), only downwards (3), biphasic negative-positive (4), or biphasic positive-negative (5). Waveform onset ``('' and end ``)'' annotations specify the waveform type in their num fields (0 for a P-wave, 1 for a QRS complex, 2 for a T wave, or 3 for a U-wave).

Table 2 documents the make-up of the QT Database, showing the record name, the source database, the time interval excerpted (shown relative to the beginning of the original record in the source database), the number of manually annotated beats, and the pattern of the dominant waveform in each record.

Table 2: Data excerpts used in the QT Database
Table 2. A: Records from MIT-BIH Arrhythmia Database
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sel100 7:00 $\rightarrow$ 22:00 30 ( p ) ( N ) t )
sel102 6:00 $\rightarrow$ 21:00 85 ( N ) t ) u )
sel103 1:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t ) u )
sel104 15:00 $\rightarrow$ 30:00 77 ( N ) t )
sel114 0:00 $\rightarrow$ 15:00 50 ( p ) ( N ) t )
sel116 0:00 $\rightarrow$ 15:00 50 ( p ) ( N ) t ) u )
sel117 15:00 $\rightarrow$ 30:00 30 ( p ) ( N ) t ) u )
sel123 3:00 $\rightarrow$ 18:00 30 ( p ) ( N ) t ) u )
sel213 0:00 $\rightarrow$ 15:00 71 ( p ) ( N ) t )
sel221 0:00 $\rightarrow$ 15:00 30 ( N ) t )
sel223 13:30 $\rightarrow$ 28:30 30 ( p ) ( N ) ( t )
sel230 7:00 $\rightarrow$ 22:00 50 ( p ) ( N ) t )
sel231 3:00 $\rightarrow$ 18:00 50 ( p ) ( N ) t )
sel232 15:00 $\rightarrow$ 30:00 30 ( A ) t )
sel233 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t )
15 3:45:00 673

Table 2. B: Records from MIT-BIH ST Change Database
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sel301 14:00 $\rightarrow$ 29:00 30 ( p ) ( N ) ( t )
sel302 8:00 $\rightarrow$ 23:00 30 ( p ) ( N ) t )
sel306 0:00 $\rightarrow$ 15:00 36 ( p ) ( N ) t )
sel307 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t )
sel308 13:00 $\rightarrow$ 28:00 50 ( p ) ( N ) t )
sel310 4:00 $\rightarrow$ 19:00 30 ( N ) t ) u )
6 1:30:00 206

Table 2. C:
Records from MIT-BIH Supraventricular Arrhythmia Database
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sel803 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t )
sel808 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t ) u )
sel811 7:30 $\rightarrow$ 22:30 30 ( p ) ( N ) t )
sel820 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t )
sel821 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) ( t )
sel840 14:50 $\rightarrow$ 29:50 70 ( p ) ( N ) ( t )
sel847 14:22 $\rightarrow$ 29:22 33 ( p ) ( N ) t )
sel853 5:00 $\rightarrow$ 20:00 30 ( p ) ( N ) t )
sel871 4:00 $\rightarrow$ 19:00 70 ( p ) ( N ) t )
sel872 0:00 $\rightarrow$ 15:00 30 ( p ) ( N ) t )
sel873 0:00 $\rightarrow$ 15:00 33 ( p ) ( N ) t )
sel883 13:30 $\rightarrow$ 28:30 30 ( p ) ( N ) t )
sel891 14:59 $\rightarrow$ 29:59 71 ( p ) ( N ) t )
13 3:15:00 517

