![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1968272%2Fc09a04a97164c5cfa3d94c89a5761ec2%2FScreenshot%202020-08-10%20at%2018.10.24.png?generation=1597079461193612&alt=media\")
Do metadata files created from DICOM images help? If yes how can they improve the parameters in general?
", "messages": [ { "id": "965479", "postDate": "08/10/2020 17:03:28", "content": "Do metadata files created from DICOM images help? If yes how can they improve the parameters in general?
", "rawMarkdown": "Do metadata files created from DICOM images help? If yes how can they improve the parameters in general?", "votes": null }, { "id": "965495", "postDate": "08/10/2020 17:12:58", "content": "
I'd be interested if anyone has got anything useful out of the metadata but also even if they had it wouldn't surprise me if distribution shift in the test data killed the benefit i.e. there might be correlation between scanner type and fvc by chance but if the distribution is different in test data then it'' hurt you.
\n\nThe only hypothesis I can come up with for any of these fields adding information about FVC is if the scanner position informed us about e.g. the subject's BMI which might be useful as per domain expert thread
", "rawMarkdown": "\n\nI'd be interested if anyone has got anything useful out of the metadata but also even if they had it wouldn't surprise me if distribution shift in the test data killed the benefit i.e. there might be correlation between scanner type and fvc by chance but if the distribution is different in test data then it'' hurt you.\n\nThe only hypothesis I can come up with for any of these fields adding information about FVC is if the scanner position informed us about e.g. the subject's BMI which might be useful as per domain expert thread", "votes": null }, { "id": "965971", "postDate": "08/11/2020 02:52:01", "content": "@jameschapman19 I did actually create one using some references but idk if it will be of any help 😅
", "rawMarkdown": "jameschapman19 I did actually create one using some references but idk if it will be of any help 😅", "votes": null }, { "id": "989916", "postDate": "08/29/2020 07:59:48", "content": "I am doing EDA on metadata extracted from the DICOM files. There are many discrepancies in the CT scans like number of slices, image size, aspect ratio of slices, etc. The meta-data, if nothing else can at least be useful in normalizing the CT scans.
\nI have already created the dataset, you can find it here. Any help in EDA will be appreciated.
", "rawMarkdown": "I am doing EDA on metadata extracted from the DICOM files. There are many discrepancies in the CT scans like number of slices, image size, aspect ratio of slices, etc. The meta-data, if nothing else can at least be useful in normalizing the CT scans. \n\nI have already created the dataset, you can find it [here](https://www.kaggle.com/ankursingh12/osic-metadata). Any help in EDA will be appreciated.", "votes": null } ], "comments": [ { "id": 965495, "author_name": "jameschapman19", "author_url": "", "post_date": "08/10/2020 17:12:58", "content": "
I'd be interested if anyone has got anything useful out of the metadata but also even if they had it wouldn't surprise me if distribution shift in the test data killed the benefit i.e. there might be correlation between scanner type and fvc by chance but if the distribution is different in test data then it'' hurt you.
\n\nThe only hypothesis I can come up with for any of these fields adding information about FVC is if the scanner position informed us about e.g. the subject's BMI which might be useful as per domain expert thread
", "votes": null, "replies": [ { "id": 965971, "author_name": "digvijayyadav", "author_url": "", "post_date": "08/11/2020 02:52:01", "content": "@jameschapman19 I did actually create one using some references but idk if it will be of any help 😅
", "votes": null, "replies": [] }, { "id": 989916, "author_name": "ankursingh12", "author_url": "", "post_date": "08/29/2020 07:59:48", "content": "I am doing EDA on metadata extracted from the DICOM files. There are many discrepancies in the CT scans like number of slices, image size, aspect ratio of slices, etc. The meta-data, if nothing else can at least be useful in normalizing the CT scans.
\nI have already created the dataset, you can find it here. Any help in EDA will be appreciated.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "965479": "Do metadata files created from DICOM images help? If yes how can they improve the parameters in general?", "965495": "\n\nI'd be interested if anyone has got anything useful out of the metadata but also even if they had it wouldn't surprise me if distribution shift in the test data killed the benefit i.e. there might be correlation between scanner type and fvc by chance but if the distribution is different in test data then it'' hurt you.\n\nThe only hypothesis I can come up with for any of these fields adding information about FVC is if the scanner position informed us about e.g. the subject's BMI which might be useful as per domain expert thread", "965971": "jameschapman19 I did actually create one using some references but idk if it will be of any help 😅", "989916": "I am doing EDA on metadata extracted from the DICOM files. There are many discrepancies in the CT scans like number of slices, image size, aspect ratio of slices, etc. The meta-data, if nothing else can at least be useful in normalizing the CT scans. \n\nI have already created the dataset, you can find it [here](https://www.kaggle.com/ankursingh12/osic-metadata). Any help in EDA will be appreciated." }, "source": "meta" }