Such fields like ImageOrientationPatient, ImagePositionPatient, FlipAngle, and etc look useful, but how to understand them correctly? Do I need to figure it out manually or there are the description?
", "messages": [ { "id": 1389614, "postDate": "2021-07-15T21:25:53.740Z", "content": "ImagePositionPatient and ImageOrientationPatient are important tags for MR because they relate 'slices' .. or images to each other.
\nThe ImagePositionPatient tag is used to relate the position of 'slices' relative to the patient's body in 3D space. It's made up of three separate values that are distances from an arbitrary center point. This tag tells us where an image lies along all three axis in 3D space.
\nWhen a patient is positioned inside the MR gantry (or CT), the technologist sets a reference point that all subsequent images are positioned from .. these measurements represent the three distances (X,Y,Z) from the center point (0,0,0).
\nImageOrientationPatient defines how a 2D image is rotated in 3D 'patient' space. That is, how the image is oriented relative to the patient's body. It has six values that are cosines which specify the degree of rotation. Three for each of the X and Y coordinates in the image. This allows for images to be rotated obliquely in odd angles, rather than just orthogonal perspectives.
\nThese tags will be of limited usage here because we only have the post-processed images in this dataset. Creating multiplanar reformations without the original volume of voxels is suboptimal.
\nHere's a notebook where I use ImageOrientationPatient tag to determine the plane of the series in this competition-> https://www.kaggle.com/davidbroberts/determining-mr-image-planes
\nHere's the complete list of DICOM tags -> http://dicom.nema.org/medical/Dicom/2017e/output/chtml/part06/chapter_6.html
", "rawMarkdown": "ImagePositionPatient and ImageOrientationPatient are important tags for MR because they relate 'slices' .. or images to each other.\n\nThe *ImagePositionPatient* tag is used to relate the position of 'slices' relative to the patient's body in 3D space. It's made up of three separate values that are distances from an arbitrary center point. This tag tells us where an image lies along all three axis in 3D space.\n\nWhen a patient is positioned inside the MR gantry (or CT), the technologist sets a reference point that all subsequent images are positioned from .. these measurements represent the three distances (X,Y,Z) from the center point (0,0,0).\n\n*ImageOrientationPatient* defines how a 2D image is rotated in 3D 'patient' space. That is, how the image is oriented relative to the patient's body. It has six values that are cosines which specify the degree of rotation. Three for each of the X and Y coordinates in the image. This allows for images to be rotated obliquely in odd angles, rather than just orthogonal perspectives. \n\nThese tags will be of limited usage here because we only have the post-processed images in this dataset. Creating multiplanar reformations without the original volume of voxels is suboptimal.\n\nHere's a notebook where I use ImageOrientationPatient tag to determine the plane of the series in this competition-> https://www.kaggle.com/davidbroberts/determining-mr-image-planes\n\nHere's the complete list of DICOM tags -> http://dicom.nema.org/medical/Dicom/2017e/output/chtml/part06/chapter_6.html", "votes": 7, "replies": [ { "id": 1390657, "postDate": "2021-07-17T00:06:36.530Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 1391547, "postDate": "2021-07-17T17:40:38.630Z", "content": "I learned quite a bit from your response @davidbroberts ! Thanks for sharing the two additional links as well. I am learning as I go in this competition since I am new to CV. It's been a pleasure to find your posts so informative! Thank you!
", "rawMarkdown": "I learned quite a bit from your response @davidbroberts ! Thanks for sharing the two additional links as well. I am learning as I go in this competition since I am new to CV. It's been a pleasure to find your posts so informative! Thank you!", "votes": 1 } ] }, { "id": 1389682, "postDate": "2021-07-16T00:16:17.640Z", "content": "One other thing to note about the ImagePostitionPatient tag is that it's the proper tag to use when ordering slices. It involves a few multiplanar calculations in some cases, but is the only accurate method to do so. In most cases, we can rely on the InstanceNumber tag (0x0010,0x0013) to order slices or the SliceLocation tag (0x0020,1041), but they're not always accurate. Of course ordering slices by filename is wildly inaccurate.
\nThe PatientPosition tag specifies the patient's orientation to the MR gantry .. HFS = Head First Supine, FFP = Feet First Prone.
\nProne vs Supine might be useful to know since gravity will prevail on the juicy brain bits somewhat. However, I doubt many (if any) patients were scanned in the prone position.
", "rawMarkdown": "One other thing to note about the ImagePostitionPatient tag is that it's *the* proper tag to use when ordering slices. It involves a few multiplanar calculations in some cases, but is the only accurate method to do so. In most cases, we can rely on the InstanceNumber tag (0x0010,0x0013) to order slices or the SliceLocation tag (0x0020,1041), but they're not always accurate. Of course ordering slices by filename is wildly inaccurate.\n\nThe PatientPosition tag specifies the patient's orientation to the MR gantry .. HFS = Head First Supine, FFP = Feet First Prone.\n\nProne vs Supine might be useful to know since gravity will prevail on the juicy brain bits somewhat. However, I doubt many (if any) patients were scanned in the prone position.", "votes": 3 }, { "id": 1389479, "postDate": "2021-07-15T18:34:31.750Z", "content": "Such fields like ImageOrientationPatient, ImagePositionPatient, FlipAngle, and etc look useful, but how to understand them correctly? Do I need to figure it out manually or there are the description?
