After reviewing the vessel segmentations, I noticed for some of them, when you load in the .nii file, the segmentations are actually reversed from the images along axis 2.
\nFor example:
\n = nib.load().get_fdata()\n = nib.load().get_fdata()\n\n\n = img[:, :, ::-]\nI have attached a list of these series here.
", "messages": [ { "id": 3258630, "postDate": "2025-07-31T01:27:36.367Z", "content": "After reviewing the vessel segmentations, I noticed for some of them, when you load in the .nii file, the segmentations are actually reversed from the images along axis 2.
\nFor example:
\n = nib.load().get_fdata()\n = nib.load().get_fdata()\n\n\n = img[:, :, ::-]\nI have attached a list of these series here.
", "rawMarkdown": "After reviewing the vessel segmentations, I noticed for some of them, when you load in the .nii file, the segmentations are actually reversed from the images along axis 2.\n\nFor example:\n```\nimg = nib.load(\"a.nii\").get_fdata()\nseg = nib.load(\"a_cowseg.nii\").get_fdata()\n\n# need to reverse img (or seg) along axis 2\nimg = img[:, :, ::-1]\n```\n\nI have attached a list of these series here.", "votes": 38 }, { "id": 3258932, "postDate": "2025-07-31T14:08:13.793Z", "content": "We are looking into this and will provide a fix as soon as possible. Should be a straightforward fix, but we will want to test/review to make sure its correct.
", "rawMarkdown": "We are looking into this and will provide a fix as soon as possible. Should be a straightforward fix, but we will want to test/review to make sure its correct.", "votes": 2, "replies": [ { "id": 3266238, "postDate": "2025-08-08T16:36:52.933Z", "content": "Hi Evan, Could you confirm this has been fixed so the current dataset is the final? Thanks in advance.
", "rawMarkdown": "Hi Evan, Could you confirm this has been fixed so the current dataset is the final? Thanks in advance.", "replies": [ { "id": 3266269, "postDate": "2025-08-08T17:26:58.757Z", "content": "The fix has not been implemented yet, as we are trying to release a single fix for multiple small issues given the complexity of all the data consistency checks required whenever we update the dataset.
\nHowever, using the list and code snippet provided by @vaillant you can run the fix on your local copy of the data.
", "rawMarkdown": "The fix has not been implemented yet, as we are trying to release a single fix for multiple small issues given the complexity of all the data consistency checks required whenever we update the dataset.\n\nHowever, using the list and code snippet provided by @vaillant you can run the fix on your local copy of the data.", "votes": 2 } ] } ] }, { "id": 3277585, "postDate": "2025-08-28T10:04:22.370Z", "content": "To ensure consistent alignment between images and segmentations, you can use MONAI’s Orientationd transform, which standardizes orientation across both files.
\ntransforms = monai.transforms.Compose([\n monai.transforms.LoadImaged(\n keys=[, ], \n =, \n =\n ),\n monai.transforms.Orientationd(\n keys=[, ], \n = # Reorients a common RAS orientation\n ),\n])\nThank you for your sharing!
", "rawMarkdown": "Thank you for your sharing!" }, { "id": 3259097, "postDate": "2025-07-31T21:29:16.577Z", "content": "Thank you for your sharing!
", "rawMarkdown": "Thank you for your sharing!" }, { "id": 3258872, "postDate": "2025-07-31T11:22:55.077Z", "content": "Thank you so much for your findings!
", "rawMarkdown": "Thank you so much for your findings!" } ], "comments": [ { "id": 3258932, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-07-31T14:08:13.793000", "content": "We are looking into this and will provide a fix as soon as possible. Should be a straightforward fix, but we will want to test/review to make sure its correct.
", "votes": 2, "replies": [ { "id": 3266238, "author_name": "FullEmpty", "author_url": "", "post_date": "2025-08-08T16:36:52.933000", "content": "Hi Evan, Could you confirm this has been fixed so the current dataset is the final? Thanks in advance.
", "votes": 0, "replies": [ { "id": 3266269, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-08-08T17:26:58.757000", "content": "The fix has not been implemented yet, as we are trying to release a single fix for multiple small issues given the complexity of all the data consistency checks required whenever we update the dataset.
\nHowever, using the list and code snippet provided by @vaillant you can run the fix on your local copy of the data.
", "votes": 2, "replies": [] } ] } ] }, { "id": 3277585, "author_name": "Dias Tashev, MD", "author_url": "", "post_date": "2025-08-28T10:04:22.370000", "content": "To ensure consistent alignment between images and segmentations, you can use MONAI’s Orientationd transform, which standardizes orientation across both files.
\ntransforms = monai.transforms.Compose([\n monai.transforms.LoadImaged(\n keys=[, ], \n =, \n =\n ),\n monai.transforms.Orientationd(\n keys=[, ], \n = # Reorients a common RAS orientation\n ),\n])\nThank you for your sharing!
", "votes": 0, "replies": [] }, { "id": 3259097, "author_name": "ABDELHAK OUANZOUGUI", "author_url": "", "post_date": "2025-07-31T21:29:16.577000", "content": "Thank you for your sharing!
", "votes": 0, "replies": [] }, { "id": 3258872, "author_name": "Khang Nguyen Le", "author_url": "", "post_date": "2025-07-31T11:22:55.077000", "content": "Thank you so much for your findings!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3258630": "After reviewing the vessel segmentations, I noticed for some of them, when you load in the .nii file, the segmentations are actually reversed from the images along axis 2.\n\nFor example:\n```\nimg = nib.load(\"a.nii\").get_fdata()\nseg = nib.load(\"a_cowseg.nii\").get_fdata()\n\n# need to reverse img (or seg) along axis 2\nimg = img[:, :, ::-1]\n```\n\nI have attached a list of these series here.", "3258932": "We are looking into this and will provide a fix as soon as possible. Should be a straightforward fix, but we will want to test/review to make sure its correct.", "3277585": "To ensure consistent alignment between images and segmentations, you can use MONAI’s Orientationd transform, which standardizes orientation across both files.\n```\ntransforms = monai.transforms.Compose([\n monai.transforms.LoadImaged(\n keys=[\"img_file\", \"seg_file\"], \n image_only=True, \n ensure_channel_first=True\n ),\n monai.transforms.Orientationd(\n keys=[\"img_file\", \"seg_file\"], \n axcodes=\"RAS\" # Reorients to a common RAS orientation\n ),\n])\n```", "3261709": "Thank you for your sharing!", "3259097": "Thank you for your sharing!", "3258872": "Thank you so much for your findings!" } }