{
  "id": 110068,
  "title": "DICOM information: ImageOrientation",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/110068",
  "author_name": "srs",
  "post_date": "2019-09-24T13:51:36.567000",
  "votes": 0,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Dear all,</p>\n\n<p>I have a (possibly very naive) question regarding the ImageOrientation and ImagePosition fields. </p>\n\n<p>If I understand it correctly they are information about the coordinate system of the CT-image. That is describing the position (i.e. size) of pixels / voxel, as well as the axis etc.  Described here: \n<a href=\"https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037\">https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037</a></p>\n\n<p>So I might be able to use the information to define the position of a single voxel in space, and use it to calculate an affine transformation, to rotate and translate the original voxel to its matching position in a template image.</p>\n\n<p>Question:\nIf I want to have my images with the nose facing upwards and the head in the middle, the ImageOrientation, and ImagePosition won't help me that much, correct?</p>\n\n<p>The slices will be shifted and rotated, so that the voxels are matching the other space, however, the relative head position might not be affected that much. To achieve my goal, I would need to do some more complex (coregistration,  matching) procedures.</p>",
  "messages": [
    {
      "id": 635263,
      "postDate": "2019-09-27T10:21:16.907Z",
      "content": "<p>ImageOrientation just describe the dicom image border direction。It isn't related to the face and nose direction。</p>",
      "rawMarkdown": "ImageOrientation just describe the dicom image border direction。It isn't related to the face and nose direction。",
      "votes": 1,
      "replies": [
        {
          "id": 635327,
          "postDate": "2019-09-27T11:38:59.707Z",
          "content": "<p>Thanks for the clarification! I thought so, but wasn't exactly sure. On a previous project I spent way too much time on affine transformation and was afraid to go done that rabbit hole again^^.</p>",
          "rawMarkdown": "Thanks for the clarification! I thought so, but wasn't exactly sure. On a previous project I spent way too much time on affine transformation and was afraid to go done that rabbit hole again^^."
        }
      ]
    },
    {
      "id": 633502,
      "postDate": "2019-09-25T02:58:36.477Z",
      "content": "<p>or you can just use vertical flip augmentation </p>",
      "rawMarkdown": "or you can just use vertical flip augmentation ",
      "replies": [
        {
          "id": 633836,
          "postDate": "2019-09-25T13:28:40.583Z",
          "content": "<p>Probably, I was more coming from the classical neuro-imaging approach, where you commonly co-register your images to a template, so that at least the orientation and the position of your participant's / patient's heads are roughly the same. </p>",
          "rawMarkdown": "Probably, I was more coming from the classical neuro-imaging approach, where you commonly co-register your images to a template, so that at least the orientation and the position of your participant's / patient's heads are roughly the same. ",
          "votes": 1
        }
      ]
    },
    {
      "id": 633161,
      "postDate": "2019-09-24T13:51:36.567Z",
      "content": "<p>Dear all,</p>\n\n<p>I have a (possibly very naive) question regarding the ImageOrientation and ImagePosition fields. </p>\n\n<p>If I understand it correctly they are information about the coordinate system of the CT-image. That is describing the position (i.e. size) of pixels / voxel, as well as the axis etc.  Described here: \n<a href=\"https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037\">https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037</a></p>\n\n<p>So I might be able to use the information to define the position of a single voxel in space, and use it to calculate an affine transformation, to rotate and translate the original voxel to its matching position in a template image.</p>\n\n<p>Question:\nIf I want to have my images with the nose facing upwards and the head in the middle, the ImageOrientation, and ImagePosition won't help me that much, correct?</p>\n\n<p>The slices will be shifted and rotated, so that the voxels are matching the other space, however, the relative head position might not be affected that much. To achieve my goal, I would need to do some more complex (coregistration,  matching) procedures.</p>",
      "rawMarkdown": "Dear all,\n\nI have a (possibly very naive) question regarding the ImageOrientation and ImagePosition fields. \n\nIf I understand it correctly they are information about the coordinate system of the CT-image. That is describing the position (i.e. size) of pixels / voxel, as well as the axis etc.  Described here: \n[https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037](https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037)\n\nSo I might be able to use the information to define the position of a single voxel in space, and use it to calculate an affine transformation, to rotate and translate the original voxel to its matching position in a template image.\n\nQuestion:\nIf I want to have my images with the nose facing upwards and the head in the middle, the ImageOrientation, and ImagePosition won't help me that much, correct?\n\nThe slices will be shifted and rotated, so that the voxels are matching the other space, however, the relative head position might not be affected that much. To achieve my goal, I would need to do some more complex (coregistration,  matching) procedures.\n"
    }
  ],
  "comments": [
    {
      "id": 635263,
      "author_name": "csuzbw",
      "author_url": "",
      "post_date": "2019-09-27T10:21:16.907000",
      "content": "<p>ImageOrientation just describe the dicom image border direction。It isn't related to the face and nose direction。</p>",
      "votes": 1,
      "replies": [
        {
          "id": 635327,
          "author_name": "srs",
          "author_url": "",
          "post_date": "2019-09-27T11:38:59.707000",
          "content": "<p>Thanks for the clarification! I thought so, but wasn't exactly sure. On a previous project I spent way too much time on affine transformation and was afraid to go done that rabbit hole again^^.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 633502,
      "author_name": "DatNT",
      "author_url": "",
      "post_date": "2019-09-25T02:58:36.477000",
      "content": "<p>or you can just use vertical flip augmentation </p>",
      "votes": 0,
      "replies": [
        {
          "id": 633836,
          "author_name": "srs",
          "author_url": "",
          "post_date": "2019-09-25T13:28:40.583000",
          "content": "<p>Probably, I was more coming from the classical neuro-imaging approach, where you commonly co-register your images to a template, so that at least the orientation and the position of your participant's / patient's heads are roughly the same. </p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "635263": "ImageOrientation just describe the dicom image border direction。It isn't related to the face and nose direction。",
    "633502": "or you can just use vertical flip augmentation ",
    "633161": "Dear all,\n\nI have a (possibly very naive) question regarding the ImageOrientation and ImagePosition fields. \n\nIf I understand it correctly they are information about the coordinate system of the CT-image. That is describing the position (i.e. size) of pixels / voxel, as well as the axis etc.  Described here: \n[https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037](https://dicom.innolitics.com/ciods/ct-image/image-plane/00200037)\n\nSo I might be able to use the information to define the position of a single voxel in space, and use it to calculate an affine transformation, to rotate and translate the original voxel to its matching position in a template image.\n\nQuestion:\nIf I want to have my images with the nose facing upwards and the head in the middle, the ImageOrientation, and ImagePosition won't help me that much, correct?\n\nThe slices will be shifted and rotated, so that the voxels are matching the other space, however, the relative head position might not be affected that much. To achieve my goal, I would need to do some more complex (coregistration,  matching) procedures.\n"
  }
}