{
  "id": 614515,
  "title": "How are segmentation masks created",
  "url": "/competitions/physionet-ecg-image-digitization/discussion/614515",
  "author_name": "",
  "post_date": "2025-11-04T13:45:04.661682100Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi Everyone,\nI’m currently exploring how to train a U-Net architecture for ECG image segmentation. One key part I’m trying to understand better is how segmentation masks are actually built for ECG charts.</p>\n<p>Specifically, how do we derive or generate these masks from raw ECG images so that a U-Net can be trained effectively? Are there common approaches or preprocessing steps (like edge detection, Hough transforms, or pixel-level labeling) used to define the waveform regions?</p>\n<p>Would really appreciate any insights, papers, or practical examples that explain how mask generation for ECG data is typically done.</p>\n<p>Thanks in advance!</p>",
  "messages": [
    {
      "id": "3311213",
      "postDate": "11/04/2025 13:45:04",
      "content": "<p>Hi Everyone,\nI’m currently exploring how to train a U-Net architecture for ECG image segmentation. One key part I’m trying to understand better is how segmentation masks are actually built for ECG charts.</p>\n<p>Specifically, how do we derive or generate these masks from raw ECG images so that a U-Net can be trained effectively? Are there common approaches or preprocessing steps (like edge detection, Hough transforms, or pixel-level labeling) used to define the waveform regions?</p>\n<p>Would really appreciate any insights, papers, or practical examples that explain how mask generation for ECG data is typically done.</p>\n<p>Thanks in advance!</p>",
      "rawMarkdown": "Hi Everyone,\nI’m currently exploring how to train a U-Net architecture for ECG image segmentation. One key part I’m trying to understand better is how segmentation masks are actually built for ECG charts.\n\nSpecifically, how do we derive or generate these masks from raw ECG images so that a U-Net can be trained effectively? Are there common approaches or preprocessing steps (like edge detection, Hough transforms, or pixel-level labeling) used to define the waveform regions?\n\nWould really appreciate any insights, papers, or practical examples that explain how mask generation for ECG data is typically done.\n\nThanks in advance!",
      "votes": null
    },
    {
      "id": "3312766",
      "postDate": "11/07/2025 22:40:20",
      "content": "<p>We modified ecg-image-kit to generate segmentation masks.</p>\n<p><a href=\"https://github.com/Ahus-AIM/ecg-image-kit\" target=\"_blank\">https://github.com/Ahus-AIM/ecg-image-kit</a></p>",
      "rawMarkdown": "We modified ecg-image-kit to generate segmentation masks.\n\nhttps://github.com/Ahus-AIM/ecg-image-kit",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3312766,
      "author_name": "eliasstenhede",
      "author_url": "",
      "post_date": "11/07/2025 22:40:20",
      "content": "<p>We modified ecg-image-kit to generate segmentation masks.</p>\n<p><a href=\"https://github.com/Ahus-AIM/ecg-image-kit\" target=\"_blank\">https://github.com/Ahus-AIM/ecg-image-kit</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3311213": "Hi Everyone,\nI’m currently exploring how to train a U-Net architecture for ECG image segmentation. One key part I’m trying to understand better is how segmentation masks are actually built for ECG charts.\n\nSpecifically, how do we derive or generate these masks from raw ECG images so that a U-Net can be trained effectively? Are there common approaches or preprocessing steps (like edge detection, Hough transforms, or pixel-level labeling) used to define the waveform regions?\n\nWould really appreciate any insights, papers, or practical examples that explain how mask generation for ECG data is typically done.\n\nThanks in advance!",
    "3312766": "We modified ecg-image-kit to generate segmentation masks.\n\nhttps://github.com/Ahus-AIM/ecg-image-kit"
  },
  "source": "meta"
}