{
  "id": 267257,
  "title": "DICOM to NifiTi with TorchIO",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/267257",
  "author_name": "",
  "post_date": "2021-08-22T13:17:09.587082500Z",
  "votes": 5,
  "comment_count": 2,
  "views": 0,
  "content": "<p>TorchIO makes converting and normalizing 3D medical images pretty easy.  </p>\n<p><a href=\"https://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio\" target=\"_blank\">https://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio</a></p>",
  "messages": [
    {
      "id": "1485843",
      "postDate": "08/22/2021 13:17:09",
      "content": "<p>TorchIO makes converting and normalizing 3D medical images pretty easy.  </p>\n<p><a href=\"https://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio\" target=\"_blank\">https://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio</a></p>",
      "rawMarkdown": "TorchIO makes converting and normalizing 3D medical images pretty easy.  \n\nhttps://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio",
      "votes": null
    },
    {
      "id": "1503135",
      "postDate": "09/05/2021 03:27:17",
      "content": "<p>Nice! How did you choose the args for <code>Crop</code> and <code>CropOrPad</code>?</p>",
      "rawMarkdown": "Nice! How did you choose the args for `Crop` and `CropOrPad`?",
      "votes": null
    },
    {
      "id": "1506117",
      "postDate": "09/07/2021 21:42:35",
      "content": "<p><a href=\"https://www.kaggle.com/fepegar\" target=\"_blank\">@fepegar</a> Currently I am using 128,128,64  for <code>CropOrPad</code> I chose these values after looking at the average shapes of the NifiTi files I created using dcmstack. Also, I am working under the assumption I needed a standard shape to train a CNN and to use <code>tio.data.UniformSampler(128,128,64)</code>.  </p>\n<p>I will update the notebook to reflect this, but I would recommend anyone reading this to <strong>not</strong> use this notebook, and instead use the notebook you just created \"<em>Preprocessing MRI with TorchIO</em>\".  After all you are the author of TorchIO!</p>",
      "rawMarkdown": "fepegar Currently I am using 128,128,64  for `CropOrPad` I chose these values after looking at the average shapes of the NifiTi files I created using dcmstack. Also, I am working under the assumption I needed a standard shape to train a CNN and to use `tio.data.UniformSampler(128,128,64)`.  \n\nI will update the notebook to reflect this, but I would recommend anyone reading this to **not** use this notebook, and instead use the notebook you just created \"*Preprocessing MRI with TorchIO*\".  After all you are the author of TorchIO!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1503135,
      "author_name": "fepegar",
      "author_url": "",
      "post_date": "09/05/2021 03:27:17",
      "content": "<p>Nice! How did you choose the args for <code>Crop</code> and <code>CropOrPad</code>?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1506117,
          "author_name": "ohbewise",
          "author_url": "",
          "post_date": "09/07/2021 21:42:35",
          "content": "<p><a href=\"https://www.kaggle.com/fepegar\" target=\"_blank\">@fepegar</a> Currently I am using 128,128,64  for <code>CropOrPad</code> I chose these values after looking at the average shapes of the NifiTi files I created using dcmstack. Also, I am working under the assumption I needed a standard shape to train a CNN and to use <code>tio.data.UniformSampler(128,128,64)</code>.  </p>\n<p>I will update the notebook to reflect this, but I would recommend anyone reading this to <strong>not</strong> use this notebook, and instead use the notebook you just created \"<em>Preprocessing MRI with TorchIO</em>\".  After all you are the author of TorchIO!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1485843": "TorchIO makes converting and normalizing 3D medical images pretty easy.  \n\nhttps://www.kaggle.com/ohbewise/dicom-to-nifiti-with-torchio",
    "1503135": "Nice! How did you choose the args for `Crop` and `CropOrPad`?",
    "1506117": "fepegar Currently I am using 128,128,64  for `CropOrPad` I chose these values after looking at the average shapes of the NifiTi files I created using dcmstack. Also, I am working under the assumption I needed a standard shape to train a CNN and to use `tio.data.UniformSampler(128,128,64)`.  \n\nI will update the notebook to reflect this, but I would recommend anyone reading this to **not** use this notebook, and instead use the notebook you just created \"*Preprocessing MRI with TorchIO*\".  After all you are the author of TorchIO!"
  },
  "source": "meta"
}