{
  "id": 18402,
  "title": "Availability of SliceLocation metadata",
  "url": "/competitions/second-annual-data-science-bowl/discussion/18402",
  "author_name": "",
  "post_date": "2016-01-15T19:49:51.203Z",
  "votes": 2,
  "comment_count": 4,
  "views": 1091,
  "content": "<p>From what I saw until now it seems that the DICOM metadata field SliceLocation is present in all training and validation data (I checked in the first frame in all sax_XX folders for all studies). Can we safely assume that it will be also present in test data? This would simplify the code.</p>\n\n<p>I am asking as the Dataset class in the tutorials does not assume that but considers various fallback cases...</p>",
  "messages": [
    {
      "id": "104719",
      "postDate": "01/15/2016 19:49:51",
      "content": "<p>From what I saw until now it seems that the DICOM metadata field SliceLocation is present in all training and validation data (I checked in the first frame in all sax_XX folders for all studies). Can we safely assume that it will be also present in test data? This would simplify the code.</p>\n\n<p>I am asking as the Dataset class in the tutorials does not assume that but considers various fallback cases...</p>",
      "rawMarkdown": "From what I saw until now it seems that the DICOM metadata field SliceLocation is present in all training and validation data (I checked in the first frame in all sax_XX folders for all studies). Can we safely assume that it will be also present in test data? This would simplify the code.\r\n\r\nI am asking as the Dataset class in the tutorials does not assume that but considers various fallback cases...",
      "votes": null
    },
    {
      "id": "104720",
      "postDate": "01/15/2016 20:39:14",
      "content": "<p>Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.</p>",
      "rawMarkdown": "Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.",
      "votes": null
    },
    {
      "id": "104863",
      "postDate": "01/17/2016 14:14:02",
      "content": "<p>[quote=William Cukierski;104720]</p>\n\n<p>Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! That is very reassuring and pleasant in comparision with some other Kaggle competition which is ending now... :-)</p>\n\n<p>However since there is a possibility that there would be a small number of studies without the SliceLocation metadata which would get randomly put into the other testing half perhaps the organizers who prepared the dataset could make a statement whether it is always there?   </p>",
      "rawMarkdown": "[quote=William Cukierski;104720]\r\n\r\nTrain, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.\r\n\r\n[/quote]\r\n\r\nThanks! That is very reassuring and pleasant in comparision with some other Kaggle competition which is ending now... :-)\r\n\r\nHowever since there is a possibility that there would be a small number of studies without the SliceLocation metadata which would get randomly put into the other testing half perhaps the organizers who prepared the dataset could make a statement whether it is always there?",
      "votes": null
    },
    {
      "id": "105193",
      "postDate": "01/20/2016 20:47:27",
      "content": "<p>I just checked the test set. SliceLocation is present in all images.</p>",
      "rawMarkdown": "I just checked the test set. SliceLocation is present in all images.",
      "votes": null
    },
    {
      "id": "105194",
      "postDate": "01/20/2016 20:52:43",
      "content": "<p>Thanks a lot!</p>",
      "rawMarkdown": "Thanks a lot!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 104720,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "01/15/2016 20:39:14",
      "content": "<p>Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 104863,
      "author_name": "romualdj",
      "author_url": "",
      "post_date": "01/17/2016 14:14:02",
      "content": "<p>[quote=William Cukierski;104720]</p>\n\n<p>Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.</p>\n\n<p>[/quote]</p>\n\n<p>Thanks! That is very reassuring and pleasant in comparision with some other Kaggle competition which is ending now... :-)</p>\n\n<p>However since there is a possibility that there would be a small number of studies without the SliceLocation metadata which would get randomly put into the other testing half perhaps the organizers who prepared the dataset could make a statement whether it is always there?   </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 105193,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "01/20/2016 20:47:27",
      "content": "<p>I just checked the test set. SliceLocation is present in all images.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 105194,
      "author_name": "romualdj",
      "author_url": "",
      "post_date": "01/20/2016 20:52:43",
      "content": "<p>Thanks a lot!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "104719": "From what I saw until now it seems that the DICOM metadata field SliceLocation is present in all training and validation data (I checked in the first frame in all sax_XX folders for all studies). Can we safely assume that it will be also present in test data? This would simplify the code.\r\n\r\nI am asking as the Dataset class in the tutorials does not assume that but considers various fallback cases...",
    "104720": "Train, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.",
    "104863": "[quote=William Cukierski;104720]\r\n\r\nTrain, validation, and test were all picked uniform randomly, so it's likely that the test images will have the same properties as train/validation.\r\n\r\n[/quote]\r\n\r\nThanks! That is very reassuring and pleasant in comparision with some other Kaggle competition which is ending now... :-)\r\n\r\nHowever since there is a possibility that there would be a small number of studies without the SliceLocation metadata which would get randomly put into the other testing half perhaps the organizers who prepared the dataset could make a statement whether it is always there?",
    "105193": "I just checked the test set. SliceLocation is present in all images.",
    "105194": "Thanks a lot!"
  },
  "source": "meta"
}