{
  "id": 591557,
  "title": "How to utilize segmentations/ or nii files?",
  "url": "/competitions/rsna-intracranial-aneurysm-detection/discussion/591557",
  "author_name": "Ryo Nakamura",
  "post_date": "2025-07-29T04:44:15.409000",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>According to the data description,</p>\n<blockquote>\n  <p>segmentations/ Contains NifTI vessel segmentations for a subset of the DICOM series in series/.</p>\n</blockquote>\n<p>So some DICOM series don't have segmentations, meaning we can't necessarily rely on segmentations right? Should we use those segmentation info as additional like using GRU or something like that?</p>",
  "messages": [
    {
      "id": 3255839,
      "postDate": "2025-07-29T14:14:12.110Z",
      "content": "<p>Thanks for bringing this up. The segmentations are provided for only a small subset of DICOM series. The idea is that some approaches may want to consider training a segmentation algorithm to assist with aneurysm parent vessel identification. So, for example, you could use the provided segmentation data to train a segmentation model, and then run inference with this model as part of your algorithm for localizing aneurysms. Hopefully challenge participants will find interesting and unique ways to use these segmentations that we have not yet considered.</p>",
      "rawMarkdown": "Thanks for bringing this up. The segmentations are provided for only a small subset of DICOM series. The idea is that some approaches may want to consider training a segmentation algorithm to assist with aneurysm parent vessel identification. So, for example, you could use the provided segmentation data to train a segmentation model, and then run inference with this model as part of your algorithm for localizing aneurysms. Hopefully challenge participants will find interesting and unique ways to use these segmentations that we have not yet considered.",
      "votes": 1,
      "replies": [
        {
          "id": 3258570,
          "postDate": "2025-07-30T22:14:22.383Z",
          "content": "<p>Thank you so much for your reply, I'll try to find ways to use it effectively. </p>",
          "rawMarkdown": "Thank you so much for your reply, I'll try to find ways to use it effectively. "
        }
      ]
    },
    {
      "id": 3255615,
      "postDate": "2025-07-29T04:44:15.410Z",
      "content": "<p>According to the data description,</p>\n<blockquote>\n  <p>segmentations/ Contains NifTI vessel segmentations for a subset of the DICOM series in series/.</p>\n</blockquote>\n<p>So some DICOM series don't have segmentations, meaning we can't necessarily rely on segmentations right? Should we use those segmentation info as additional like using GRU or something like that?</p>",
      "rawMarkdown": "According to the data description,\n>segmentations/ Contains NifTI vessel segmentations for a subset of the DICOM series in series/.\n\nSo some DICOM series don't have segmentations, meaning we can't necessarily rely on segmentations right? Should we use those segmentation info as additional like using GRU or something like that?",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 3255839,
      "author_name": "Evan Calabrese",
      "author_url": "",
      "post_date": "2025-07-29T14:14:12.110000",
      "content": "<p>Thanks for bringing this up. The segmentations are provided for only a small subset of DICOM series. The idea is that some approaches may want to consider training a segmentation algorithm to assist with aneurysm parent vessel identification. So, for example, you could use the provided segmentation data to train a segmentation model, and then run inference with this model as part of your algorithm for localizing aneurysms. Hopefully challenge participants will find interesting and unique ways to use these segmentations that we have not yet considered.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 3258570,
          "author_name": "Ryo Nakamura",
          "author_url": "",
          "post_date": "2025-07-30T22:14:22.383000",
          "content": "<p>Thank you so much for your reply, I'll try to find ways to use it effectively. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3255839": "Thanks for bringing this up. The segmentations are provided for only a small subset of DICOM series. The idea is that some approaches may want to consider training a segmentation algorithm to assist with aneurysm parent vessel identification. So, for example, you could use the provided segmentation data to train a segmentation model, and then run inference with this model as part of your algorithm for localizing aneurysms. Hopefully challenge participants will find interesting and unique ways to use these segmentations that we have not yet considered.",
    "3255615": "According to the data description,\n>segmentations/ Contains NifTI vessel segmentations for a subset of the DICOM series in series/.\n\nSo some DICOM series don't have segmentations, meaning we can't necessarily rely on segmentations right? Should we use those segmentation info as additional like using GRU or something like that?"
  }
}