{
  "id": 437344,
  "title": "probably wrong image level label?",
  "url": "/competitions/rsna-2023-abdominal-trauma-detection/discussion/437344",
  "author_name": "hengck23",
  "post_date": "2023-09-06T12:47:29.811000",
  "votes": 11,
  "comment_count": 1,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Ff02214e482e7975a19e1d7fa091b9b0a%2FSelection_999(3094).png?generation=1694004391970292&amp;alt=media\" alt=\"\"></p>\n<p>above is bowel injury image level label for series id  21228. <br>\nIt is different from others (see below)</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fe2206368f93072054413bc3858bc0f0c%2FSelection_999(3095).png?generation=1694004446292244&amp;alt=media\" alt=\"\"></p>\n<p>also:<br>\npatient id = 45303,  series id  = 46118 has no image label (but it has bowel injury=1)</p>\n<p>there is less than 100 positive samples for  bowel injury. This is very small sample size. hence any error in label may have large effects</p>\n<pre><code>''' \n\nno image bowel label ????\npatient id, series id\n???? 8684 50377\n???? 45303 46118\n???? 10065 46839\n???? 20951 6575\n???? 54371 55647\n???? 54525 50667\n???? 58465 62600\n???? 8684 50377\n\n\n'''\n</code></pre>",
  "messages": [
    {
      "id": 2426155,
      "postDate": "2023-09-06T12:47:29.810Z",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Ff02214e482e7975a19e1d7fa091b9b0a%2FSelection_999(3094).png?generation=1694004391970292&amp;alt=media\" alt=\"\"></p>\n<p>above is bowel injury image level label for series id  21228. <br>\nIt is different from others (see below)</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fe2206368f93072054413bc3858bc0f0c%2FSelection_999(3095).png?generation=1694004446292244&amp;alt=media\" alt=\"\"></p>\n<p>also:<br>\npatient id = 45303,  series id  = 46118 has no image label (but it has bowel injury=1)</p>\n<p>there is less than 100 positive samples for  bowel injury. This is very small sample size. hence any error in label may have large effects</p>\n<pre><code>''' \n\nno image bowel label ????\npatient id, series id\n???? 8684 50377\n???? 45303 46118\n???? 10065 46839\n???? 20951 6575\n???? 54371 55647\n???? 54525 50667\n???? 58465 62600\n???? 8684 50377\n\n\n'''\n</code></pre>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Ff02214e482e7975a19e1d7fa091b9b0a%2FSelection_999(3094).png?generation=1694004391970292&alt=media)\n\nabove is bowel injury image level label for series id  21228. \nIt is different from others (see below)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fe2206368f93072054413bc3858bc0f0c%2FSelection_999(3095).png?generation=1694004446292244&alt=media)\n\nalso:\npatient id = 45303,  series id  = 46118 has no image label (but it has bowel injury=1)\n\nthere is less than 100 positive samples for  bowel injury. This is very small sample size. hence any error in label may have large effects\n\n```\n''' \n\nno image bowel label ????\npatient id, series id\n???? 8684 50377\n???? 45303 46118\n???? 10065 46839\n???? 20951 6575\n???? 54371 55647\n???? 54525 50667\n???? 58465 62600\n???? 8684 50377\n\n\n'''\n```",
      "votes": 10
    },
    {
      "id": 2427373,
      "postDate": "2023-09-07T07:28:50.530Z",
      "content": "<p>train.csv has labels at the patient id level so bowel_injury == 1 for 45303</p>\n<p>image_level_labels.csv has Bowel and Active_Extravasation injury_name at the patient id, series id and instance number level.<br>\nfor patient id 45303, unique series id is only one 44615 and unique injury name is Bowel</p>\n<p>train_series_meta.csv has aortic_hu  and (for train only not test meta)     incomplete_organ 0 or 1 at the patient id, series id level<br>\nfor patient id  45303 series id 44615  aortic_hu is 229.0  for series id 46118     aortic_hu is 170.0</p>\n<p>from this host discussion -<br>\n<a href=\"https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226\" target=\"_blank\">https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226</a></p>\n<p>\"Multiphasic CT: This is a more specialized protocol that involves capturing images at different times after injecting the contrast.</p>\n<p>One way to determine phases in a multiphasic CT is to measure the density of blood in the abdominal aorta in Hounsfield units. This is provided in the [train/test]_series_meta.csv file in the aortic_hu column. For a multiphasic CT scan, the higher value typically indicates the late arterial phase. </p>\n<p>Late arterial phase: Images are taken approximately 35 seconds after injecting IV contrast. The phase helps to evaluate the blood vessels and find areas of active bleeding.\"</p>\n<p>So for this patient 45303 series id 44615 seems to be the Late arterial phase and based on the image level labels is the series id that shows active bleeding in Bowel and why the train patient has bowel injury label == 1.</p>\n<p>It may be worthwhile considering which series id to use for training for bowel_injury and extravasation_injury or at least consider differences in the series if there are multiples. </p>",
