{
  "id": 252884,
  "title": "Valid Images",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/252884",
  "author_name": "",
  "post_date": "2021-07-14T04:22:20.827425600Z",
  "votes": 5,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Not sure if I did this correctly, but I have never dealt with dicoms before.  Noticed when trying to display some images so were just arrays of zeroes so I made a dataset that should provide only the valid images in train and test set.  Let me know what I did wrong.</p>\n<p><a href=\"https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images\" target=\"_blank\">https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images</a></p>",
  "messages": [
    {
      "id": "1387281",
      "postDate": "07/14/2021 04:22:20",
      "content": "<p>Not sure if I did this correctly, but I have never dealt with dicoms before.  Noticed when trying to display some images so were just arrays of zeroes so I made a dataset that should provide only the valid images in train and test set.  Let me know what I did wrong.</p>\n<p><a href=\"https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images\" target=\"_blank\">https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images</a></p>",
      "rawMarkdown": "Not sure if I did this correctly, but I have never dealt with dicoms before.  Noticed when trying to display some images so were just arrays of zeroes so I made a dataset that should provide only the valid images in train and test set.  Let me know what I did wrong.\n\n[https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images](https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images)",
      "votes": null
    },
    {
      "id": "1387324",
      "postDate": "07/14/2021 05:00:40",
      "content": "<p>A lot of the images are empty space in MR. This is normal since they're reconstructed in multiple planes and the anatomy doesn't always occupy the entire space. A simple solution is to ignore any image with a pixel mean() of 0.</p>",
      "rawMarkdown": "A lot of the images are empty space in MR. This is normal since they're reconstructed in multiple planes and the anatomy doesn't always occupy the entire space. A simple solution is to ignore any image with a pixel mean() of 0.",
      "votes": null
    },
    {
      "id": "1387713",
      "postDate": "07/14/2021 11:38:07",
      "content": "<p>Yeah, I figured they were just blank slices of the brain scans.  I felt like filtering the images ahead of time would allow me to make a properly sized np array when it came to training time, but the time savings might be negligible.</p>",
      "rawMarkdown": "Yeah, I figured they were just blank slices of the brain scans.  I felt like filtering the images ahead of time would allow me to make a properly sized np array when it came to training time, but the time savings might be negligible.",
      "votes": null
    },
    {
      "id": "1387738",
      "postDate": "07/14/2021 12:01:08",
      "content": "<p>Thanks for explaining.  I was wondering because some images have np.max(img) == 0. But you answered my question.</p>",
      "rawMarkdown": "Thanks for explaining.  I was wondering because some images have np.max(img) == 0. But you answered my question.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1387324,
      "author_name": "davidbroberts",
      "author_url": "",
      "post_date": "07/14/2021 05:00:40",
      "content": "<p>A lot of the images are empty space in MR. This is normal since they're reconstructed in multiple planes and the anatomy doesn't always occupy the entire space. A simple solution is to ignore any image with a pixel mean() of 0.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1387713,
          "author_name": "craigwickizer",
          "author_url": "",
          "post_date": "07/14/2021 11:38:07",
          "content": "<p>Yeah, I figured they were just blank slices of the brain scans.  I felt like filtering the images ahead of time would allow me to make a properly sized np array when it came to training time, but the time savings might be negligible.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1387738,
          "author_name": "lucamtb",
          "author_url": "",
          "post_date": "07/14/2021 12:01:08",
          "content": "<p>Thanks for explaining.  I was wondering because some images have np.max(img) == 0. But you answered my question.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1387281": "Not sure if I did this correctly, but I have never dealt with dicoms before.  Noticed when trying to display some images so were just arrays of zeroes so I made a dataset that should provide only the valid images in train and test set.  Let me know what I did wrong.\n\n[https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images](https://www.kaggle.com/craigwickizer/rsnamiccai-brain-tumor-radiogenomic-valid-images)",
    "1387324": "A lot of the images are empty space in MR. This is normal since they're reconstructed in multiple planes and the anatomy doesn't always occupy the entire space. A simple solution is to ignore any image with a pixel mean() of 0.",
    "1387713": "Yeah, I figured they were just blank slices of the brain scans.  I felt like filtering the images ahead of time would allow me to make a properly sized np array when it came to training time, but the time savings might be negligible.",
    "1387738": "Thanks for explaining.  I was wondering because some images have np.max(img) == 0. But you answered my question."
  },
  "source": "meta"
}