I've seen a lot of talk about ordering slices within a series.
\nI made this notebook to show how to do it using DICOM tags -> https://www.kaggle.com/davidbroberts/determining-dicom-image-order
", "messages": [ { "id": "1486420", "postDate": "08/23/2021 00:52:11", "content": "I've seen a lot of talk about ordering slices within a series.
\nI made this notebook to show how to do it using DICOM tags -> https://www.kaggle.com/davidbroberts/determining-dicom-image-order
", "rawMarkdown": "I've seen a lot of talk about ordering slices within a series. \n\nI made this notebook to show how to do it using DICOM tags -> https://www.kaggle.com/davidbroberts/determining-dicom-image-order", "votes": null }, { "id": "1486536", "postDate": "08/23/2021 03:57:06", "content": "Thank you David,
\nIMHO, your approach only makes sense if we already know which coordinate system we are using, ie. in your code, you assume it is LPS (most of my Googling results the same for DICOM).
\nBut the competition data preparing paper says it uses RAI, and I found nothing in the DICOM tags to clearly state which coordinate system actually is. That's the part I'm still confused.
You can find the competition data preparation paper here: https://arxiv.org/pdf/2107.02314.pdf
", "rawMarkdown": "Thank you David,\nIMHO, your approach only makes sense if we already know which coordinate system we are using, ie. in your code, you assume it is LPS (most of my Googling results the same for DICOM). \nBut the competition data preparing paper says it uses RAI, and I found nothing in the DICOM tags to clearly state which coordinate system actually is. That's the part I'm still confused.\n\nYou can find the competition data preparation paper here: https://arxiv.org/pdf/2107.02314.pdf", "votes": null }, { "id": "1486555", "postDate": "08/23/2021 04:28:00", "content": "I think that paper is referencing the other part of this challenge where the data has been converted to NIfTI format. These are in DICOM format which is always in LPS.
", "rawMarkdown": "I think that paper is referencing the other part of this challenge where the data has been converted to NIfTI format. These are in DICOM format which is always in LPS.", "votes": null }, { "id": "1489045", "postDate": "08/24/2021 17:05:36", "content": "for some reason it seems to me that the coordinate system, especially for png images, is not important. if only the data fit into the model, and then you still process it in the same way, that is, in the same format as during training. I could be wrong.
", "rawMarkdown": "for some reason it seems to me that the coordinate system, especially for png images, is not important. if only the data fit into the model, and then you still process it in the same way, that is, in the same format as during training. I could be wrong.", "votes": null }, { "id": "1490046", "postDate": "08/25/2021 13:18:07", "content": "It's important if you care about the ordering of slices or intend on performing any type of measurement between slices.
", "rawMarkdown": "It's important if you care about the ordering of slices or intend on performing any type of measurement between slices.", "votes": null } ], "comments": [ { "id": 1486536, "author_name": "tungvs", "author_url": "", "post_date": "08/23/2021 03:57:06", "content": "Thank you David,
\nIMHO, your approach only makes sense if we already know which coordinate system we are using, ie. in your code, you assume it is LPS (most of my Googling results the same for DICOM).
\nBut the competition data preparing paper says it uses RAI, and I found nothing in the DICOM tags to clearly state which coordinate system actually is. That's the part I'm still confused.
You can find the competition data preparation paper here: https://arxiv.org/pdf/2107.02314.pdf
", "votes": null, "replies": [ { "id": 1486555, "author_name": "davidbroberts", "author_url": "", "post_date": "08/23/2021 04:28:00", "content": "I think that paper is referencing the other part of this challenge where the data has been converted to NIfTI format. These are in DICOM format which is always in LPS.
", "votes": null, "replies": [] }, { "id": 1489045, "author_name": "zaakciiru", "author_url": "", "post_date": "08/24/2021 17:05:36", "content": "for some reason it seems to me that the coordinate system, especially for png images, is not important. if only the data fit into the model, and then you still process it in the same way, that is, in the same format as during training. I could be wrong.
", "votes": null, "replies": [] }, { "id": 1490046, "author_name": "davidbroberts", "author_url": "", "post_date": "08/25/2021 13:18:07", "content": "It's important if you care about the ordering of slices or intend on performing any type of measurement between slices.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1486420": "I've seen a lot of talk about ordering slices within a series. \n\nI made this notebook to show how to do it using DICOM tags -> https://www.kaggle.com/davidbroberts/determining-dicom-image-order", "1486536": "Thank you David,\nIMHO, your approach only makes sense if we already know which coordinate system we are using, ie. in your code, you assume it is LPS (most of my Googling results the same for DICOM). \nBut the competition data preparing paper says it uses RAI, and I found nothing in the DICOM tags to clearly state which coordinate system actually is. That's the part I'm still confused.\n\nYou can find the competition data preparation paper here: https://arxiv.org/pdf/2107.02314.pdf", "1486555": "I think that paper is referencing the other part of this challenge where the data has been converted to NIfTI format. These are in DICOM format which is always in LPS.", "1489045": "for some reason it seems to me that the coordinate system, especially for png images, is not important. if only the data fit into the model, and then you still process it in the same way, that is, in the same format as during training. I could be wrong.", "1490046": "It's important if you care about the ordering of slices or intend on performing any type of measurement between slices." }, "source": "meta" }