Long Term ST Database 1.0.0

File: <base>/legendre/doc/ltstdb_annotation_files.txt (8,407 bytes)


Here is updated description of the annotation files and ST annotation
codes (and related measurements) according to the annotation protocol
used for the records of the LTST DB.

Each record in the LTST database includes the following files:
(The format of the files is the MIT DB format.)

sXXXXX.dat   Data file   (Two channel ECG data sampled at 250 Hz.)
sXXXXX.hea   Header file
sXXXXX.ari   ARISTOTLE's file (QRS labels attached to each heart
             beat, plus rhythm changes.)
sXXXXX.atr   Atruth file (QRS labels as manually determined attached
             to each heart beat, plus rhythm changes.)
sXXXXX.sta   ST annotation atruth file (This is the ST ANNOTATION file.)
sXXXXX.16a   ST measurement atruth file
             (The ST measurements were derived from the AVERAGE heart
             beats. These average heart beats are composed for each
             single heart beat (those which passed the preprocessing
             phase as clean) from the 16 second time window surounding each
             heart beat. The ST measurements thus obtained are attached
             back to single heart beats.)

 XXXXX - the record number

1. sXXXXX.16a file: (What it contains):

   First, we computed average heart beats in the time windows of 16 seconds
   for each clean heart beat. Clean heart beats are those which passed
   filtering in the KL space during preprocesing phase. After that, we
   manually set the position of the point defining the isoelectric level
   and J-point for these average heart beats. Not for all of them, but
   very frequently, i.e., prior to ST episode, at the peak of ST episode,
   after the ST episode, prior to axis shift, after the axis shift,
   otherwise approximately every 20 minutes. This manual seting of the
   two fiducial points is performed during dummy annotation phase. The
   positions of the points defining the isoelectric level and J-points
   for the rest of average beats were computed using linear interpolation.
   After that, we computed a number of ST measurements for each average
   beat. These ST measurements were attached back to each single heart
   beat. So, each ST annotation in the file sXXXXX.16a is attached to a
   heart beat and contains ST measurements obtained on the corresponding
   average heart beat. Noisy heart beats are not attached with these
   annotations. They were judged as noisy during the preprocessing phase.
    Time index of each ST annotation: ARISTOTLEs fiducial point plus
    16 mili seconds.

    The format of ST annotation is:
               (a)    (b)    (c)    (d)    (e)    (f) 

                               (g)    (h)  (i)    (j)   (k)   (l)    (m)

  (a) ST amp.at J+80(60)(*)
  (b) ST amp.at J-point
  (c) ST amp.at J+20
  (d) ST amp.at J+40
  (e) ST amp.at J+60
  (f) ST amp.at J+80
  (g) ST amp.at J+100
  (h) ST amp.at J+120
  (i) Position of isoelectric level (mili seconds prior to ARISTOTLE's
      fiducial point)
  (j) Position of J-point (mili seconds after the ARISTOTLE's fiducial
  (k) Position of the point of measurement (mili seconds after the
      ARISTOTLE's fiducial point)
  (l) Number of heart beats left of current heart beat included into
      the average beat
  (m) Number of heart beats right of current heart beat included into
      the average beat (**)

  (The ST amplitude measurements (isoelectric level, J+XX ms) were
   obtained as means of five ECG samples (20 mili seconds) in the
   average waveforms.)
  x:   Lead number

  [DMY]: Whether or not the average beat corresponds with manually
         obtained measurements
         (none:  Does not, the measurements were obtained according
                 to prior interpolation of the positions of the points
                 defining the isoelectric level and J-points
          DMY:   The average beat corresponds to manually determined
                 measurements. The positions of the point defining
                 the isoelectric level and J-point were determined
                 manually for this average heart beat.)
 (*)  These measurments were obtained at the J-point plus 80 (or 60ms).
      The point of measurement was determined according to heart rate
      by the following rule:

      J+80   if  HR <  100 bpm
      J+72   if  HR >= 100 bpm   and   HR < 110 bpm
      J+64   if  HR >= 110 bpm   and   HR < 120 bpm
      J+60   if  HR >= 120 bpm

    These measurements (J+80(60)) also serve as a basis for the ST
    LEVEL FUNCTION which was used for ST annotating regarding ST
    episodes. These ST annotations are described below in item 2.
 (**) Thus, nleft + nright + 1 equals total number of heart beats
      included into this average.

2. sXXXXX.sta file (What it contains):

   ST annotations in this file mark: global reference, local
   reference, the beginning of an ST episode, its extrema, its end,
   noise, axis shift, or change due to QRS conduction change.

   The ST measurements, i.e., ST amplitudes at J+80(60) obtained on
   average heart beats are the basis for the raw ST LEVEL FUNCTION.
   The J+80(60) samples were then smoothed, resampled (0.5 Hz) and
   smoothed (7 point moving average - window of 12 seconds) thus
   forming smooth ST LEVEL FUNCTION (or just ST LEVEL FUNCTION).

   Next, we set a global reference and many local reference
   annotations in the ST LEVEL FUNCTION for each ECG lead throughout
   the record. By inserting the local references we tracked the
   ST REFERENCE FUNCTION of the ST segment level (reference ST
   level). After that, we subtracted the ST REFERENCE FUNCTION
   (piecewisely linearly approximated ST LEVEL FUNCTION from local
   reference to local reference) from the ST LEVEL FUNCTION. Thus,
   we ended in the ST DEVIATION FUNCTION (or subtracted function).

   Next, we annotated ISCHEMIC and RATE-RELATED episodes in the ST
   DEVIATION FUNCTION. We also annotated noises and unreadable
   intervals. Previously, during tracking the ST REFERENCE
   FUNCTIN, we annotated axis shifts and shifts due to QRS
   conduction changes.
Description of annotation codes:

  Time index:  Index of sample in the ST DEVIATION FUNCTION.
  ST annotation codes:

   The annotations could be in lower or in upper case characters.
   Upper case: Annotations were set with regard to global reference
               in the ST LEVEL FUNCTION. (These annotations are
               global reference and local references.)
   Lower case: Annotations were set with regard to the ST REFERENCE
               FUNCTION (reference ST level) in the ST DEVIATION
               (These are annotations for ischemic episodes,
                rate-related episodes, axis shifts, conduction
                change shifts, noises and unreadable events.)
  Code                Meaning
     GRST x           Global reference
     LRST x +- llll   Local reference
  s[cc]st x           Significant ST shift

  ([rt]st x +- dddd   Beginning of significant ST episode
  a[rt]st x +- dddd   Extrema of significant ST episode
   [rt]st x +- dddd)  End of significant ST episode

      noi x +- dddd   Noise
     (und x           Beginning of unreadable interval
      und x )         Beginning of unreadable interval

                  x:  lead number (0 or 1 or 2)

            +- llll:  ST level in microV
            +- dddd:  ST deviation in microV

               [cc]:  type of ST shift
              (none:  axis shift
                 cc:  conduction change)

               [rt]:  type of episode
              (none:  ischemic 
                 rt:  rate-related)

Critera for episodes:

    1.An episode begins when the magnitude of the ST deviation
      function first exceeds 50 uV; 
    2.The deviation must reach a magnitude of Vmin or more throughout
      a continuous interval of at least Tmin; 
    3.The episode ends when the deviation becomes smaller than 50 uV,
      provided that it does not exceed 50 uV in the following 30 seconds.

3. sXXXXX.stb file

    Vmin = 100 uV, Tmin = 30 s

4. sXXXXX.stc file

    Vmin = 100 uV, Tmin = 60 s