{
  "id": 668650,
  "title": "How should we digitize the extracted waveforms?",
  "url": "/competitions/physionet-ecg-image-digitization/discussion/668650",
  "author_name": "nakao syoya",
  "post_date": "2026-01-18T06:09:14.305000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I would like to ask for advice from the community. I was able to extract ECG waveforms using a U-Net model. As shown in the attached image, I can now extract four long waveforms from a single ECG image.</p>\n<p>My question is: what methods are effective for digitizing these extracted waveforms into time-series signals?\nAdditionally, when waveforms overlap or partially merge with each other, what kinds of post-processing techniques are recommended?</p>\n<p>I would greatly appreciate any suggestions or insights. Thank you very much.<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22506629%2F235b2a2d7bf50e6fc8478a59d60a743d%2F2026-01-18%20150837.png?generation=1768716543955335&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": 3393032,
      "postDate": "2026-01-18T06:09:14.307Z",
      "content": "<p>I would like to ask for advice from the community. I was able to extract ECG waveforms using a U-Net model. As shown in the attached image, I can now extract four long waveforms from a single ECG image.</p>\n<p>My question is: what methods are effective for digitizing these extracted waveforms into time-series signals?\nAdditionally, when waveforms overlap or partially merge with each other, what kinds of post-processing techniques are recommended?</p>\n<p>I would greatly appreciate any suggestions or insights. Thank you very much.<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22506629%2F235b2a2d7bf50e6fc8478a59d60a743d%2F2026-01-18%20150837.png?generation=1768716543955335&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "I would like to ask for advice from the community. I was able to extract ECG waveforms using a U-Net model. As shown in the attached image, I can now extract four long waveforms from a single ECG image.\n\nMy question is: what methods are effective for digitizing these extracted waveforms into time-series signals?\nAdditionally, when waveforms overlap or partially merge with each other, what kinds of post-processing techniques are recommended?\n\nI would greatly appreciate any suggestions or insights. Thank you very much.![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22506629%2F235b2a2d7bf50e6fc8478a59d60a743d%2F2026-01-18%20150837.png?generation=1768716543955335&alt=media)",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3393032": "I would like to ask for advice from the community. I was able to extract ECG waveforms using a U-Net model. As shown in the attached image, I can now extract four long waveforms from a single ECG image.\n\nMy question is: what methods are effective for digitizing these extracted waveforms into time-series signals?\nAdditionally, when waveforms overlap or partially merge with each other, what kinds of post-processing techniques are recommended?\n\nI would greatly appreciate any suggestions or insights. Thank you very much.![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F22506629%2F235b2a2d7bf50e6fc8478a59d60a743d%2F2026-01-18%20150837.png?generation=1768716543955335&alt=media)"
  }
}