{
  "fetched_at": "2026-10-18T13:32:08Z",
  "source_url": "https://www.kaggle.com/competitions/bhf-data-science-centre-ecg-challenge/data",
  "pages": [
    {
      "name": "data-description",
      "content": "This challenge will use the [GenECG dataset](https://www.medrxiv.org/content/10.1101/2023.12.27.23300581v1 ). This dataset contains synthetically generated ECG images which have been created to incorporate artifacts common on paper ECG images which are typically scanned or photographed. The GenECG dataset contains 21,799 ECG images which have passed a clinical Turing test where expert observers could not distinguish between synthetic and real ECGs. The wave form data to create this dataset is part of the [PTB-XL dataset](https://www.nature.com/articles/s41597-020-0495-6) containing diagnoses that have been adjudicated by two cardiologists.\n\n**Images**\nThe dataset consists of ECG images as .png format. The dataset has been divided into **training images (train_000000.png, train_000001.png, onwards)** and **testing images (test_000000.png, test_000001.png, onwards)**. An external testing dataset has also been withheld for final testing of the models. \n\n**Diagnoses**\nThe GenECG and PTB-XL dataset contain ECG images with a wide variety of diagnoses. For the purpose of this challenge we would like you to identify 5 important diagnoses. These represent the four key superclasses of diagnosis, and the most common rhythm abnormality. The diagnoses which you will be asked to identify in the dataset are as follows:\n-\tMyocardial infarction. This is a medical emergency which happens when the blood supply to the heart is blocked. https://www.bhf.org.uk/informationsupport/conditions/heart-attack\n-\tAtrial fibrillation. This is caused by irregular electrical signals in the heart which leads to the heart beating irregularly and often too fast. https://www.bhf.org.uk/informationsupport/conditions/atrial-fibrillation\n-\tHypertrophy. This is an enlargement of the heart muscle which can have a number of causes including high blood pressure, heart valve disease, and inherited conditions. \n-\tConduction disturbance. This is a problem with the electrical signals that control the heart's rhythm which leads to an irregularity in the heart rhythm. \n-\tST/T changes. This is an abnormality of the wave pattern of electrical signals from the heart where there is a shift in the shape or height of certain lines on an ECG.\n\nThe diagnoses are provided in the **train_final.csv** file (with ID 0, 1, onwards). An ECG can none, one or more than one of these diagnoses. Other diagnoses may be present on the images, but they are not considered as part of this challenge. The train_final.csv file is structured as follows. An example of the format of the submission can be found in sample_submission.csv.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F96d0c4d757419dd949b255b18b1e2d35%2F2024_Kaggle_data.png?generation=1728117655968153&alt=media)\n\n**Example images**\nExamples of these diagnoses are shown in the images below:\n\nNormal ECG:\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F78c344d6f3f02fc5e3b7d094cf5b98e4%2F000000_NORMAL.jpg?generation=1728310607438247&alt=media)\n\nST/T changes (STTC):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F0954f70c7ed29c52c3ca23a64e527110%2F000088_STTC.jpg?generation=1728310672156698&alt=media)\n\nHypertrophy (HYP):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2Fb86a7de0ecf2339436e9d7843170272f%2F000032_HYP.jpg?generation=1728310715264829&alt=media)\n\nMyocardial infarction (MI): \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F79b71563507d38c0db2a3dc70d16abb6%2F000183_MI.jpg?generation=1728310744883264&alt=media)\n\nAtrial fibrillation (AF):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F90f4bb98b69c1aca74169eb64bb07d13%2F001703_AF.jpg?generation=1728310838020668&alt=media)\n\nAtrial Fibrillation (AF), left ventricular hypertrophy (HYP), ST/T changes (STTC),  myocardial infarction (MI) and conduction delay (CD):\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F1928d9e710f2b3f31440b2dcd12fc692%2F002034%20STTC%20HYP%20MI%20CD%20AF.jpg?generation=1728310899837340&alt=media)\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "Heart and circulatory disease (also called cardiovascular diseases) are the commonest cause of death around the world. It is estimated that 1 in 3 people are living with cardiovascular disease and since 1997 the number of people living with cardiovascular disease around the world has doubled. More information about heart and circulatory disease can be found on the British Heart Foundation (BHF) [website](https://www.bhf.org.uk/-/media/files/for-professionals/research/heart-statistics/bhf-cvd-statistics-global-factsheet.pdf?rev=5ae062031a884f1d9a98d3f9ecd08804&hash=0670F13FCB5E0BF4F2CB149D24F794F3). \n \nFor many people with suspected cardiovascular disease the first test that they have is an electrocardiogram (also called an electrokardiogram, ECG or EKG).  An ECG is a simple test which records the electrical activity of the heart, including its rate and rhythm. More information about ECGs can be found [here](https://www.bhf.org.uk/informationsupport/tests/ecg).