Hi everyone im new in Kaggle, Im wondering if there is a way optimal to convert the dicom images to png without losing important data, I was reading about crop but i would like some suggestion or tips to the preprocess part. If there is someone that want to help me i would be so grateful.
", "messages": [ { "id": 2116438, "postDate": "2023-01-26T13:47:13.743Z", "content": "Hi @alejandrolunamtz
\nAll the scans do not have the same pixel size or pixel interpretation so you will need to make a preprocessing pipeline that convert everything. Also they have a lot of useless data (background pixels) so cropping is very important. You can check out my own pipeline here, use it or get some inspiration to make your own steps.
Good luck!
", "rawMarkdown": "Hi @alejandrolunamtz\nAll the scans do not have the same pixel size or pixel interpretation so you will need to make a preprocessing pipeline that convert everything. Also they have a lot of useless data (background pixels) so cropping is very important. You can check out my own pipeline [here](https://www.kaggle.com/code/paulbacher/custom-preprocessor-rsna-breast-cancer), use it or get some inspiration to make your own steps.\n\nGood luck!", "votes": 3, "replies": [ { "id": 2116596, "postDate": "2023-01-26T15:37:09.597Z", "content": "Thank uuu, that helps me a lot.
", "rawMarkdown": "Thank uuu, that helps me a lot.", "votes": 1 } ] }, { "id": 2114293, "postDate": "2023-01-24T20:50:38.450Z", "content": "Hey,
\nI did not start yet, but I wanted to give you a fast answer (did not test it myself).
\nI found this, it might help you: https://www.tensorflow.org/io/tutorials/dicom
I will check it out, thanks for your time.
", "rawMarkdown": "I will check it out, thanks for your time." } ] }, { "id": 2114049, "postDate": "2023-01-24T17:18:03.483Z", "content": "Hi everyone im new in Kaggle, Im wondering if there is a way optimal to convert the dicom images to png without losing important data, I was reading about crop but i would like some suggestion or tips to the preprocess part. If there is someone that want to help me i would be so grateful.
", "rawMarkdown": "Hi everyone im new in Kaggle, Im wondering if there is a way optimal to convert the dicom images to png without losing important data, I was reading about crop but i would like some suggestion or tips to the preprocess part. If there is someone that want to help me i would be so grateful." } ], "comments": [ { "id": 2116438, "author_name": "Paul Bacher", "author_url": "", "post_date": "2023-01-26T13:47:13.743000", "content": "Hi @alejandrolunamtz
\nAll the scans do not have the same pixel size or pixel interpretation so you will need to make a preprocessing pipeline that convert everything. Also they have a lot of useless data (background pixels) so cropping is very important. You can check out my own pipeline here, use it or get some inspiration to make your own steps.
Good luck!
", "votes": 3, "replies": [ { "id": 2116596, "author_name": "Alex Luna", "author_url": "", "post_date": "2023-01-26T15:37:09.597000", "content": "Thank uuu, that helps me a lot.
", "votes": 1, "replies": [] } ] }, { "id": 2114293, "author_name": "P-W-T", "author_url": "", "post_date": "2023-01-24T20:50:38.450000", "content": "Hey,
\nI did not start yet, but I wanted to give you a fast answer (did not test it myself).
\nI found this, it might help you: https://www.tensorflow.org/io/tutorials/dicom
I will check it out, thanks for your time.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "2116438": "Hi @alejandrolunamtz\nAll the scans do not have the same pixel size or pixel interpretation so you will need to make a preprocessing pipeline that convert everything. Also they have a lot of useless data (background pixels) so cropping is very important. You can check out my own pipeline [here](https://www.kaggle.com/code/paulbacher/custom-preprocessor-rsna-breast-cancer), use it or get some inspiration to make your own steps.\n\nGood luck!", "2114293": "Hey, \n\nI did not start yet, but I wanted to give you a fast answer (did not test it myself). \nI found this, it might help you: https://www.tensorflow.org/io/tutorials/dicom", "2114049": "Hi everyone im new in Kaggle, Im wondering if there is a way optimal to convert the dicom images to png without losing important data, I was reading about crop but i would like some suggestion or tips to the preprocess part. If there is someone that want to help me i would be so grateful." } }