![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fa3cd2919291cc62787667479bae21e76%2FSelection_096.png?generation=1721049749114050&alt=media\")
i am not a doctor, but l read that some people (?%) has L6?
\n
i am not a doctor, but l read that some people (?%) has L6?
\n
based on data doc, It seems named s1， in stead of L6?
", "rawMarkdown": "based on data doc, It seems named s1， in stead of L6?", "votes": null }, { "id": "2923139", "postDate": "07/15/2024 17:21:48", "content": "That's interesting. 10% seems too much and it could be in this dataset.
", "rawMarkdown": "That's interesting. 10% seems too much and it could be in this dataset.", "votes": null }, { "id": "2923289", "postDate": "07/15/2024 18:41:25", "content": "It can be hard to tell on a lumbar spine MRI without seeing the patient's entire spine. If a patient did have six lumbar-type vertebral bodies, then \"L6\" would probably end up being designated L5 or S1.
", "rawMarkdown": "It can be hard to tell on a lumbar spine MRI without seeing the patient's entire spine. If a patient did have six lumbar-type vertebral bodies, then \"L6\" would probably end up being designated L5 or S1.", "votes": null }, { "id": "2923298", "postDate": "07/15/2024 18:48:09", "content": "yes. i agree. this is also why sometimes it is difficult to apply AI in medical.
\nassume I build a shape/graph model with 5 points (L1 to L5), it may get confused.
abnormalities are OOD and AI predictions are unpredictable.
", "rawMarkdown": "yes. i agree. this is also why sometimes it is difficult to apply AI in medical. \nassume I build a shape/graph model with 5 points (L1 to L5), it may get confused. \n \n\nabnormalities are OOD and AI predictions are unpredictable.", "votes": null }, { "id": "2923435", "postDate": "07/15/2024 22:04:16", "content": "Looking through out-of-fold predictions, our model seems to get the vertebrae level incorrect on series_id 2421706412. Hard for me to tell, but maybe this patient has an L6?
![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5570735%2Fb5edda2fbdc16899b681f2ec3e2fb94b%2FCapture.JPG?generation=1721080796788452&alt=media\")
after googling, i managed to find how L6 (and L4 , … about 3% of population has only FOUR! ) look like:
\n![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F66c7830bc50e100d32d695051bef033d%2FSelection_107.png?generation=1721117852056545&alt=media\")
you really need to identify where is S1 and T12
", "rawMarkdown": "after googling, i managed to find how L6 (and L4 , ... about 3% of population has only FOUR! ) look like:\n\n\nyou really need to identify where is S1 and T12", "votes": null }, { "id": "2924321", "postDate": "07/16/2024 13:03:53", "content": "When encountering variant number of vertebral bodies, usually a radiology report will contain statements like \"There are six non-rib bearing vertebra. For the purpose of reporting, the first non-rib bearing vertebra is designated as L1\" or \"There are six non-rib bearing vertebra. For the purpose of reporting, the last non-rib bearing vertebra is designated as L5\". For this competition, since they have declared an L5-S1 intervertebral level to be graded, I think it is safe to assume that the vertebral body above the sacrum is always L5.
", "rawMarkdown": "When encountering variant number of vertebral bodies, usually a radiology report will contain statements like \"There are six non-rib bearing vertebra. For the purpose of reporting, the first non-rib bearing vertebra is designated as L1\" or \"There are six non-rib bearing vertebra. For the purpose of reporting, the last non-rib bearing vertebra is designated as L5\". For this competition, since they have declared an L5-S1 intervertebral level to be graded, I think it is safe to assume that the vertebral body above the sacrum is always L5.", "votes": null }, { "id": "2924923", "postDate": "07/16/2024 19:54:07", "content": "Here is another interesting example series_id=1212326388. In this series it looks like the left and right labels are given to different levels of the spine. Maybe locating L5 is hard for experts too..
Similar examples: [3893961572, 3620276777, 3657125167, 3625340069, 1724042057]
![\"image\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5570735%2F58e8723b146015d8e577221f372ac528%2FCapture.JPG?generation=1721159426224378&alt=media\")
i read that there is such a thing called fused L5. it is naturally fused into S1. anyway, thanks for showing these. i need to adjust my method a bit.
", "rawMarkdown": "i read that there is such a thing called fused L5. it is naturally fused into S1. anyway, thanks for showing these. i need to adjust my method a bit.", "votes": null }, { "id": "2925814", "postDate": "07/17/2024 12:02:19", "content": "just to add few more, this is where my model got confused ![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F6259210%2F8e6c688e9b1d8c65429534e30f701031%2Ffew%20more.png?generation=1721217684312651&alt=media\")
delete. I'll have to take a look at the dataset a little more.
", "rawMarkdown": "delete. I'll have to take a look at the dataset a little more.", "votes": null } ], "comments": [ { "id": 2922797, "author_name": "dragonzhang", "author_url": "", "post_date": "07/15/2024 13:56:27", "content": "based on data doc, It seems named s1， in stead of L6?
