<!--#set var="TITLE" value="Introduction to SEMIA, Version 3.0.1"-->
<!--#include virtual="/head.shtml"-->
<p>
SEMIA, version 3.0.1, is a tool for viewing time series of diagnostic and
morphology parameters of long-term ambulatory recordings, and ST
segment annotations with their corresponding ECG waveforms of the <a
href="../">Long-Term ST Database</a> (LTST DB). SEMIA was
created during the development of the LTST DB, a project supported by
Medtronic, Inc. (Minneapolis, MN, USA) and Zymed, Inc. (Camarillo, CA,
USA).
<h2>Compiling SEMIA</h2>
<p>
If you have not already installed the <a
href="/physiotools/wfdb.shtml">WFDB software package</a> and the
<a href="/physiotools/xview/">XView toolkit</a>, do so now.
<p>
If you are running under GNU/Linux, you may not need to compile SEMIA.
Try running the precompiled version of SEMIA available <a
href="semia">here</a>. (You will still need to have installed WFDB
and XView in order to do this successfully.)
<p>
If you need to compile SEMIA, make a directory for the SEMIA sources,
then copy all files from this directory into it. (For convenience,
you may download this <a href="../semia-3.0.1.tar.gz">gzip-compressed
tar archive</a> of all files in this directory rather than downloading
the individual files.)
<p>
Enter your directory of SEMIA sources and type '<tt>make</tt>'.
<p>
You may encounter many warnings related to improperly formatted
comments in XView #include files. These warnings may be ignored. If
you wish to eliminate them, download and apply this <a
href="/physiotools/wfdb/wave/xview-patches">patch</a> (read and follow
the instructions at the beginning of the patch to see how to apply
it).
<p>
When '<tt>make</tt>' finishes, copy <tt>semia</tt> into a directory in your
<b>PATH</b>, and copy <tt>semia.opt</tt> and <tt>semia.hlp</tt> into
the directory where you wish to use SEMIA.
<h2>Using SEMIA</h2>
<p>
To run SEMIA successfully, the following files of the LTST DB are needed in the
current working directory:
<ul>
<li> <tt>s<i>nnnnn</i>.hea</tt> (the header file for record s<i>nnnnn</i>).
This file also contains comments of human annotators, diagnostic data of
the patient, and technical data of the record.
<li> <tt>s<i>nnnnn</i>.dat</tt> (digitized ECGs for the record)
<li> <tt>s<i>nnnnn</i>.cnt</tt> (numbers of annotated ST events and their
durations according to annotation protocols A, B, and C, and numbers of
normal and abnormal heart beats)
<li> <tt>s<i>nnnnn</i>_fin.dmy</tt> (time series of "fine" diagnostic
and morphology parameters obtained on average heart beats)
<li> <tt>s<i>nnnnn</i>_raw.dmy</tt> (time series of "raw" (resampled - 0.5
samples/sec - and smoothed) diagnostic and morphology parameters)
<li> <tt>s<i>nnnnn</i>_1.sta</tt> (ST annotation markers according to
annotation protocol A)
</ul>
<p>
The last three of these files are contained within the <tt>.tsr.zip</tt>
files in the LTST DB directory on PhysioNet (or on the "Subset" CD-ROM).
Download the <tt>.tsr.zip</tt> file and unpack it using a command such as:
<pre>
unzip s20011.tsr.zip
</pre>
<p>
In addition, you will need copies of <tt>semia.opt</tt> (a text file containing
SEMIA's options) and <tt>semia.hlp</tt> (a text file containing on-line help
for SEMIA) in the current working directory, as noted above.
<p>
Run SEMIA by typing "<tt>semia</tt>". Click on the "Help" button to view
SEMIA's on-line manual. To open a record, click on "Open", then enter the
record name (<tt>s<i>nnnnn</i></tt> only; do not include <tt>.hea</tt> or any
other suffix in the record name). Depending on the speed of your system, it
may take up to a minute for SEMIA to load the data after you click "OK".
<p>
Once the data have been loaded, several buttons that were initially
inactive become available for use. Each of them ( "Lead0", "Lead1",
"Lead2", "Data", and "KL") opens a window when clicked. ("Lead2" remains
inactive unless you have opened a record with 3 ECG signals.)
