{
  "id": 266570,
  "title": "PNG DATASET?",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/266570",
  "author_name": "",
  "post_date": "2021-08-19T15:18:37.187992200Z",
  "votes": 5,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Has anyone converted this whole dicom dataset into a png dataset? If yes, can anyone give a link to such a dataset ?</p>",
  "messages": [
    {
      "id": "1481548",
      "postDate": "08/19/2021 15:18:37",
      "content": "<p>Has anyone converted this whole dicom dataset into a png dataset? If yes, can anyone give a link to such a dataset ?</p>",
      "rawMarkdown": "Has anyone converted this whole dicom dataset into a png dataset? If yes, can anyone give a link to such a dataset ?",
      "votes": null
    },
    {
      "id": "1481560",
      "postDate": "08/19/2021 15:23:48",
      "content": "<p>Here is a PNG dataset and resizing into 256x256 2D png images and 36 slices per modality per patient: <a href=\"http://www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a\" target=\"_blank\">www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a</a></p>\n<p>You can find the Notebook on how to do this here: <a href=\"https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36\" target=\"_blank\">https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36</a></p>\n<p>hope this helps.</p>\n<p>PS files are also converted into AXIAL plane for all modalities and patients.</p>",
      "rawMarkdown": "Here is a PNG dataset and resizing into 256x256 2D png images and 36 slices per modality per patient: www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a\n\nYou can find the Notebook on how to do this here: https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36\n\nhope this helps.\n\nPS files are also converted into AXIAL plane for all modalities and patients.",
      "votes": null
    },
    {
      "id": "1481736",
      "postDate": "08/19/2021 16:41:02",
      "content": "<p>How much time did it take to convert these images?? It is killing me !!!</p>",
      "rawMarkdown": "How much time did it take to convert these images?? It is killing me !!!",
      "votes": null
    },
    {
      "id": "1481753",
      "postDate": "08/19/2021 16:50:09",
      "content": "<p>It is slow indeed. It takes about 2 hours to do the entire conversion to voxel space (in this case AXIAL), resizing to 256x256x36 and then saving train and test keeping the same directory structure as the original dataset. The bottleneck is the IO for sure, saving each slice individually as png files. I don't really know how to speed this up…sorry.</p>\n<p>You can use the dataset I created though directly, just keep in mind that an intermediate conversion to the same plane and size is being performed.</p>",
      "rawMarkdown": "It is slow indeed. It takes about 2 hours to do the entire conversion to voxel space (in this case AXIAL), resizing to 256x256x36 and then saving train and test keeping the same directory structure as the original dataset. The bottleneck is the IO for sure, saving each slice individually as png files. I don't really know how to speed this up...sorry.\n\nYou can use the dataset I created though directly, just keep in mind that an intermediate conversion to the same plane and size is being performed.",
      "votes": null
    },
    {
      "id": "1481876",
      "postDate": "08/19/2021 17:48:43",
      "content": "<p>Is that (the slowness of the process) the reason why you have 36 images per mri type or is their some other reason ?</p>",
      "rawMarkdown": "Is that (the slowness of the process) the reason why you have 36 images per mri type or is their some other reason ?",
      "votes": null
    },
    {
      "id": "1482546",
      "postDate": "08/20/2021 06:25:54",
      "content": "<p>No, that was a choice a made semi-randomly really. I thought 36 slices is a small but indicative amount of slices. You can choose whatever you see most appropriate.</p>",
      "rawMarkdown": "No, that was a choice a made semi-randomly really. I thought 36 slices is a small but indicative amount of slices. You can choose whatever you see most appropriate.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1481560,
      "author_name": "smoschou55",
      "author_url": "",
      "post_date": "08/19/2021 15:23:48",
      "content": "<p>Here is a PNG dataset and resizing into 256x256 2D png images and 36 slices per modality per patient: <a href=\"http://www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a\" target=\"_blank\">www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a</a></p>\n<p>You can find the Notebook on how to do this here: <a href=\"https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36\" target=\"_blank\">https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36</a></p>\n<p>hope this helps.</p>\n<p>PS files are also converted into AXIAL plane for all modalities and patients.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1481736,
          "author_name": "abhranta",
          "author_url": "",
          "post_date": "08/19/2021 16:41:02",
          "content": "<p>How much time did it take to convert these images?? It is killing me !!!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1481753,
          "author_name": "smoschou55",
          "author_url": "",
          "post_date": "08/19/2021 16:50:09",
          "content": "<p>It is slow indeed. It takes about 2 hours to do the entire conversion to voxel space (in this case AXIAL), resizing to 256x256x36 and then saving train and test keeping the same directory structure as the original dataset. The bottleneck is the IO for sure, saving each slice individually as png files. I don't really know how to speed this up…sorry.</p>\n<p>You can use the dataset I created though directly, just keep in mind that an intermediate conversion to the same plane and size is being performed.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1481876,
          "author_name": "abhranta",
          "author_url": "",
          "post_date": "08/19/2021 17:48:43",
          "content": "<p>Is that (the slowness of the process) the reason why you have 36 images per mri type or is their some other reason ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1482546,
          "author_name": "smoschou55",
          "author_url": "",
          "post_date": "08/20/2021 06:25:54",
          "content": "<p>No, that was a choice a made semi-randomly really. I thought 36 slices is a small but indicative amount of slices. You can choose whatever you see most appropriate.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1481548": "Has anyone converted this whole dicom dataset into a png dataset? If yes, can anyone give a link to such a dataset ?",
    "1481560": "Here is a PNG dataset and resizing into 256x256 2D png images and 36 slices per modality per patient: www.kaggle.com/dataset/d12293c3d1f77b9ff3448b8c7aa8b84312a0b8f15f1c4de9a6140ac7680c354a\n\nYou can find the Notebook on how to do this here: https://www.kaggle.com/smoschou55/dicom-to-2d-resized-axial-pngs-256x256-x36\n\nhope this helps.\n\nPS files are also converted into AXIAL plane for all modalities and patients.",
    "1481736": "How much time did it take to convert these images?? It is killing me !!!",
    "1481753": "It is slow indeed. It takes about 2 hours to do the entire conversion to voxel space (in this case AXIAL), resizing to 256x256x36 and then saving train and test keeping the same directory structure as the original dataset. The bottleneck is the IO for sure, saving each slice individually as png files. I don't really know how to speed this up...sorry.\n\nYou can use the dataset I created though directly, just keep in mind that an intermediate conversion to the same plane and size is being performed.",
    "1481876": "Is that (the slowness of the process) the reason why you have 36 images per mri type or is their some other reason ?",
    "1482546": "No, that was a choice a made semi-randomly really. I thought 36 slices is a small but indicative amount of slices. You can choose whatever you see most appropriate."
  },
  "source": "meta"
}