Table 2. D: Records from MIT-BIH Normal Sinus Rhythm Database
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sel16265 10:37:30 $\rightarrow$ 10:52:30 30 ( p ) ( N ) t )
sel16272 11:13:10 $\rightarrow$11:28:10 30 ( p ) ( N ) t )
sel16273 11:22:36.5 $\rightarrow$ 11:37:36.5 30 ( p ) ( N ) t )
sel16420 14:24:00 $\rightarrow$ 14:39:00 30 ( p ) ( N ) t )
sel16483 13:32:00 $\rightarrow$ 13:47:00 30 ( p ) ( N ) t )
sel16539 20:56:00 $\rightarrow$ 21:11:00 30 ( p ) ( N ) ( t )
sel16773 10:23:40 $\rightarrow$ 10:38:40 30 ( p ) ( N ) t ) u )
sel16786 18:22:00 $\rightarrow$ 18:37:00 30 ( p ) ( N ) ( t )
sel16795 15:00:00$\rightarrow$ 15:15:00 30 ( p ) ( N ) ( t )
sel17453 14:15:00 $\rightarrow$ 14:30:00 30 ( p ) ( N ) ( t ) u )
10 2:30:00 300

Table 2. E: Records from European ST-T Database
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sele0104 1:35:00 $\rightarrow$ 1:50:00 30 ( p ) ( N ) t )
sele0106 1:31:00 $\rightarrow$ 1:46:00 30 ( p ) ( N ) ( t ) u )
sele0107 1:9:30 $\rightarrow$ 1:24:30 34 ( p ) ( N ) t ) u )
sele0110 1:32:00 $\rightarrow$ 1:47:00 30 ( p ) ( N ) t )
sele0111 1:19:00 $\rightarrow$ 1:34:00 30 ( p ) ( N ) t ) u )
sele0112 1:32:10 $\rightarrow$ 1:47:10 50 ( p ) ( N ) t ) u )
sele0114 1:38:00 $\rightarrow$ 1:53:00 30 ( p ) ( N ) ( t ) ( u )
sele0116 00:37:30 $\rightarrow$ 00:52:30 30 ( p ) ( N ) ( t )
sele0121 1:07:30 $\rightarrow$ 1:22:30 30 ( p ) ( N ) t )
sele0122 40:00 $\rightarrow$ 55:00 30 ( p ) ( N ) t ) u )
sele0124 1:41:00 $\rightarrow$ :56:00 50 ( p ) ( N ) ( t )
sele0126 1:36:40 $\rightarrow$ 1:51:40 30 ( p ) ( N ) t ) u )
sele0129 00:26:30$\rightarrow$ 00:41:30 30 ( p ) ( N ) t )
sele0133 00:37:30 $\rightarrow$ 00:52:30 30 ( p ) ( N ) t )
sele0136 00:56:00 $\rightarrow$ 1:11:00 30 ( p ) ( N ) ( t )
sele0166 00:52:30 $\rightarrow$ 1:07:30 36 ( p ) ( N ) ( t )
sele0170 00:22:30$\rightarrow$ 00:37:30 30 ( p ) ( N ) t )
sele0203 00:52:30 $\rightarrow$ 1:07:30 30 ( p ) ( N ) t )
sele0210 1:45:00 $\rightarrow$ 2:00:00 30 ( p ) ( N ) ( t )
sele0211 1:45:00 $\rightarrow$ 2:00:00 30 ( p ) ( N ) ( t )
sele0303 00:34:30 $\rightarrow$ 00:49:30 30 ( p ) ( N ) ( t )
sele0405 00:22:30 $\rightarrow$ 00:37:30 30 ( p ) ( N ) t )
sele0406 1:00:00$\rightarrow$ 1:15:00 31 ( p ) ( N ) t )
sele0409 1:20:00$\rightarrow$ 1:35:00 30 ( p ) ( N ) ( t )
sele0411 1:30:00 $\rightarrow$ 1:45:00 30 ( p ) ( N ) t )
sele0509 00:15:00 $\rightarrow$ 00:30:00 30 ( p ) ( N ) t )
sele0603 00:17:00 $\rightarrow$ 00:32:00 30 ( p ) ( N ) t ) u )
sele0604 00:00:30 $\rightarrow$ 00:15:30 30 ( p ) ( N ) t ) u )
sele0606 1:31:00 $\rightarrow$ 1:46:00 30 ( p ) ( N ) t )
sele0607 1:00:10 $\rightarrow$ 1:15:10 30 ( p ) ( N ) t )
sele0609 00:35:30 $\rightarrow$ 00:50:30 30 ( p ) ( N ) ( t )
sele0612 00:19:30 $\rightarrow$ 00:34:30 30 ( p ) ( N ) ( t )
sele0704 1:16:30 $\rightarrow$ 1:31:30 30 ( p ) ( N ) t ) u )
33 8:15:00 1041