<ul>
<li> The "Lead" windows display time series of diagnostic parameters
(raw and fine ST level functions, raw and fine ST slope functions,
raw linearly interpolated ST reference functions, raw ST deviation
functions, raw and fine heart rate functions, and annotated ST
episodes) and ST segment annotations according to annotation
protocol A of the LTST DB.
<p>
Fine time series or functions are those obtained on average heart
beats after noise detection during preprocessing phase. Raw time
series or functions are those after resampling and smoothing of
fine functions.
<p>
ST segment annotations are: global reference, local references
that define the ST reference function, ST segment annotations
indicating ischemic and heart-rate related ST episodes (beginnings,
extrema, ends), axis shift annotations, conduction change
annotations, noise annotations, and annotations indicating
unreadable intervals (beginnings, ends).
<p>
Three different display modes for "Lead" windows are possible:
<ul>
<li> Unsubtracted mode displaying ST level function (linearly interpolated
ST reference function is not subtracted from the ST level function).
In this mode, the active reference to compare ST segment waveforms is
the global reference annotation.
<li> Unsubtracted mode displaying ST level function (linearly interpolated
ST reference function is not subtracted from the ST level function).
In this mode, the active reference to compare ST segment waveforms is
a local reference annotation.
<li> Subtracted mode displaying ST deviation function (linearly interpolated
ST reference function is subtracted from the ST level function to form
the ST deviation function). In this mode, the active reference to
compare ST segment waveforms is the global reference annotation
</ul>
<p>
The ECG signal corresponding to current active reference annotation
is displayed in the upper part of the "Data" window.
<p>
Time series displayed in the "Lead" windows can be examined by
pointing them using mouse cursor or setting the <b>Marker</b> at the
time instance of interest. The corresponding ECG signal is displayed in
the lower part of the "Data" window.
<p>
ST segment annotations can be examined using the <b>Exm</b> option. The
examined ST annotation becomes the currently active annotation, and
its corresponding ECG signal is displayed in the lower part of the
"Data" window as well.
<p>
The time scale may be adjusted to permit between 1 minute and 96 hours
of data to be seen at once. Time series from "other" ECG leads may be
overlaid in each "Lead" window.
<li> The "Data" window displays the ECG signal corresponding to the time
of the active reference annotation in the upper portion of the window,
and that corresponding to the time of the current active annotation or
marker in the lower portion of the window. These two ECG waveforms may be
overlaid. A variety of scales may be selected for comparing ST segment
morphologies. The time scale may be adjusted to permit between 1 and 60
seconds of data to be seen at once.
<p>
The center heart beat in each ECG trace is the beat average over the chosen
time window. ECG signals from "other" ECG leads may be overlaid.
<li> The "KL" window displays time series of morphology (waveform shape)
parameters. Two sets of parameters (for the ST segment and the QRS complex)
can be displayed; the parameters are coefficients of a Karhunen-Loève
transform (principal components).
<p>
The ST segment or QRS complex distance function (Mahalanobis distance
function) can be overlaid in the KL window.
</ul>
<hr>
<h3>References:</h3>
<ol>
<li> Franc Jager, George B. Moody, Alessandro Taddei, Gorazd Antolic, Michele Emdin,
Ales Smrdel, Boris Glavic, Carlo Marchesi, and Roger G. Mark.
A Long-Term ST Database for Development and Evaluation of Ischemia Detectors.
<i>Computers in Cardiology 1998</i>, pp. 301-304, IEEE Press. ISSN 0276-6547.
<li> Franc Jager, Alessandro Taddei, Michele Emdin, Gorazd Antolic, Roman Dorn,
George B. Moody, Boris Glavic, Ales Smrdel, M Varanini, Mitja Zabukovec,
Simone Bordigiago, Carlo Marchesi, and Roger G. Mark.
The Long-Term ST Database: A Research Resource for Algorithm Development and
Physiologic Studies of Transient Myocardial Ischemia.
<i>Computers in Cardiology 2000</i>, pp. 841-844. IEEE Press. ISSN 0276-6547.</ol>