I was looking into differencess between dicom and jpg images and I wonder: \n1) Are the image in dicom somehow different than the ones in jpg? (Excluding metadata information)\n2) How to transform existing dicom to the same color scale as jpg? \n3) What is the advantage of dicom over the jpg? \n4) What is the base for calculating the tfrecord: dicom or jpg?
", "messages": [ { "id": "870284", "postDate": "06/01/2020 15:53:28", "content": "I was looking into differencess between dicom and jpg images and I wonder: \n1) Are the image in dicom somehow different than the ones in jpg? (Excluding metadata information)\n2) How to transform existing dicom to the same color scale as jpg? \n3) What is the advantage of dicom over the jpg? \n4) What is the base for calculating the tfrecord: dicom or jpg?
", "rawMarkdown": "I was looking into differencess between dicom and jpg images and I wonder: \n1) Are the image in dicom somehow different than the ones in jpg? (Excluding metadata information)\n2) How to transform existing dicom to the same color scale as jpg? \n3) What is the advantage of dicom over the jpg? \n4) What is the base for calculating the tfrecord: dicom or jpg?", "votes": null }, { "id": "870291", "postDate": "06/01/2020 15:59:45", "content": "This interests me
", "rawMarkdown": "This interests me", "votes": null }, { "id": "885871", "postDate": "06/14/2020 14:29:54", "content": "I had similar questions and tried to answer them in my kernel here. Please have a look and let me know your thoughts. Thanks.
", "rawMarkdown": "I had similar questions and tried to answer them in my kernel [here](https://www.kaggle.com/supreethmanyam/dicom-vs-jpeg-vs-tfrecord-image-quality-analysis). Please have a look and let me know your thoughts. Thanks.", "votes": null }, { "id": "887282", "postDate": "06/15/2020 15:36:00", "content": "I was about to post link to your notebook here. Great Job!
", "rawMarkdown": "I was about to post link to your notebook here. Great Job!", "votes": null }, { "id": "887417", "postDate": "06/15/2020 17:00:29", "content": "I covered the Dicom files in quite a bit of detail in this notebook (including image transformations).
", "rawMarkdown": "I covered the Dicom files in quite a bit of detail in [this notebook](https://www.kaggle.com/jbfarrar/dicom-format-eda-and-image-analysis) (including image transformations).", "votes": null }, { "id": "892304", "postDate": "06/18/2020 19:36:46", "content": "Hi! Thanks for your questions. I can answer a few of them.
\n\nAll images in the competition were built from the same set of source JPEGs.
\n\nThe JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.
", "rawMarkdown": "Hi! Thanks for your questions. I can answer a few of them.\n\nAll images in the competition were built from the same set of source JPEGs.\n\n1. The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.\n3. DICOM provides additional metadata. For more complicated images (3D images, etc.) it's a great way to tie complex pixel arrays to a single image. If your pipeline deals with DICOM files directly it could be more easily useable in medical contexts.\n4. The TFRecords were based on the source JPEGs as well.\n\nThe JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.", "votes": null }, { "id": "915682", "postDate": "07/05/2020 02:45:33", "content": "This Discussion showed that saving the DICOM arrays at 75% quality yielded an array that matched the supplied JPGs. I don't understand how both could have been created from source JPGs, unless the supplied JPGs were created from source JPGs using a lower quality than that used to create the DICOM files. Can you elaborate?
", "rawMarkdown": "[This Discussion](https://www.kaggle.com/supreethmanyam/dicom-vs-jpeg-vs-tfrecord-image-quality-analysis) showed that saving the DICOM arrays at 75% quality yielded an array that matched the supplied JPGs. I don't understand how both could have been created from source JPGs, unless the supplied JPGs were created from source JPGs using a lower quality than that used to create the DICOM files. Can you elaborate?", "votes": null }, { "id": "915772", "postDate": "07/05/2020 05:01:39", "content": "I can't speak for Phil, but my guess is that the DICOM is the original source JPEG\n> The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.
\n\nAnd the competition JPEG is the source JPEG saved at 75% quality\n> The JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.
\n\nIt appears (from the analysis in the linked notebook) that when they wrote the new JPEG they wrote it at 75%
", "rawMarkdown": "I can't speak for Phil, but my guess is that the DICOM is the original source JPEG\n> The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.\n\nAnd the competition JPEG is the source JPEG saved at 75% quality\n> The JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.\n\nIt appears (from the analysis in the linked notebook) that when they wrote the new JPEG they wrote it at 75%", "votes": null }, { "id": "942250", "postDate": "07/23/2020 17:02:18", "content": "@philculliton do you know what transformation was applied to the input JPEG to get the TFRecords?
