{
  "id": 207752,
  "title": "Are there multiple images fom same patient?",
  "url": "/competitions/vinbigdata-chest-xray-abnormalities-detection/discussion/207752",
  "author_name": "Awsaf",
  "post_date": "2020-12-31T05:43:10.415000",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>It's mentioned that <code>The dataset comprises 18,000 postero-anterior (PA) CXR scans in DICOM format, which were de-identified to protect patient privacy</code>. If we can't tell if one patient has multiple images, it may create a  leak on train-test split. So, are there multiple images from the  same patient? </p>",
  "messages": [
    {
      "id": 1133351,
      "postDate": "2020-12-31T06:14:15.647Z",
      "content": "<p>No, there is only 1 image per patient in our dataset 😀</p>",
      "rawMarkdown": "No, there is only 1 image per patient in our dataset 😀",
      "votes": 5,
      "replies": [
        {
          "id": 1133449,
          "postDate": "2020-12-31T08:19:58.447Z",
          "content": "<p>thanks <a href=\"https://www.kaggle.com/andy2709\" target=\"_blank\">@andy2709</a> </p>",
          "rawMarkdown": "thanks @andy2709 ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1133324,
      "postDate": "2020-12-31T05:43:10.417Z",
      "content": "<p>It's mentioned that <code>The dataset comprises 18,000 postero-anterior (PA) CXR scans in DICOM format, which were de-identified to protect patient privacy</code>. If we can't tell if one patient has multiple images, it may create a  leak on train-test split. So, are there multiple images from the  same patient? </p>",
      "rawMarkdown": "It's mentioned that `The dataset comprises 18,000 postero-anterior (PA) CXR scans in DICOM format, which were de-identified to protect patient privacy`. If we can't tell if one patient has multiple images, it may create a  leak on train-test split. So, are there multiple images from the  same patient? ",
      "votes": 2
    },
    {
      "id": 1134126,
      "postDate": "2020-12-31T21:37:14.657Z",
      "content": "<p>There is only one image per patient, however i noticed duplicate image ids while doing some EDA, that maybe because of multiple bounding boxes in the same image, each having different co-ordinates. </p>",
      "rawMarkdown": "There is only one image per patient, however i noticed duplicate image ids while doing some EDA, that maybe because of multiple bounding boxes in the same image, each having different co-ordinates. ",
      "replies": [
        {
          "id": 1134129,
          "postDate": "2020-12-31T21:44:23.143Z",
          "content": "<p>yes I've noticed that too</p>",
          "rawMarkdown": "yes I've noticed that too"
        },
        {
          "id": 1134678,
          "postDate": "2021-01-01T13:39:56.040Z",
          "content": "<p>That's indeed the reason, 3 readers per image, possibly with multiple classes to give a bounding box for. For more, see some of the EDA notebooks like <a href=\"https://www.kaggle.com/bjoernholzhauer/eda-dicom-reading-vinbigdata-chest-x-ray\" target=\"_blank\">this one</a>.</p>",
          "rawMarkdown": "That's indeed the reason, 3 readers per image, possibly with multiple classes to give a bounding box for. For more, see some of the EDA notebooks like [this one](https://www.kaggle.com/bjoernholzhauer/eda-dicom-reading-vinbigdata-chest-x-ray)."
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1133351,
      "author_name": "NguyenThanhNhan",
      "author_url": "",
      "post_date": "2020-12-31T06:14:15.647000",
      "content": "<p>No, there is only 1 image per patient in our dataset 😀</p>",
      "votes": 5,
      "replies": [
        {
          "id": 1133449,
          "author_name": "Awsaf",
          "author_url": "",
          "post_date": "2020-12-31T08:19:58.447000",
          "content": "<p>thanks <a href=\"https://www.kaggle.com/andy2709\" target=\"_blank\">@andy2709</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1134126,
      "author_name": "Anmol",
      "author_url": "",
      "post_date": "2020-12-31T21:37:14.657000",
      "content": "<p>There is only one image per patient, however i noticed duplicate image ids while doing some EDA, that maybe because of multiple bounding boxes in the same image, each having different co-ordinates. </p>",
      "votes": 0,
      "replies": [
        {
          "id": 1134129,
          "author_name": "Awsaf",
          "author_url": "",
          "post_date": "2020-12-31T21:44:23.143000",
          "content": "<p>yes I've noticed that too</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1134678,
          "author_name": "Björn",
          "author_url": "",
          "post_date": "2021-01-01T13:39:56.040000",
          "content": "<p>That's indeed the reason, 3 readers per image, possibly with multiple classes to give a bounding box for. For more, see some of the EDA notebooks like <a href=\"https://www.kaggle.com/bjoernholzhauer/eda-dicom-reading-vinbigdata-chest-x-ray\" target=\"_blank\">this one</a>.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1133351": "No, there is only 1 image per patient in our dataset 😀",
    "1133324": "It's mentioned that `The dataset comprises 18,000 postero-anterior (PA) CXR scans in DICOM format, which were de-identified to protect patient privacy`. If we can't tell if one patient has multiple images, it may create a  leak on train-test split. So, are there multiple images from the  same patient? ",
    "1134126": "There is only one image per patient, however i noticed duplicate image ids while doing some EDA, that maybe because of multiple bounding boxes in the same image, each having different co-ordinates. "
  }
}