Why only subset of the data is provided? Why not provide the data for all training set?
\ntrain_bounding_boxes.csv Bounding boxes for a subset of the training set.\n\nsegmentations/ Pixel level annotations for a subset of the training set.\nYou have to keep in mind that segmentation of medical images is very time consuming, especially for 3D data.
\nAnd it must be done by a medical expert, so it's also potentially expensive.
I guess the host tried to give as much as possible but the entire dataset was just too much.
", "rawMarkdown": "You have to keep in mind that segmentation of medical images is very time consuming, especially for 3D data.\nAnd it must be done by a medical expert, so it's also potentially expensive.\n\nI guess the host tried to give as much as possible but the entire dataset was just too much.", "votes": 4, "replies": [ { "id": 1893189, "postDate": "2022-08-10T15:52:07.467Z", "content": "This is exactly right. You can see more detail about just how much work is involved by looking at the length of the acknowledgements page. Medical datasets generally take a great deal of work to prepare.
", "rawMarkdown": "This is exactly right. You can see more detail about just how much work is involved by looking at the length of [the acknowledgements page](https://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/overview/acknowledgements). Medical datasets generally take a great deal of work to prepare.", "votes": 2 }, { "id": 1896608, "postDate": "2022-08-13T01:20:21.530Z", "content": "Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??
\nAlso I used \"DICOM\" file viewer to view scans for case \"1.2.826.0.1.3680043.10633\"… these files show blank image?? 1.dcm, 10.dcm, 100.dcm, 101.dcm, 144.dcm …
\nIf they are not blank, can you recommend DICOM file viewer, so that I can view these images.
", "rawMarkdown": "Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??\n\nAlso I used \"DICOM\" file viewer to view scans for case \"1.2.826.0.1.3680043.10633\"... these files show blank image?? 1.dcm, 10.dcm, 100.dcm, 101.dcm, 144.dcm ...\n\nIf they are not blank, can you recommend DICOM file viewer, so that I can view these images." }, { "id": 1898940, "postDate": "2022-08-15T00:25:08.703Z", "content": "Try mango
\nhttps://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/342311
Thanks, I can view the images now using Mango.
\nBoundingboxes and Segmentation data is provided only for subset of training set. Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??
", "rawMarkdown": "Thanks, I can view the images now using Mango.\n\nBoundingboxes and Segmentation data is provided only for subset of training set. Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??" }, { "id": 1903748, "postDate": "2022-08-17T16:13:00.027Z", "content": "That's up to you. You could use the bounding boxes to train for fractured verterbrae. The segmentations could be used for training a segmentation algorithm. Obviously, they can only be used for training, as you won't be provided those in the test set.
\nAs @jirkaborovec discusses in this post, you have a fraction of the data, but maybe useful.
\nhttps://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/344819
Why only subset of the data is provided? Why not provide the data for all training set?
\ntrain_bounding_boxes.csv Bounding boxes for a subset of the training set.\n\nsegmentations/ Pixel level annotations for a subset of the training set.\nYou have to keep in mind that segmentation of medical images is very time consuming, especially for 3D data.
\nAnd it must be done by a medical expert, so it's also potentially expensive.
I guess the host tried to give as much as possible but the entire dataset was just too much.
", "votes": 4, "replies": [ { "id": 1893189, "author_name": "Sohier Dane", "author_url": "", "post_date": "2022-08-10T15:52:07.467000", "content": "This is exactly right. You can see more detail about just how much work is involved by looking at the length of the acknowledgements page. Medical datasets generally take a great deal of work to prepare.
", "votes": 2, "replies": [] }, { "id": 1896608, "author_name": "Prashant", "author_url": "", "post_date": "2022-08-13T01:20:21.530000", "content": "Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??
\nAlso I used \"DICOM\" file viewer to view scans for case \"1.2.826.0.1.3680043.10633\"… these files show blank image?? 1.dcm, 10.dcm, 100.dcm, 101.dcm, 144.dcm …
\nIf they are not blank, can you recommend DICOM file viewer, so that I can view these images.
", "votes": 0, "replies": [] }, { "id": 1898940, "author_name": "SolverWorld", "author_url": "", "post_date": "2022-08-15T00:25:08.703000", "content": "Try mango
\nhttps://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/342311
Thanks, I can view the images now using Mango.
\nBoundingboxes and Segmentation data is provided only for subset of training set. Does this mean ML solution for this problem is not supposed to use the given \"train_bounding_boxes.csv\" and \"segmentations\"??
", "votes": 0, "replies": [] }, { "id": 1903748, "author_name": "SolverWorld", "author_url": "", "post_date": "2022-08-17T16:13:00.027000", "content": "That's up to you. You could use the bounding boxes to train for fractured verterbrae. The segmentations could be used for training a segmentation algorithm. Obviously, they can only be used for training, as you won't be provided those in the test set.
\nAs @jirkaborovec discusses in this post, you have a fraction of the data, but maybe useful.
\nhttps://www.kaggle.com/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/344819