", "rawMarkdown": "philculliton do you know what transformation was applied to the input JPEG to get the TFRecords?", "votes": null } ], "comments": [ { "id": 870291, "author_name": "maximofn", "author_url": "", "post_date": "06/01/2020 15:59:45", "content": "This interests me
", "votes": null, "replies": [] }, { "id": 885871, "author_name": "supreethmanyam", "author_url": "", "post_date": "06/14/2020 14:29:54", "content": "I had similar questions and tried to answer them in my kernel here. Please have a look and let me know your thoughts. Thanks.
", "votes": null, "replies": [ { "id": 887282, "author_name": "janidziak", "author_url": "", "post_date": "06/15/2020 15:36:00", "content": "I was about to post link to your notebook here. Great Job!
", "votes": null, "replies": [] } ] }, { "id": 887417, "author_name": "jbfarrar", "author_url": "", "post_date": "06/15/2020 17:00:29", "content": "I covered the Dicom files in quite a bit of detail in this notebook (including image transformations).
", "votes": null, "replies": [] }, { "id": 892304, "author_name": "philculliton", "author_url": "", "post_date": "06/18/2020 19:36:46", "content": "Hi! Thanks for your questions. I can answer a few of them.
\n\nAll images in the competition were built from the same set of source JPEGs.
\n\nThe JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.
", "votes": null, "replies": [ { "id": 915682, "author_name": "ericzarahn", "author_url": "", "post_date": "07/05/2020 02:45:33", "content": "This Discussion showed that saving the DICOM arrays at 75% quality yielded an array that matched the supplied JPGs. I don't understand how both could have been created from source JPGs, unless the supplied JPGs were created from source JPGs using a lower quality than that used to create the DICOM files. Can you elaborate?
", "votes": null, "replies": [] }, { "id": 915772, "author_name": "cdeotte", "author_url": "", "post_date": "07/05/2020 05:01:39", "content": "I can't speak for Phil, but my guess is that the DICOM is the original source JPEG\n> The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.
\n\nAnd the competition JPEG is the source JPEG saved at 75% quality\n> The JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.
\n\nIt appears (from the analysis in the linked notebook) that when they wrote the new JPEG they wrote it at 75%
", "votes": null, "replies": [] }, { "id": 942250, "author_name": "janidziak", "author_url": "", "post_date": "07/23/2020 17:02:18", "content": "@philculliton do you know what transformation was applied to the input JPEG to get the TFRecords?
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "870284": "I was looking into differencess between dicom and jpg images and I wonder: \n1) Are the image in dicom somehow different than the ones in jpg? (Excluding metadata information)\n2) How to transform existing dicom to the same color scale as jpg? \n3) What is the advantage of dicom over the jpg? \n4) What is the base for calculating the tfrecord: dicom or jpg?", "870291": "This interests me", "885871": "I had similar questions and tried to answer them in my kernel [here](https://www.kaggle.com/supreethmanyam/dicom-vs-jpeg-vs-tfrecord-image-quality-analysis). Please have a look and let me know your thoughts. Thanks.", "887282": "I was about to post link to your notebook here. Great Job!", "887417": "I covered the Dicom files in quite a bit of detail in [this notebook](https://www.kaggle.com/jbfarrar/dicom-format-eda-and-image-analysis) (including image transformations).", "892304": "Hi! Thanks for your questions. I can answer a few of them.\n\nAll images in the competition were built from the same set of source JPEGs.\n\n1. The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.\n3. DICOM provides additional metadata. For more complicated images (3D images, etc.) it's a great way to tie complex pixel arrays to a single image. If your pipeline deals with DICOM files directly it could be more easily useable in medical contexts.\n4. The TFRecords were based on the source JPEGs as well.\n\nThe JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.", "915682": "[This Discussion](https://www.kaggle.com/supreethmanyam/dicom-vs-jpeg-vs-tfrecord-image-quality-analysis) showed that saving the DICOM arrays at 75% quality yielded an array that matched the supplied JPGs. I don't understand how both could have been created from source JPGs, unless the supplied JPGs were created from source JPGs using a lower quality than that used to create the DICOM files. Can you elaborate?", "915772": "I can't speak for Phil, but my guess is that the DICOM is the original source JPEG\n> The source JPEGs' pixel information was extracted and written into the Pixel Data of the DICOM files.\n\nAnd the competition JPEG is the source JPEG saved at 75% quality\n> The JPEGs provided are also based on the source JPEGs, but they've had their pixel information extracted and written into a new JPEG to remove any pre-existing metadata.\n\nIt appears (from the analysis in the linked notebook) that when they wrote the new JPEG they wrote it at 75%", "942250": "philculliton do you know what transformation was applied to the input JPEG to get the TFRecords?" }, "source": "meta" }