{ "id": 611032, "title": "Fun fact? Two Dicom series discovered to be from the same patient!", "url": "/competitions/rsna-intracranial-aneurysm-detection/discussion/611032", "author_name": "MakeLoveAndPeace", "post_date": "2025-10-08T10:25:27.445000", "votes": 5, "comment_count": 4, "views": 0, "content": "

By inspecting the images, we discovered two Dicom series with different modalities are actually from the same patient:

\n

SeriesInstanceUID PatientAge PatientSex Modality
\n1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896 58 Female CTA
\n1.2.826.0.1.3680043.8.498.28062264737209503799520847652265280868 59 Female MRA

\n

These two Dicom series even have different ages. One is 58, and the other is 59. Assuming the age is not anonymized, then the two scans might be follow-up scans one year apart.

\n

I post this observation as a fun fact. But on a serious note, is the test and train split strictly by patient? If not, the train and test sets can have leakage like in this post. It will be good if the organizers can clarify on the split criteria.

\n

Thanks!

", "messages": [ { "id": 3299539, "postDate": "2025-10-08T10:25:27.447Z", "content": "

By inspecting the images, we discovered two Dicom series with different modalities are actually from the same patient:

\n

SeriesInstanceUID PatientAge PatientSex Modality
\n1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896 58 Female CTA
\n1.2.826.0.1.3680043.8.498.28062264737209503799520847652265280868 59 Female MRA

\n

These two Dicom series even have different ages. One is 58, and the other is 59. Assuming the age is not anonymized, then the two scans might be follow-up scans one year apart.

\n

I post this observation as a fun fact. But on a serious note, is the test and train split strictly by patient? If not, the train and test sets can have leakage like in this post. It will be good if the organizers can clarify on the split criteria.

\n

Thanks!

", "rawMarkdown": "By inspecting the images, we discovered two Dicom series with different modalities are actually **from the same patient**:\n\nSeriesInstanceUID\tPatientAge\tPatientSex\tModality\n1.2.826.0.1.3680043.8.498.**11283060280124241416746713304089513896**\t**58\tFemale\tCTA**\n1.2.826.0.1.3680043.8.498.**28062264737209503799520847652265280868**\t**59\tFemale\tMRA**\n\nThese two Dicom series even have different ages. One is 58, and the other is 59. Assuming the age is not anonymized, then the two scans might be follow-up scans one year apart.\n\nI post this observation as a fun fact. But on a serious note, is the test and train split strictly by patient? If not, the train and test sets can have leakage like in this post. It will be good if the organizers can clarify on the split criteria.\n\nThanks!", "votes": 5 }, { "id": 3300175, "postDate": "2025-10-09T18:24:03.250Z", "content": "

No overlap between patient is in the train and test

", "rawMarkdown": "No overlap between patient is in the train and test", "votes": 2, "replies": [ { "id": 3300841, "postDate": "2025-10-11T17:13:51.670Z", "content": "

Many thanks for the clarification!

", "rawMarkdown": "Many thanks for the clarification!" } ] }, { "id": 3299548, "postDate": "2025-10-08T10:33:51.437Z", "content": "

Note: While these two Dicom series have the same aneurysm location labels, the \"coordinates\" in x,y are labelled differently. Especially for 1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896, where the aneurysm coordinates for BA-tip and Posterior aneurysms are pointing to the SAME location:

\n

{'x': 292.77092978395063, 'y': 230.5624125916281} Posterior
\n{'x': 292.77092978395063, 'y': 231.35253604841822} BA Tip

\n

The other Dicom series gives a much more reasonable coordinate that are actually NOT marking the same location.

\n

It is thus reasonable to assume that the coordinates for 11283060's BA and POS aneurysms are wrong.

