\n
![\"from](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3163942%2F036863bb8e72d78518d3495931612437%2F1.2.826.0.1.3680043.8.498.11887329867812275491160566603814454129.png?generation=1759219393008646&alt=media\")
We found as many as 28 training images have wrong orientations as viewed in DICOM viewers. They are all:
\nT2 modality and multi-frame DCM.
Example 1: Left and Right (LR) is wrong
\n
Example 2: Anterior and Posterior (AP) is wrong
\n![\"from](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3163942%2F79f6d70e21305e0c22eac7ea92b88838%2F1.2.826.0.1.3680043.8.498.11915319973409844345177713085783065237.png?generation=1759219460676415&alt=media\")
While writing this post, we discovered that two previous posts in the discussion had raised similar issues:
\nThis is beyond just a response \"This reflects real world data\" from organizer's most recent response, as the same organizer replied 2 months ago with \"OK thanks for update. We will look into these and see if we can find a fix.\" Can the organizers clarify if they have been finding a fix as stated 2 months ago, or are we supposed to just pretend this systematic error is \"just real world data\"? Also, it will be very helpful if the organizers can clarify if this error in the DCM orientation in T2 MultiFrame Dicom is an issue in the test set with the same prevalance?
\nThank you!
", "messages": [ { "id": 3296115, "postDate": "2025-09-30T08:16:19.153Z", "content": "We found as many as 28 training images have wrong orientations as viewed in DICOM viewers. They are all:
\nT2 modality and multi-frame DCM.
Example 1: Left and Right (LR) is wrong
\n
Example 2: Anterior and Posterior (AP) is wrong
\n
While writing this post, we discovered that two previous posts in the discussion had raised similar issues:
\nThis is beyond just a response \"This reflects real world data\" from organizer's most recent response, as the same organizer replied 2 months ago with \"OK thanks for update. We will look into these and see if we can find a fix.\" Can the organizers clarify if they have been finding a fix as stated 2 months ago, or are we supposed to just pretend this systematic error is \"just real world data\"? Also, it will be very helpful if the organizers can clarify if this error in the DCM orientation in T2 MultiFrame Dicom is an issue in the test set with the same prevalance?
\nThank you!
", "rawMarkdown": "We found as many as 28 training images have wrong orientations as viewed in DICOM viewers. They are all:\n**T2 modality and multi-frame DCM**.\n\nExample 1: Left and Right (LR) is wrong\n\n\nExample 2: Anterior and Posterior (AP) is wrong\n\n\nWhile writing this post, we discovered that two previous posts in the discussion had raised similar issues:\n1. [2 months ago posted, no update from organizer] MultiFrame ImageOrientationPatient wrong (https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/598828)\n2. [5 days ago posted, organizer said this is \"just real world data\"] Incorrect ImageOrientationPatient for multi-frame dicom (https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/609271)\n\nThis is beyond just a response \"This reflects real world data\" from organizer's most recent response, as the same organizer replied 2 months ago with \"OK thanks for update. We will look into these and see if we can find a fix.\" Can the organizers clarify if they have been finding a fix as stated 2 months ago, or are we supposed to just pretend this systematic error is \"just real world data\"? Also, it will be very helpful if the organizers can clarify if this error in the DCM orientation in T2 MultiFrame Dicom is an issue in the test set with the same prevalance?\n\nThank you!\n\n", "votes": 13 }, { "id": 3296798, "postDate": "2025-10-01T16:20:19.643Z", "content": "We looked into it, and this is how the data was received from the clinical site, so it reflects real world data. We are not planning to fix this, particularly given the short time left in the competition. I can't comment on the contents of the test set.
", "rawMarkdown": "We looked into it, and this is how the data was received from the clinical site, so it reflects real world data. We are not planning to fix this, particularly given the short time left in the competition. I can't comment on the contents of the test set.", "votes": 1, "replies": [ { "id": 3300842, "postDate": "2025-10-11T17:15:05.740Z", "content": "Thanks for your clarification!
", "rawMarkdown": "Thanks for your clarification!" } ] } ], "comments": [ { "id": 3296798, "author_name": "Evan Calabrese", "author_url": "", "post_date": "2025-10-01T16:20:19.643000", "content": "We looked into it, and this is how the data was received from the clinical site, so it reflects real world data. We are not planning to fix this, particularly given the short time left in the competition. I can't comment on the contents of the test set.
", "votes": 1, "replies": [ { "id": 3300842, "author_name": "MakeLoveAndPeace", "author_url": "", "post_date": "2025-10-11T17:15:05.740000", "content": "Thanks for your clarification!
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "3296115": "We found as many as 28 training images have wrong orientations as viewed in DICOM viewers. They are all:\n**T2 modality and multi-frame DCM**.\n\nExample 1: Left and Right (LR) is wrong\n\n\nExample 2: Anterior and Posterior (AP) is wrong\n\n\nWhile writing this post, we discovered that two previous posts in the discussion had raised similar issues:\n1. [2 months ago posted, no update from organizer] MultiFrame ImageOrientationPatient wrong (https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/598828)\n2. [5 days ago posted, organizer said this is \"just real world data\"] Incorrect ImageOrientationPatient for multi-frame dicom (https://www.kaggle.com/competitions/rsna-intracranial-aneurysm-detection/discussion/609271)\n\nThis is beyond just a response \"This reflects real world data\" from organizer's most recent response, as the same organizer replied 2 months ago with \"OK thanks for update. We will look into these and see if we can find a fix.\" Can the organizers clarify if they have been finding a fix as stated 2 months ago, or are we supposed to just pretend this systematic error is \"just real world data\"? Also, it will be very helpful if the organizers can clarify if this error in the DCM orientation in T2 MultiFrame Dicom is an issue in the test set with the same prevalance?\n\nThank you!\n\n", "3296798": "We looked into it, and this is how the data was received from the clinical site, so it reflects real world data. We are not planning to fix this, particularly given the short time left in the competition. I can't comment on the contents of the test set." } }