", "rawMarkdown": "Such fields like ImageOrientationPatient, ImagePositionPatient, FlipAngle, and etc look useful, but how to understand them correctly? Do I need to figure it out manually or there are the description?", "votes": 4 } ], "comments": [ { "id": 1389614, "author_name": "David Roberts", "author_url": "", "post_date": "2021-07-15T21:25:53.740000", "content": "ImagePositionPatient and ImageOrientationPatient are important tags for MR because they relate 'slices' .. or images to each other.
\nThe ImagePositionPatient tag is used to relate the position of 'slices' relative to the patient's body in 3D space. It's made up of three separate values that are distances from an arbitrary center point. This tag tells us where an image lies along all three axis in 3D space.
\nWhen a patient is positioned inside the MR gantry (or CT), the technologist sets a reference point that all subsequent images are positioned from .. these measurements represent the three distances (X,Y,Z) from the center point (0,0,0).
\nImageOrientationPatient defines how a 2D image is rotated in 3D 'patient' space. That is, how the image is oriented relative to the patient's body. It has six values that are cosines which specify the degree of rotation. Three for each of the X and Y coordinates in the image. This allows for images to be rotated obliquely in odd angles, rather than just orthogonal perspectives.
\nThese tags will be of limited usage here because we only have the post-processed images in this dataset. Creating multiplanar reformations without the original volume of voxels is suboptimal.
\nHere's a notebook where I use ImageOrientationPatient tag to determine the plane of the series in this competition-> https://www.kaggle.com/davidbroberts/determining-mr-image-planes
\nHere's the complete list of DICOM tags -> http://dicom.nema.org/medical/Dicom/2017e/output/chtml/part06/chapter_6.html
", "votes": 7, "replies": [ { "id": 1390657, "author_name": "", "author_url": "", "post_date": "2021-07-17T00:06:36.530000", "content": "", "votes": 1, "replies": [] }, { "id": 1391547, "author_name": "ElenaEB", "author_url": "", "post_date": "2021-07-17T17:40:38.630000", "content": "I learned quite a bit from your response @davidbroberts ! Thanks for sharing the two additional links as well. I am learning as I go in this competition since I am new to CV. It's been a pleasure to find your posts so informative! Thank you!
", "votes": 1, "replies": [] } ] }, { "id": 1389682, "author_name": "David Roberts", "author_url": "", "post_date": "2021-07-16T00:16:17.640000", "content": "One other thing to note about the ImagePostitionPatient tag is that it's the proper tag to use when ordering slices. It involves a few multiplanar calculations in some cases, but is the only accurate method to do so. In most cases, we can rely on the InstanceNumber tag (0x0010,0x0013) to order slices or the SliceLocation tag (0x0020,1041), but they're not always accurate. Of course ordering slices by filename is wildly inaccurate.
\nThe PatientPosition tag specifies the patient's orientation to the MR gantry .. HFS = Head First Supine, FFP = Feet First Prone.
\nProne vs Supine might be useful to know since gravity will prevail on the juicy brain bits somewhat. However, I doubt many (if any) patients were scanned in the prone position.
", "votes": 3, "replies": [] } ], "raw_markdown_by_id": { "1389614": "ImagePositionPatient and ImageOrientationPatient are important tags for MR because they relate 'slices' .. or images to each other.\n\nThe *ImagePositionPatient* tag is used to relate the position of 'slices' relative to the patient's body in 3D space. It's made up of three separate values that are distances from an arbitrary center point. This tag tells us where an image lies along all three axis in 3D space.\n\nWhen a patient is positioned inside the MR gantry (or CT), the technologist sets a reference point that all subsequent images are positioned from .. these measurements represent the three distances (X,Y,Z) from the center point (0,0,0).\n\n*ImageOrientationPatient* defines how a 2D image is rotated in 3D 'patient' space. That is, how the image is oriented relative to the patient's body. It has six values that are cosines which specify the degree of rotation. Three for each of the X and Y coordinates in the image. This allows for images to be rotated obliquely in odd angles, rather than just orthogonal perspectives. \n\nThese tags will be of limited usage here because we only have the post-processed images in this dataset. Creating multiplanar reformations without the original volume of voxels is suboptimal.\n\nHere's a notebook where I use ImageOrientationPatient tag to determine the plane of the series in this competition-> https://www.kaggle.com/davidbroberts/determining-mr-image-planes\n\nHere's the complete list of DICOM tags -> http://dicom.nema.org/medical/Dicom/2017e/output/chtml/part06/chapter_6.html", "1389682": "One other thing to note about the ImagePostitionPatient tag is that it's *the* proper tag to use when ordering slices. It involves a few multiplanar calculations in some cases, but is the only accurate method to do so. In most cases, we can rely on the InstanceNumber tag (0x0010,0x0013) to order slices or the SliceLocation tag (0x0020,1041), but they're not always accurate. Of course ordering slices by filename is wildly inaccurate.\n\nThe PatientPosition tag specifies the patient's orientation to the MR gantry .. HFS = Head First Supine, FFP = Feet First Prone.\n\nProne vs Supine might be useful to know since gravity will prevail on the juicy brain bits somewhat. However, I doubt many (if any) patients were scanned in the prone position.", "1389479": "Such fields like ImageOrientationPatient, ImagePositionPatient, FlipAngle, and etc look useful, but how to understand them correctly? Do I need to figure it out manually or there are the description?" } }