<div class="sc-cmRAlD dkqmWS"><div class="sc-UEtKG dGqiYy sc-flttKd cguEtd"><div class="sc-fqwslf gsqkEc"><div class="sc-cBQMlg kAHhUk"><h2 class="sc-dcKlJK sc-cVttbi gqEuPW ksnHgj">About Dataset</h2></div></div></div><div class="sc-jgvlka jFuPjz"><div class="sc-gzqKSP tNtjD"><div style="min-height: 80px;"><div class="sc-etVRix jqYJaa sc-bMmLMY ZURWJ"><p><strong><em>see the introduction kernel on how to access data and use annotation labels</em></strong></p>

<h3>CONTEXT</h3>

<p>This dataset contains Lead II signal (with annotations) of 201 records collected from following 3 databases available on PhysioNet under open access:</p>

<ol>

<li>MIT-BIH Arrhythmia Database [<strong>mitdb</strong>]</li>

<li>MIT-BIH Supraventricular Arrhythmia Database [<strong>svdb</strong>]</li>

<li>St Petersburg INCART 12-lead Arrhythmia Database [<strong>incartdb</strong>]</li>

</ol>

<h3>NOTE</h3>

<blockquote>

  <ol>

  <li>All signals have been resampled to 128Hz and gain has been removed.</li>

  <li>Baseline wander has been removed using Median Filtering.</li>

  <li>Denoising was NOT used.</li>

  <li>Signal data and annotation labels have been saved in numpy (.npy) format.</li>

  <li>All Signals are nearly 30 mins long.</li>

  </ol>

</blockquote>

<h3>CONTENT</h3>

<p>Data has been organised as follows:</p>

<p>parent directory <em>db_npy</em> contains 3 sub-directories each of which represent one database</p>

<blockquote>

  <p>mitdb_npy has 48 records</p>

  <p>svdb_npy has 78 records</p>

  <p>incartdb_npy has 75 records</p>

</blockquote>

<p>Each of these database directory contains a '<strong><em>RECORDS</em></strong>' file that lists the ecg records available in that database.</p>

<p>Each record has 3 files associated with it:</p>

<blockquote>

  <ol>

  <li><p><em>rec</em>_BEAT.npy: contains 'beat' annotations (R-peaks and its label) for the record.<br>

  <em>each record may have variable number of beats based on heart rate</em><br>

  <em>usually we shall be interested in beat annotation labels only. Each beat label represents one R-peak and hence one beat</em></p></li>

  <li><p><em>rec</em>_NBEAT.npy: contains 'non-beat' annotations (Other than R-peaks) for the record</p></li>

  <li><p><em>rec</em>_SIG_II.npy: contains the Lead 2 signal data of the record as a single numpy array</p></li>

  </ol>

</blockquote>

<p>(* see the introduction kernel on how to access data and use annotation labels *)</p>

<h3>Understanding Annotations:</h3>

<p>There are two types of annotations: Beat and Non-Beat annotations.<br>

Beat annotations are associated with each heart-beat. If you are working with heart-beat classifications then only Beat annotations shall be useful and Non-Beat annotations can be ignored.</p>

<p>Standard PhysioNet Annotations are described in <strong><em>db_npy/annotations.txt</em></strong> file. These are common across all databases.<br>

This file has 3 columns <br>

Column 1: Label<br>

Column 2: Type of label [ b=beat annotation; n=non-beat annotation ]<br>

Column 3: Description</p>

<p><img alt="" src="https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5417013%2F43781477b37cdca4adb187e4b39c3648%2Fstdlab.png?generation=1593929940178708&amp;alt=media"></p>

<p>There are 19 Beat annotations and 22 Non-Beat annotations. However, not all annotations may occur in data files. For example, the Label 'r' does not occur even once in any three of the database but yet its the part of standard PhysioNet labels. (might be in use in some other database). It's advised to do a full annotation count before working with data.</p>

<p>According to AAMI recommendation, each beat is classified into one of the 5 types [ N, V, S, F, Q ]. However, you are free to choose any classification strategy.</p>

