Hi,
\nI am trying to take the pixel value of the dicom files by using pixel_array, but I encounter various problems.
Thank you for your help.
", "messages": [ { "id": 1207208, "postDate": "2021-02-17T18:41:35.860Z", "content": "I don't think there is really an issue with no image in some .dicom. Here's a great notebook by GM raddar that I used for my own data pre-processing for reading the image data from the .dicom files.
\nYou could use PNG for loss-less compression.
", "rawMarkdown": "I don't think there is really an issue with no image in some .dicom. Here's a [great notebook](https://www.kaggle.com/raddar/convert-dicom-to-np-array-the-correct-way) by GM raddar that I used for [my own data pre-processing](https://www.kaggle.com/bjoernholzhauer/eda-dicom-reading-vinbigdata-chest-x-ray#7.-Creating-fast-to-read-shelve-file) for reading the image data from the .dicom files.\n\nYou could use PNG for loss-less compression.", "votes": 1, "replies": [ { "id": 1207259, "postDate": "2021-02-17T19:11:17.203Z", "content": "Thank you for your reply. I saw that notebook. But if there is no pixel_array attribute, you will have an error for calling dicom.pixel_array
\nAttributeError: 'FileMetaDataset' object has no attribute 'TransferSyntaxUID'
\nHow do you solve that if you didn't exclude them in the training set?
Do you have a file name where that's an issue. I don't think I ran out into such an issue, but it's possible I just missed it.
", "rawMarkdown": "Do you have a file name where that's an issue. I don't think I ran out into such an issue, but it's possible I just missed it." }, { "id": 1209005, "postDate": "2021-02-18T16:08:15.777Z", "content": "There are multiple files. This one (50a418190bc3fb1ef1633bf9678929b3.dicom in the train folder) is an example. I don't know if that is the problem with my download. Could you check it for me? I've been working on this for 1 week.
\nThank you for your reply.
Not sure what's going on for you there, perhaps it's indeed your download. If you create an notebook on the Kaggle site within the competition, you can test this:
\nimport pydicom\ntrain_dir = '../input/vinbigdata-chest-xray-abnormalities-detection/train'\nfilename = '50a418190bc3fb1ef1633bf9678929b3.dicom'\ndicom = pydicom.read_file(f'{train_dir}/{filename}')\ndicom.pixel_array\nand the result is
\narray([[16383, 16383, 16383, ..., 74, 66, 70],\n [16383, 16383, 16383, ..., 84, 80, 85],\n [16383, 16383, 16383, ..., 85, 78, 84],\n ...,\n [ 92, 84, 81, ..., 52, 56, 56],\n [ 80, 82, 72, ..., 54, 56, 59],\n [ 89, 91, 87, ..., 62, 64, 61]], dtype=uint16)\nfor me.
", "rawMarkdown": "Not sure what's going on for you there, perhaps it's indeed your download. If you create an notebook on the Kaggle site within the competition, you can test this:\n```\nimport pydicom\ntrain_dir = '../input/vinbigdata-chest-xray-abnormalities-detection/train'\nfilename = '50a418190bc3fb1ef1633bf9678929b3.dicom'\ndicom = pydicom.read_file(f'{train_dir}/{filename}')\ndicom.pixel_array\n```\nand the result is \n```\narray([[16383, 16383, 16383, ..., 74, 66, 70],\n [16383, 16383, 16383, ..., 84, 80, 85],\n [16383, 16383, 16383, ..., 85, 78, 84],\n ...,\n [ 92, 84, 81, ..., 52, 56, 56],\n [ 80, 82, 72, ..., 54, 56, 59],\n [ 89, 91, 87, ..., 62, 64, 61]], dtype=uint16)\n```\nfor me." }, { "id": 1215578, "postDate": "2021-02-23T18:45:41.880Z", "content": "Thank you so much. I almost gave up.
\nHave a nice day.
Hi,
\nI am trying to take the pixel value of the dicom files by using pixel_array, but I encounter various problems.
Thank you for your help.
", "rawMarkdown": "Hi,\nI am trying to take the pixel value of the dicom files by using pixel_array, but I encounter various problems.\n1. Some files don't have the pixel_array attribute => Is that the problem of the data set and we just skip those files?\n2. NotImplementedError: The pixel data with transfer syntax JPEG 2000 Image Compression (Lossless Only), cannot be read because Pillow lacks the JPEG 2000 plugin => I am not sure how to solve this problem\n\nThank you for your help.\n", "votes": 1 } ], "comments": [ { "id": 1207208, "author_name": "Björn", "author_url": "", "post_date": "2021-02-17T18:41:35.860000", "content": "I don't think there is really an issue with no image in some .dicom. Here's a great notebook by GM raddar that I used for my own data pre-processing for reading the image data from the .dicom files.
\nYou could use PNG for loss-less compression.
", "votes": 1, "replies": [ { "id": 1207259, "author_name": "Julia L", "author_url": "", "post_date": "2021-02-17T19:11:17.203000", "content": "Thank you for your reply. I saw that notebook. But if there is no pixel_array attribute, you will have an error for calling dicom.pixel_array
\nAttributeError: 'FileMetaDataset' object has no attribute 'TransferSyntaxUID'
\nHow do you solve that if you didn't exclude them in the training set?
Do you have a file name where that's an issue. I don't think I ran out into such an issue, but it's possible I just missed it.
", "votes": 0, "replies": [] }, { "id": 1209005, "author_name": "Julia L", "author_url": "", "post_date": "2021-02-18T16:08:15.777000", "content": "There are multiple files. This one (50a418190bc3fb1ef1633bf9678929b3.dicom in the train folder) is an example. I don't know if that is the problem with my download. Could you check it for me? I've been working on this for 1 week.
\nThank you for your reply.
Not sure what's going on for you there, perhaps it's indeed your download. If you create an notebook on the Kaggle site within the competition, you can test this:
\nimport pydicom\ntrain_dir = '../input/vinbigdata-chest-xray-abnormalities-detection/train'\nfilename = '50a418190bc3fb1ef1633bf9678929b3.dicom'\ndicom = pydicom.read_file(f'{train_dir}/{filename}')\ndicom.pixel_array\nand the result is
\narray([[16383, 16383, 16383, ..., 74, 66, 70],\n [16383, 16383, 16383, ..., 84, 80, 85],\n [16383, 16383, 16383, ..., 85, 78, 84],\n ...,\n [ 92, 84, 81, ..., 52, 56, 56],\n [ 80, 82, 72, ..., 54, 56, 59],\n [ 89, 91, 87, ..., 62, 64, 61]], dtype=uint16)\nfor me.
", "votes": 0, "replies": [] }, { "id": 1215578, "author_name": "Julia L", "author_url": "", "post_date": "2021-02-23T18:45:41.880000", "content": "Thank you so much. I almost gave up.
\nHave a nice day.