Hello,
\nI've read a pinned discussion regarding the dataset, but I faced some problems with retrieving image from segmentation series by its InstanceNumber field value.
\nThere are some cases when the image on the right (from segmentations folder) doesn't correspond to the one on the left (series folder). It there some metadata that should be retrieved to change the order of frames in the specific series or just a data issue?
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11673033%2F9d2a8a7c6e1d69048c4d6ae8d9b258a3%2Fscreenshot.png?generation=1756240734879885&alt=media\")
That's order problem
", "rawMarkdown": "That's order problem", "votes": 1 }, { "id": 3276693, "postDate": "2025-08-26T20:39:42.573Z", "content": "Hello,
\nI've read a pinned discussion regarding the dataset, but I faced some problems with retrieving image from segmentation series by its InstanceNumber field value.
\nThere are some cases when the image on the right (from segmentations folder) doesn't correspond to the one on the left (series folder). It there some metadata that should be retrieved to change the order of frames in the specific series or just a data issue?
\n
Can you help me understand what you are trying to do? The reason that we provided the segmentation NIfTIs alongside corresponding image NIfTIs was precisely to avoid these types of issues with slice order/orientation when converting from DICOM to NIfTI.
", "rawMarkdown": "Can you help me understand what you are trying to do? The reason that we provided the segmentation NIfTIs alongside corresponding image NIfTIs was precisely to avoid these types of issues with slice order/orientation when converting from DICOM to NIfTI.", "replies": [ { "id": 3280582, "postDate": "2025-09-02T21:34:21.887Z", "content": "I was trying to retrieve the index of the slice where the aneurysm was located, along with the corresponding mask. To do this, I used the index of the SOPInstanceUID within the SeriesInstanceUID from train_localizers.csv.
\nHowever, the index and therefore the image retrieved from the NIfTI vessel segmentations did not always match the corresponding image from the series. In some cases, the slices seem to need sorting in descending order.
\nMy question is: is this purely an ordering problem, or is there metadata that can help determine whether reordering is needed?
That's order problem
", "votes": 1, "replies": [] }, { "id": 3280282, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-09-02T12:32:12.073000", "content": "Can you help me understand what you are trying to do? The reason that we provided the segmentation NIfTIs alongside corresponding image NIfTIs was precisely to avoid these types of issues with slice order/orientation when converting from DICOM to NIfTI.
", "votes": 0, "replies": [ { "id": 3280582, "author_name": "ostamax", "author_url": "", "post_date": "2025-09-02T21:34:21.887000", "content": "I was trying to retrieve the index of the slice where the aneurysm was located, along with the corresponding mask. To do this, I used the index of the SOPInstanceUID within the SeriesInstanceUID from train_localizers.csv.
\nHowever, the index and therefore the image retrieved from the NIfTI vessel segmentations did not always match the corresponding image from the series. In some cases, the slices seem to need sorting in descending order.
\nMy question is: is this purely an ordering problem, or is there metadata that can help determine whether reordering is needed?