Table 2. F: Records from ``sudden death'' patients from BIH
Record Original interval time annot Waveform Pattern
beats (p) (QRS) (t) (u)
sel30 7:39:30 $\rightarrow$ 7:54:30 30 ( p ) ( N ) ( t )
sel31 12:57:00 $\rightarrow$ 13:12:00 30 ( p ) ( N ) ( t )
sel32 20:52:20 $\rightarrow$ 21:07:20 30 ( p ) ( N ) ( t )
sel33 14:57:00 $\rightarrow$ 15:12:40 30 ( p ) ( N ) ( t )
sel34 5:53:00 $\rightarrow$ 6:08:00 30 ( p ) ( N ) ( t )
sel35 4:52:00 $\rightarrow$ 5:07:00 31 (AF) ( N )
sel36 18:43:00 $\rightarrow$ 18:58:00 31 ( B ) ( t ) u )
sel37 1:14:10 $\rightarrow$ 1:29:10 50 ( Q )
sel38 5:00:00 $\rightarrow$ 5:15:00 30 ( p ) ( N ) ( t )
sel39 00:04:30 $\rightarrow$ 00:19:30 30 ( p ) ( N ) ( t )
sel40 7:07:00 $\rightarrow$ 7:22:00 30 ( p ) ( N ) ( t )
sel41 00:05:00 $\rightarrow$ 00:20:00 30 ( p ) ( N ) ( t )
sel42 10:03:00 $\rightarrow$ 10:18:00 30 ( p ) ( N ) ( t )
sel43 08:00:00 $\rightarrow$ 08:15:00 30 ( p ) ( N ) ( t )
sel44 21:30:00 $\rightarrow$ 21:45:00 30 ( p ) ( N ) ( t )
sel45 17:30:30 $\rightarrow$ 17:45:30 30 ( p ) ( N ) ( t ) u )
sel46 03:10:00 $\rightarrow$ 03:25:00 30 ( p ) ( N ) ( t )
sel47 8:43:40 $\rightarrow$ 8:58:40 30 ( p ) ( N ) ( t ) ( u )
sel48 07:11:00 $\rightarrow$ 07:26:00 30 ( p ) ( N ) ( t )
sel49 23:10:40 $\rightarrow$ 23:25:40 30 ( p ) ( N ) ( t )
sel50 10:53:00 $\rightarrow$ 11:8:00 32 ( N ) ( t ) u )
sel51 00:55:00 $\rightarrow$ 01:10:00 30 ( p ) ( N ) ( t )
sel52 00:08:00 $\rightarrow$ 00:23:00 30 ( p ) ( N ) ( t )
sel17152 1:04:00 $\rightarrow$ 1:19:00 30 ( p ) ( N ) ( t )
24 6:00:00 744
 
Table 2. G: Records from MIT-BIH Long-Term ECG Database
Record Original interval time annot Waveform Pattern
    beats (p) (QRS) (t) (u)
sel14046 9:14:13 $\rightarrow$ 9:29:13 31 ( p ) ( N ) t )
sel14157 7:58:12 $\rightarrow$ 8:13:12 30 ( p ) ( N ) t ) u )
sel14172 17:38:43 $\rightarrow$ 17:53:43 50 ( p ) ( N ) ( t ) ( u )
sel15814 21:57:00 $\rightarrow$ 22:12:00 30 ( p ) ( N ) t )
4 1:00:00 141  
 

Holter records from the sudden death group are not calibrated with respect to amplitude; thus the signal gains recorded in the header files for these records are only estimates. Records sel102 and sel104 come from paced ECG recordings.


Next: Manual Annotations Up: A Database for Evaluation of Algorithms for Previous: The QT Database