\n\nChanges in the ECG appearance can be used to detect a variety of heart diseases. During an ECG the electrical activity of the heart is recorded, and this is displayed graphically as a series of waves and intervals. Around the world access to digital tools varies, and frequently ECG images are displayed on a screen or printed out. Existing artificial intelligence/machine learning tools which have been developed for waveform ECG data may perform poorly on ECG images. Developing tools to automate the analysis of ECG images will improve patient care by helping to improve the speed and accuracy of diagnoses. \n\nThe [British Heart Foundation Data Science Centre](https://www.hdruk.ac.uk/helping-with-health-data/bhf-data-science-centre/) is organising this challenge in collaboration with researchers from the [University of Edinburgh](https://ecci-cvs.github.io). The challenge aims to assess whether artificial intelligence/machine learning can be used to identify and classify disorders of heart rhythm from synthetic ECG images which have been created to incorporate common artefacts. \n\nThis challenge will use the [GenECG dataset](https://www.medrxiv.org/content/10.1101/2023.12.27.23300581v1). This dataset contains synthetically generated ECG images which have been created to incorporate artefacts common on paper ECG images which are typically scanned or photographed. The wave form data to create this dataset is part of the [PTB-XL dataset](https://www.nature.com/articles/s41597-020-0495-6) containing diagnoses that have been adjudicated by two cardiologists. The GenECG dataset contains 21,799 ECG images which have passed a clinical Turing test where expert observers could not distinguish between synthetic and real ECGs. \n\nThis challenge will focus on the identification of five important heart disease diagnoses. These include myocardial infarction, atrial fibrillation, hypertrophy, conduction disturbance, and ST/T change. Further information on these diagnoses can be found in the data section.\n\nThe top scoring teams will be invited to participate in a British Heart Foundation Data Science Centre webinar about the challenge and the best performing will be invited to present their results at the British Cardiovascular Society (BCS) meeting in June 2025. Participation in these activities will require code sharing.\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "Submissions will be evaluated with [area under the ROC curve metrics](http://en.wikipedia.org/wiki/Receiver_operating_characteristic). These will be averaged across the labels to be assigned to each ECG image to create the final score.   \n\n---\n\n## Submission File\nFor each ID in the test set, you must predict the presence or absence of each of the five labels. These are myocardial infarction (MI), ST/T changes (STTC), conduction delay (CD), hypertrophy (HYP) and atrial fibrillation (AF). The submission file should contain a header and have the following format:\n\n    ID,MI,STTC,CD,HYP,AF\n    1,0,0,0,0,0\n    2,1,0,0,0,0\n    3,1,1,0,0,0\n    etc.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "Competitors must not attempt to gain an advantage over other competitors through malicious or unfair methods.\n\nOne account per participant. You cannot submit from multiple Kaggle accounts. \n\nNo private sharing of code outside of teams. You can share code with other participants via the forum.\n\nCompetitors must adhere to the licence rules, which the GenECG dataset is subject to. The full list of terms and conditions are listed here: https://physionet.org/content/ptb-xl/view-license/1.0.3/\nData provided for this challenge are for non-commercial purposes only. \nAny results generated through use of the dataset must acknowledge its use in accordance with the attribution requirements.\n\nParticipants in the competition must not make copies of the data for use outwith the challenge or attempt to redistribute the data. \nCompetitors must not make any attempt to re-identify the participant's whose data was used to derive the original dataset. \n\nThe top scoring teams will be invited to participate in a British Heart Foundation Data Science Centre webinar about the challenge and the best performing will be invited to present their results at the British Cardiovascular Society (BCS) meeting in June 2025. Participation in these activities will require code sharing. Selection for participation in these activities is at the discretion of the British Heart Foundation Data Science Centre. Decisions of the British Heart Foundation Data Science Centre are final.\n\nParticipants must not use the BHF Data Science Centre branding, logo or collateral without explicit permission from the BHF Data Science Centre.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "The aim of this challenge is to create models which can be used to identify and classify important disorders of the heart from electrocardiogram (ECG, EKG) images. The dataset includes synthetic images of ECGs which simulate artefacts which might occur when taking a photo of the ECG with a smart phone. Models to automate diagnoses from these images could be used to speed up heart disease care around the world. ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Data Dictionary",
      "content": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F96d0c4d757419dd949b255b18b1e2d35%2F2024_Kaggle_data.png?generation=1728117655968153&alt=media)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Example image",
      "content": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F22715854%2F844c74ee48c3c8af60be3b0eacdd46e3%2F000000_NORMAL.jpg?generation=1728310494309672&alt=media)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}