", "rawMarkdown": "Note: While these two Dicom series have the same aneurysm location labels, the \"coordinates\" in x,y are labelled differently. Especially for 1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896, where the aneurysm coordinates for BA-tip and Posterior aneurysms are **pointing to the SAME location:**\n\n{'x': 292.77092978395063, 'y': 230.5624125916281}\tPosterior\n{'x': 292.77092978395063, 'y': 231.35253604841822}\tBA Tip\n\nThe other Dicom series gives a much more reasonable coordinate that are actually NOT marking the same location.\n\nIt is thus reasonable to assume that the coordinates for 11283060's BA and POS aneurysms are wrong.\n", "votes": 2, "replies": [ { "id": 3299627, "postDate": "2025-10-08T13:51:58.597Z", "content": "

BA-tip aneurysms are technically a subset of posterior circulation aneurysms, which is probably why they are so close.

", "rawMarkdown": "BA-tip aneurysms are technically a subset of posterior circulation aneurysms, which is probably why they are so close." } ] } ], "comments": [ { "id": 3300175, "author_name": "JeffRudie", "author_url": "", "post_date": "2025-10-09T18:24:03.250000", "content": "

No overlap between patient is in the train and test

", "votes": 2, "replies": [ { "id": 3300841, "author_name": "MakeLoveAndPeace", "author_url": "", "post_date": "2025-10-11T17:13:51.670000", "content": "

Many thanks for the clarification!

", "votes": 0, "replies": [] } ] }, { "id": 3299548, "author_name": "MakeLoveAndPeace", "author_url": "", "post_date": "2025-10-08T10:33:51.437000", "content": "

Note: While these two Dicom series have the same aneurysm location labels, the \"coordinates\" in x,y are labelled differently. Especially for 1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896, where the aneurysm coordinates for BA-tip and Posterior aneurysms are pointing to the SAME location:

\n

{'x': 292.77092978395063, 'y': 230.5624125916281} Posterior
\n{'x': 292.77092978395063, 'y': 231.35253604841822} BA Tip

\n

The other Dicom series gives a much more reasonable coordinate that are actually NOT marking the same location.

\n

It is thus reasonable to assume that the coordinates for 11283060's BA and POS aneurysms are wrong.

", "votes": 2, "replies": [ { "id": 3299627, "author_name": "aastha5657hello", "author_url": "", "post_date": "2025-10-08T13:51:58.597000", "content": "

BA-tip aneurysms are technically a subset of posterior circulation aneurysms, which is probably why they are so close.

", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "3299539": "By inspecting the images, we discovered two Dicom series with different modalities are actually **from the same patient**:\n\nSeriesInstanceUID\tPatientAge\tPatientSex\tModality\n1.2.826.0.1.3680043.8.498.**11283060280124241416746713304089513896**\t**58\tFemale\tCTA**\n1.2.826.0.1.3680043.8.498.**28062264737209503799520847652265280868**\t**59\tFemale\tMRA**\n\nThese two Dicom series even have different ages. One is 58, and the other is 59. Assuming the age is not anonymized, then the two scans might be follow-up scans one year apart.\n\nI post this observation as a fun fact. But on a serious note, is the test and train split strictly by patient? If not, the train and test sets can have leakage like in this post. It will be good if the organizers can clarify on the split criteria.\n\nThanks!", "3300175": "No overlap between patient is in the train and test", "3299548": "Note: While these two Dicom series have the same aneurysm location labels, the \"coordinates\" in x,y are labelled differently. Especially for 1.2.826.0.1.3680043.8.498.11283060280124241416746713304089513896, where the aneurysm coordinates for BA-tip and Posterior aneurysms are **pointing to the SAME location:**\n\n{'x': 292.77092978395063, 'y': 230.5624125916281}\tPosterior\n{'x': 292.77092978395063, 'y': 231.35253604841822}\tBA Tip\n\nThe other Dicom series gives a much more reasonable coordinate that are actually NOT marking the same location.\n\nIt is thus reasonable to assume that the coordinates for 11283060's BA and POS aneurysms are wrong.\n" } }