{ "id": 266729, "title": "How to group data by institution?", "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/266729", "author_name": "", "post_date": "2021-08-20T06:54:47.750023600Z", "votes": 1, "comment_count": 3, "views": 0, "content": "

Is there a way to find out which cases originate from the same institution?

\n

Many times it is mentioned that the dataset is multi-institutional, and I would like to use it for federated learning research.

", "messages": [ { "id": "1482591", "postDate": "08/20/2021 06:54:47", "content": "

Is there a way to find out which cases originate from the same institution?

\n

Many times it is mentioned that the dataset is multi-institutional, and I would like to use it for federated learning research.

", "rawMarkdown": "Is there a way to find out which cases originate from the same institution?\n\nMany times it is mentioned that the dataset is multi-institutional, and I would like to use it for federated learning research.", "votes": null }, { "id": "1483080", "postDate": "08/20/2021 12:47:19", "content": "

You could look at the DICOM StudyInstanceUID tag (0020,000D) to get some idea of where the study was done. It's not recommended in the real world, but might offer some more information here.

", "rawMarkdown": "You could look at the DICOM StudyInstanceUID tag (0020,000D) to get some idea of where the study was done. It's not recommended in the real world, but might offer some more information here.", "votes": null }, { "id": "1487571", "postDate": "08/23/2021 17:44:38", "content": "

I don't really see the pattern to extract more information from these, is there anything else?

\n

(0020, 000D)
\n(0020, 000E)

", "rawMarkdown": "I don't really see the pattern to extract more information from these, is there anything else?\n\n\n[(0020, 000D)](https://1drv.ms/t/s!AnowZL60xcbVmbgfdO_XxuRg6JPHTA?e=SJA88d)\n[(0020, 000E)](https://1drv.ms/t/s!AnowZL60xcbVmbggSDGJYojbEwn7uQ?e=uZFf8I)", "votes": null }, { "id": "1487595", "postDate": "08/23/2021 17:57:47", "content": "

You can break the UID down into it's components and compare the country and organization IDs across studies to see which were performed at the same location. You may be able to determine the specific machine the study was done on as well.

\n

These studies have been thoroughly anonymized. I have not checked whether specific UID elements
\n were anon'd though.

\n

You can look up the UID structure in the DICOM standard -> http://dicom.nema.org/dicom/2013/output/chtml/part05/chapter_C.html

", "rawMarkdown": "You can break the UID down into it's components and compare the country and organization IDs across studies to see which were performed at the same location. You may be able to determine the specific machine the study was done on as well. \n\nThese studies have been thoroughly anonymized. I have not checked whether specific UID elements \n were anon'd though. \n\nYou can look up the UID structure in the DICOM standard -> http://dicom.nema.org/dicom/2013/output/chtml/part05/chapter_C.html", "votes": null } ], "comments": [ { "id": 1483080, "author_name": "davidbroberts", "author_url": "", "post_date": "08/20/2021 12:47:19", "content": "

You could look at the DICOM StudyInstanceUID tag (0020,000D) to get some idea of where the study was done. It's not recommended in the real world, but might offer some more information here.

", "votes": null, "replies": [] }, { "id": 1487571, "author_name": "raphaelschaefer", "author_url": "", "post_date": "08/23/2021 17:44:38", "content": "

I don't really see the pattern to extract more information from these, is there anything else?

\n

(0020, 000D)
\n(0020, 000E)

", "votes": null, "replies": [ { "id": 1487595, "author_name": "davidbroberts", "author_url": "", "post_date": "08/23/2021 17:57:47", "content": "

You can break the UID down into it's components and compare the country and organization IDs across studies to see which were performed at the same location. You may be able to determine the specific machine the study was done on as well.

\n

These studies have been thoroughly anonymized. I have not checked whether specific UID elements
\n were anon'd though.

\n

You can look up the UID structure in the DICOM standard -> http://dicom.nema.org/dicom/2013/output/chtml/part05/chapter_C.html

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1482591": "Is there a way to find out which cases originate from the same institution?\n\nMany times it is mentioned that the dataset is multi-institutional, and I would like to use it for federated learning research.", "1483080": "You could look at the DICOM StudyInstanceUID tag (0020,000D) to get some idea of where the study was done. It's not recommended in the real world, but might offer some more information here.", "1487571": "I don't really see the pattern to extract more information from these, is there anything else?\n\n\n[(0020, 000D)](https://1drv.ms/t/s!AnowZL60xcbVmbgfdO_XxuRg6JPHTA?e=SJA88d)\n[(0020, 000E)](https://1drv.ms/t/s!AnowZL60xcbVmbggSDGJYojbEwn7uQ?e=uZFf8I)", "1487595": "You can break the UID down into it's components and compare the country and organization IDs across studies to see which were performed at the same location. You may be able to determine the specific machine the study was done on as well. \n\nThese studies have been thoroughly anonymized. I have not checked whether specific UID elements \n were anon'd though. \n\nYou can look up the UID structure in the DICOM standard -> http://dicom.nema.org/dicom/2013/output/chtml/part05/chapter_C.html" }, "source": "meta" }