ECG or heartbeat datasets

I have gone through all possible open source ECG datastes available for classification problem. The idea of this blog post is to share this useful information with research community.  It includes the details about the open source ECG datasets i have found and are useful to train a model. You guys can also share some other resources for good. Just drop the link in the comment section below. 

1-  Physionet datasets (physionet.org)

Freely available for the research community. They have lots of datasets with an implemented Matlab “toolbox” to import the data directly into Matlab.

2- ECG datasets from Time series classification (http://www.timeseriesclassification.com)

Dataset available in various formats such as Weka ARFF, simple text files and sktime ts file format. Data is suitable to use for univariate and multivariate classification problems. More details can be obtained from here.

3- ECG signals (1000 fragments)

This dataset is also obtained from PhysioNet (http://www.physionet.org)  MIT-BIH Arrhythmia database. 

  1. ECG signals are from 45 patients: 19 female (age: 23-89) and 26 male (age: 32-89).
  2. It contains 17 classes: normal sinus rhythm, pacemaker rhythm, and 15 types of cardiac dysfunctions
  3. All ECG signals were recorded at a sampling frequency of 360 [Hz] and a gain of 200.
  4. For the analysis, 1000, 10-second (3600 samples) fragments of the ECG signal (not overlapping) were randomly selected.
  5. Only signals derived from one lead, the MLII, were used.
  6. Data are in mat format (Matlab) and you can download it from here.

4- D1NAMO dataset

Dataset is taken using wearable device (Zephyr Bioharness 3). Data obtained from 29 people where 20 are healthy and 9 are type-1 diabetes patients [1]. Dataset consist of ECG signals, breathing signals, accelerometer outputs, Glucose measurements, and food pictures & annotations by a dietitian [1].  Data size is around 9.5 GB and you can download it from here.

5- Arrhythmia Data Set

This dataset is used in [2] to distinguish between the presence and absence of cardiac arrhythmia. The dataset contain 16 different classes where one is the normal sinus rhythm and 15 others are different classes of arrhythmia. You can download dataset from here.

6- ECG Heartbeat Categorization Dataset

This dataset consist of segmented and pre-processed ECG signals for heartbeat classification [3]. The dataset is used in [3] to classify normal and abnormal heart beat from a single heart beat. You can download dataset in csv file format from here

7- CSRC ECG datasets

CSRC ECG datasets is available freely but need approval from CSRC (a public-private partnership). A scientific oversight committee is responsible to evaluate the proposals for use of the released ECG data and to foster collaboration within the research community. You can read more information about this dataset from here.

8- ECG view dataset

Openly available for academic use. You can send them a request to obtain the data. They will be happy to collaborate with you.  They also developed a tool for extracting PDF from the ECG signals. Visit ecgview for more details.

9- Wafer and ECG time series dataset

This might be useful to test your algorithm developed to classify the normal and abnormal sequences. Wafter and ECG time series datasets are available here.

There are some other datasets such as sleep heart health study dataset and dataset from biobank  “datasets containing genuine data can only be accessed by authorized researchers who are logged onto this system”. 

References

[1] Dubosson F, Ranvier JE, Bromuri S, Calbimonte JP, Ruiz J, Schumacher M. The open D1NAMO dataset: A multi-modal dataset for research on non-invasive type 1 diabetes management. Informatics in Medicine Unlocked. 2018 Jan 1;13:92-100.

[2] H. Altay Guvenir, Burak Acar, Gulsen Demiroz, Ayhan Cekin “A Supervised Machine Learning Algorithm for Arrhythmia Analysis.” Proceedings of the Computers in Cardiology Conference, Lund, Sweden, 1997.

[3]. Mohammad Kachuee, Shayan Fazeli, and Majid Sarrafzadeh. “ECG Heartbeat Classification: A Deep Transferable Representation.” arXiv preprint arXiv:1805.00794 (2018)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.