", "votes": null, "replies": [] }, { "id": 2923139, "author_name": "gunesevitan", "author_url": "", "post_date": "07/15/2024 17:21:48", "content": "That's interesting. 10% seems too much and it could be in this dataset.
", "votes": null, "replies": [] }, { "id": 2923289, "author_name": "vaillant", "author_url": "", "post_date": "07/15/2024 18:41:25", "content": "It can be hard to tell on a lumbar spine MRI without seeing the patient's entire spine. If a patient did have six lumbar-type vertebral bodies, then \"L6\" would probably end up being designated L5 or S1.
", "votes": null, "replies": [ { "id": 2923298, "author_name": "hengck23", "author_url": "", "post_date": "07/15/2024 18:48:09", "content": "yes. i agree. this is also why sometimes it is difficult to apply AI in medical.
\nassume I build a shape/graph model with 5 points (L1 to L5), it may get confused.
abnormalities are OOD and AI predictions are unpredictable.
", "votes": null, "replies": [ { "id": 2923435, "author_name": "brendanartley", "author_url": "", "post_date": "07/15/2024 22:04:16", "content": "Looking through out-of-fold predictions, our model seems to get the vertebrae level incorrect on series_id 2421706412. Hard for me to tell, but maybe this patient has an L6?
after googling, i managed to find how L6 (and L4 , … about 3% of population has only FOUR! ) look like:
\n
you really need to identify where is S1 and T12
", "votes": null, "replies": [ { "id": 2924923, "author_name": "brendanartley", "author_url": "", "post_date": "07/16/2024 19:54:07", "content": "Here is another interesting example series_id=1212326388. In this series it looks like the left and right labels are given to different levels of the spine. Maybe locating L5 is hard for experts too..
Similar examples: [3893961572, 3620276777, 3657125167, 3625340069, 1724042057]
i read that there is such a thing called fused L5. it is naturally fused into S1. anyway, thanks for showing these. i need to adjust my method a bit.
", "votes": null, "replies": [] }, { "id": 2925915, "author_name": "yeeseng", "author_url": "", "post_date": "07/17/2024 13:22:45", "content": "delete. I'll have to take a look at the dataset a little more.
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 2925814, "author_name": "sergiosaharovskiy", "author_url": "", "post_date": "07/17/2024 12:02:19", "content": "just to add few more, this is where my model got confused
When encountering variant number of vertebral bodies, usually a radiology report will contain statements like \"There are six non-rib bearing vertebra. For the purpose of reporting, the first non-rib bearing vertebra is designated as L1\" or \"There are six non-rib bearing vertebra. For the purpose of reporting, the last non-rib bearing vertebra is designated as L5\". For this competition, since they have declared an L5-S1 intervertebral level to be graded, I think it is safe to assume that the vertebral body above the sacrum is always L5.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2922759": "i am not a doctor, but l read that some people (?%) has L6?\n\n\n", "2922797": "based on data doc, It seems named s1， in stead of L6?", "2923139": "That's interesting. 10% seems too much and it could be in this dataset.", "2923289": "It can be hard to tell on a lumbar spine MRI without seeing the patient's entire spine. If a patient did have six lumbar-type vertebral bodies, then \"L6\" would probably end up being designated L5 or S1.", "2923298": "yes. i agree. this is also why sometimes it is difficult to apply AI in medical. \nassume I build a shape/graph model with 5 points (L1 to L5), it may get confused. \n \n\nabnormalities are OOD and AI predictions are unpredictable.", "2923435": "Looking through out-of-fold predictions, our model seems to get the vertebrae level incorrect on series_id `2421706412`. Hard for me to tell, but maybe this patient has an L6?\n\n\n", "2923971": "after googling, i managed to find how L6 (and L4 , ... about 3% of population has only FOUR! ) look like:\n\n\nyou really need to identify where is S1 and T12", "2924321": "When encountering variant number of vertebral bodies, usually a radiology report will contain statements like \"There are six non-rib bearing vertebra. For the purpose of reporting, the first non-rib bearing vertebra is designated as L1\" or \"There are six non-rib bearing vertebra. For the purpose of reporting, the last non-rib bearing vertebra is designated as L5\". For this competition, since they have declared an L5-S1 intervertebral level to be graded, I think it is safe to assume that the vertebral body above the sacrum is always L5.", "2924923": "Here is another interesting example `series_id=1212326388`. In this series it looks like the left and right labels are given to different levels of the spine. Maybe locating L5 is hard for experts too..\n\nSimilar examples: `[3893961572, 3620276777, 3657125167, 3625340069, 1724042057]`\n\n", "2925180": "i read that there is such a thing called fused L5. it is naturally fused into S1. anyway, thanks for showing these. i need to adjust my method a bit.", "2925814": "just to add few more, this is where my model got confused ", "2925915": "delete. I'll have to take a look at the dataset a little more." }, "source": "meta" }