{
  "id": 263002,
  "title": "3D UNet Classisfication",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/263002",
  "author_name": "",
  "post_date": "2021-08-08T06:08:58.495556500Z",
  "votes": 14,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I was searching for information on 3D UNets and found very little online so I thought to make a discussion topic on it. I am surprised we have not talked it about more, since it's often used in medical imaging and has won various competitions. So the question is what resources are recommended for 3D UNets, what software is used for masking and how would one go about training this dataset.</p>",
  "messages": [
    {
      "id": "1458906",
      "postDate": "08/08/2021 06:08:58",
      "content": "<p>I was searching for information on 3D UNets and found very little online so I thought to make a discussion topic on it. I am surprised we have not talked it about more, since it's often used in medical imaging and has won various competitions. So the question is what resources are recommended for 3D UNets, what software is used for masking and how would one go about training this dataset.</p>",
      "rawMarkdown": "I was searching for information on 3D UNets and found very little online so I thought to make a discussion topic on it. I am surprised we have not talked it about more, since it's often used in medical imaging and has won various competitions. So the question is what resources are recommended for 3D UNets, what software is used for masking and how would one go about training this dataset.",
      "votes": null
    },
    {
      "id": "1460349",
      "postDate": "08/08/2021 20:09:30",
      "content": "<p>Check <a href=\"https://github.com/MIC-DKFZ/nnUNet\" target=\"_blank\">https://github.com/MIC-DKFZ/nnUNet</a>, is the same net used for creating the segmentations. </p>",
      "rawMarkdown": "Check https://github.com/MIC-DKFZ/nnUNet, is the same net used for creating the segmentations.",
      "votes": null
    },
    {
      "id": "1460918",
      "postDate": "08/09/2021 06:12:02",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/ranafago\" target=\"_blank\">@ranafago</a> 🙂</p>",
      "rawMarkdown": "Thanks @ranafago 🙂",
      "votes": null
    },
    {
      "id": "1461258",
      "postDate": "08/09/2021 09:57:47",
      "content": "<p>I think it's only for linux os users, windows is not fit ((</p>",
      "rawMarkdown": "I think it's only for linux os users, windows is not fit ((",
      "votes": null
    },
    {
      "id": "1461342",
      "postDate": "08/09/2021 10:57:40",
      "content": "<p>As far as i know, U-net's architecture is specifically designed to perform segmentation tasks. People probably do not talk much about it because the task of this competition is binary classification, although there may be use cases of U-net for classification that I am not aware of. </p>",
      "rawMarkdown": "As far as i know, U-net's architecture is specifically designed to perform segmentation tasks. People probably do not talk much about it because the task of this competition is binary classification, although there may be use cases of U-net for classification that I am not aware of.",
      "votes": null
    },
    {
      "id": "1469659",
      "postDate": "08/13/2021 04:18:55",
      "content": "<p>Yes, it's true that U-nets are used for segmentation tasks, however, by combining the voxelwise classes as it's been done in this paper, <a href=\"http://www.ajnr.org/content/42/5/845\" target=\"_blank\">http://www.ajnr.org/content/42/5/845</a> you could make a better classification model than most 3D models.</p>",
      "rawMarkdown": "Yes, it's true that U-nets are used for segmentation tasks, however, by combining the voxelwise classes as it's been done in this paper, http://www.ajnr.org/content/42/5/845 you could make a better classification model than most 3D models.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1460349,
      "author_name": "ranafago",
      "author_url": "",
      "post_date": "08/08/2021 20:09:30",
      "content": "<p>Check <a href=\"https://github.com/MIC-DKFZ/nnUNet\" target=\"_blank\">https://github.com/MIC-DKFZ/nnUNet</a>, is the same net used for creating the segmentations. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1460918,
          "author_name": "aristotle609",
          "author_url": "",
          "post_date": "08/09/2021 06:12:02",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/ranafago\" target=\"_blank\">@ranafago</a> 🙂</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1461258,
          "author_name": "kevinleekrus",
          "author_url": "",
          "post_date": "08/09/2021 09:57:47",
          "content": "<p>I think it's only for linux os users, windows is not fit ((</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1461342,
      "author_name": "jinuahn",
      "author_url": "",
      "post_date": "08/09/2021 10:57:40",
      "content": "<p>As far as i know, U-net's architecture is specifically designed to perform segmentation tasks. People probably do not talk much about it because the task of this competition is binary classification, although there may be use cases of U-net for classification that I am not aware of. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1469659,
          "author_name": "aristotle609",
          "author_url": "",
          "post_date": "08/13/2021 04:18:55",
          "content": "<p>Yes, it's true that U-nets are used for segmentation tasks, however, by combining the voxelwise classes as it's been done in this paper, <a href=\"http://www.ajnr.org/content/42/5/845\" target=\"_blank\">http://www.ajnr.org/content/42/5/845</a> you could make a better classification model than most 3D models.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1458906": "I was searching for information on 3D UNets and found very little online so I thought to make a discussion topic on it. I am surprised we have not talked it about more, since it's often used in medical imaging and has won various competitions. So the question is what resources are recommended for 3D UNets, what software is used for masking and how would one go about training this dataset.",
    "1460349": "Check https://github.com/MIC-DKFZ/nnUNet, is the same net used for creating the segmentations.",
    "1460918": "Thanks @ranafago 🙂",
    "1461258": "I think it's only for linux os users, windows is not fit ((",
    "1461342": "As far as i know, U-net's architecture is specifically designed to perform segmentation tasks. People probably do not talk much about it because the task of this competition is binary classification, although there may be use cases of U-net for classification that I am not aware of.",
    "1469659": "Yes, it's true that U-nets are used for segmentation tasks, however, by combining the voxelwise classes as it's been done in this paper, http://www.ajnr.org/content/42/5/845 you could make a better classification model than most 3D models."
  },
  "source": "meta"
}