      "rawMarkdown": "train.csv has labels at the patient id level so bowel_injury == 1 for 45303\n\nimage_level_labels.csv has Bowel and Active_Extravasation injury_name at the patient id, series id and instance number level.\nfor patient id 45303, unique series id is only one 44615 and unique injury name is Bowel\n\ntrain_series_meta.csv has aortic_hu  and (for train only not test meta) \tincomplete_organ 0 or 1 at the patient id, series id level\nfor patient id  45303 series id 44615  aortic_hu is 229.0  for series id 46118 \taortic_hu is 170.0\n\nfrom this host discussion -\nhttps://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226\n\n\"Multiphasic CT: This is a more specialized protocol that involves capturing images at different times after injecting the contrast.\n\nOne way to determine phases in a multiphasic CT is to measure the density of blood in the abdominal aorta in Hounsfield units. This is provided in the [train/test]_series_meta.csv file in the aortic_hu column. For a multiphasic CT scan, the higher value typically indicates the late arterial phase. \n\nLate arterial phase: Images are taken approximately 35 seconds after injecting IV contrast. The phase helps to evaluate the blood vessels and find areas of active bleeding.\"\n\nSo for this patient 45303 series id 44615 seems to be the Late arterial phase and based on the image level labels is the series id that shows active bleeding in Bowel and why the train patient has bowel injury label == 1.\n\nIt may be worthwhile considering which series id to use for training for bowel_injury and extravasation_injury or at least consider differences in the series if there are multiples. \n",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 2427373,
      "author_name": "something4kag",
      "author_url": "",
      "post_date": "2023-09-07T07:28:50.530000",
      "content": "<p>train.csv has labels at the patient id level so bowel_injury == 1 for 45303</p>\n<p>image_level_labels.csv has Bowel and Active_Extravasation injury_name at the patient id, series id and instance number level.<br>\nfor patient id 45303, unique series id is only one 44615 and unique injury name is Bowel</p>\n<p>train_series_meta.csv has aortic_hu  and (for train only not test meta)     incomplete_organ 0 or 1 at the patient id, series id level<br>\nfor patient id  45303 series id 44615  aortic_hu is 229.0  for series id 46118     aortic_hu is 170.0</p>\n<p>from this host discussion -<br>\n<a href=\"https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226\" target=\"_blank\">https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226</a></p>\n<p>\"Multiphasic CT: This is a more specialized protocol that involves capturing images at different times after injecting the contrast.</p>\n<p>One way to determine phases in a multiphasic CT is to measure the density of blood in the abdominal aorta in Hounsfield units. This is provided in the [train/test]_series_meta.csv file in the aortic_hu column. For a multiphasic CT scan, the higher value typically indicates the late arterial phase. </p>\n<p>Late arterial phase: Images are taken approximately 35 seconds after injecting IV contrast. The phase helps to evaluate the blood vessels and find areas of active bleeding.\"</p>\n<p>So for this patient 45303 series id 44615 seems to be the Late arterial phase and based on the image level labels is the series id that shows active bleeding in Bowel and why the train patient has bowel injury label == 1.</p>\n<p>It may be worthwhile considering which series id to use for training for bowel_injury and extravasation_injury or at least consider differences in the series if there are multiples. </p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2426155": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Ff02214e482e7975a19e1d7fa091b9b0a%2FSelection_999(3094).png?generation=1694004391970292&alt=media)\n\nabove is bowel injury image level label for series id  21228. \nIt is different from others (see below)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fe2206368f93072054413bc3858bc0f0c%2FSelection_999(3095).png?generation=1694004446292244&alt=media)\n\nalso:\npatient id = 45303,  series id  = 46118 has no image label (but it has bowel injury=1)\n\nthere is less than 100 positive samples for  bowel injury. This is very small sample size. hence any error in label may have large effects\n\n```\n''' \n\nno image bowel label ????\npatient id, series id\n???? 8684 50377\n???? 45303 46118\n???? 10065 46839\n???? 20951 6575\n???? 54371 55647\n???? 54525 50667\n???? 58465 62600\n???? 8684 50377\n\n\n'''\n```",
    "2427373": "train.csv has labels at the patient id level so bowel_injury == 1 for 45303\n\nimage_level_labels.csv has Bowel and Active_Extravasation injury_name at the patient id, series id and instance number level.\nfor patient id 45303, unique series id is only one 44615 and unique injury name is Bowel\n\ntrain_series_meta.csv has aortic_hu  and (for train only not test meta) \tincomplete_organ 0 or 1 at the patient id, series id level\nfor patient id  45303 series id 44615  aortic_hu is 229.0  for series id 46118 \taortic_hu is 170.0\n\nfrom this host discussion -\nhttps://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection/discussion/427226\n\n\"Multiphasic CT: This is a more specialized protocol that involves capturing images at different times after injecting the contrast.\n\nOne way to determine phases in a multiphasic CT is to measure the density of blood in the abdominal aorta in Hounsfield units. This is provided in the [train/test]_series_meta.csv file in the aortic_hu column. For a multiphasic CT scan, the higher value typically indicates the late arterial phase. \n\nLate arterial phase: Images are taken approximately 35 seconds after injecting IV contrast. The phase helps to evaluate the blood vessels and find areas of active bleeding.\"\n\nSo for this patient 45303 series id 44615 seems to be the Late arterial phase and based on the image level labels is the series id that shows active bleeding in Bowel and why the train patient has bowel injury label == 1.\n\nIt may be worthwhile considering which series id to use for training for bowel_injury and extravasation_injury or at least consider differences in the series if there are multiples. \n"
  }
}