<p><img alt="" src="https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5417013%2F58397c44c2e877ddb61418270eb2f3fd%2FRelation-between-MIT-BIH-heartbeats-and-AAMI-standards.png?generation=1593940064290748&amp;alt=media"></p>

<h3>IMPORTANT</h3>

<blockquote>

  <ol>

  <li><p>mitdb's record '102' and '104' DO NOT have lead 2 signal available hence files '102_SIG_II.npy' and '104_SIG_II.npy' are not present. However, they have BEAT and NBEAT files present. Its advised not to use those two records</p></li>

  <li><p>mitdb's record '102', '104', '107' and '217' are paced records</p></li>

  <li><p>mitdb's record '207' is the only record with 'Flutter' waves that are not marked by beat-annotations (no R-peaks marked). However, they are marked by non-beat annotations.</p></li>

  </ol>

</blockquote>

<h3>Acknowledgements</h3>

<p>**PhysioNet **[<a rel="noreferrer nofollow" aria-label="https://physionet.org/ (opens in a new tab)" target="_blank" href="https://physionet.org/">https://physionet.org/</a>]<br>

MLA    Goldberger, A., et al. "PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220." (2000).<br>

APA        Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., … &amp; 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.<br>

Chicago        Goldberger, A., L. Amaral, L. Glass, J. Hausdorff, P. C. Ivanov, R. Mark, J. E. Mietus, G. B. Moody, C. K. Peng, and H. E. Stanley. "PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220." (2000).<br>

Harvard        Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P.C., Mark, R., Mietus, J.E., Moody, G.B., Peng, C.K. and 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.<br>

Vancouver    Goldberger A, Amaral L, Glass L, Hausdorff J, Ivanov PC, Mark R, Mietus JE, Moody GB, Peng CK, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.</p>

<p><strong>MIT-BIH Arrhythmia Database</strong><br>

Moody GB, Mark RG. The impact of the MIT-BIH Arrhythmia Database. IEEE Eng in Med and Biol 20(3):45-50 (May-June 2001). (PMID: 11446209) : Hosted at <a rel="noreferrer nofollow" aria-label="https://physionet.org/content/mitdb/1.0.0/ (opens in a new tab)" target="_blank" href="https://physionet.org/content/mitdb/1.0.0/">https://physionet.org/content/mitdb/1.0.0/</a></p>

<p><strong>MIT-BIH Supraventricular Arrhythmia Database</strong><br>

Greenwald SD. Improved detection and classification of arrhythmias in noise-corrupted electrocardiograms using contextual information. Ph.D. thesis, Harvard-MIT Division of Health Sciences and Technology, 1990 : Hosted at <a rel="noreferrer nofollow" aria-label="https://physionet.org/content/svdb/1.0.0/ (opens in a new tab)" target="_blank" href="https://physionet.org/content/svdb/1.0.0/">https://physionet.org/content/svdb/1.0.0/</a></p>

<p><strong>St Petersburg INCART 12-lead Arrhythmia Database</strong><br>

Hosted at <a rel="noreferrer nofollow" aria-label="https://physionet.org/content/incartdb/1.0.0/ (opens in a new tab)" target="_blank" href="https://physionet.org/content/incartdb/1.0.0/">https://physionet.org/content/incartdb/1.0.0/</a></p>

<p><em>DOI</em><br>

<a rel="noreferrer nofollow" aria-label="https://doi.org/10.13026/C2F305 (opens in a new tab)" target="_blank" href="https://doi.org/10.13026/C2F305">https://doi.org/10.13026/C2F305</a><br>

<a rel="noreferrer nofollow" aria-label="https://doi.org/10.13026/C2V30W (opens in a new tab)" target="_blank" href="https://doi.org/10.13026/C2V30W">https://doi.org/10.13026/C2V30W</a><br>

<a rel="noreferrer nofollow" aria-label="https://doi.org/10.13026/C2V88N (opens in a new tab)" target="_blank" href="https://doi.org/10.13026/C2V88N">https://doi.org/10.13026/C2V88N</a></p></div></div></div>