How can I feed dicom Images to CNN
", "messages": [ { "id": "1338508", "postDate": "06/06/2021 14:01:24", "content": "How can I feed dicom Images to CNN
", "rawMarkdown": "How can I feed dicom Images to CNN", "votes": null }, { "id": "1338865", "postDate": "06/06/2021 19:41:05", "content": "After loading the DICOM file, you have the image as numpy array in ds.pixel_array. From here you can do anything you want - you can feed it to a CNN as it, resize it, or save it in another format.
", "rawMarkdown": "After loading the DICOM file, you have the image as numpy array in ds.pixel_array. From here you can do anything you want - you can feed it to a CNN as it, resize it, or save it in another format.", "votes": null }, { "id": "1338994", "postDate": "06/06/2021 23:52:45", "content": "Hi @nur988
\nno need to use the raw files/ Dataset (dicom Images)
\nsee here some Kagglers made jpg , png images from the dataset.
\nhttps://www.kaggle.com/h053473666/siimcovid19-512-img-png-600-study-png
\nhttps://www.kaggle.com/awsaf49/siimcovid19-1024-jpg-image-dataset
\nhttps://www.kaggle.com/awsaf49/siimcovid19-256-jpg-image-dataset
More here
\nhttps://www.kaggle.com/search?q=siimcovid19+in%3Adatasets
To work with dicom images, there is a python package called pydicom.
\nThe pydicom package allows us to extract all of the metadata from a set of DICOM files and store them in a way that is easy to access.
Read a DICOM file
\nmy_dicom = pydicom.dcmread(“MY_DICOM.dcm”)
\nAccess the actual pixel data of the image
\nmy_dicom_image = my_dicom.pixel_array
\nShow the Image
\nmatplotlib.pyplot.imshow(my_dicom_image, cmap = ‘gray’)
P.S. : data returned in my_dicom.pixel_array will be in the coordinate format [y,x].
", "rawMarkdown": "To work with dicom images, there is a python package called [pydicom](https://pypi.org/project/pydicom/).\nThe pydicom package allows us to extract all of the metadata from a set of DICOM files and store them in a way that is easy to access.\n\n**Read a DICOM file**\n`my_dicom = pydicom.dcmread(“MY_DICOM.dcm”)`\n**Access the actual pixel data of the image**\n`my_dicom_image = my_dicom.pixel_array `\n**Show the Image**\n`matplotlib.pyplot.imshow(my_dicom_image, cmap = ‘gray’)`\n\n**P.S.** : data returned in my_dicom.pixel_array will be in the coordinate format [y,x].", "votes": null }, { "id": "1340380", "postDate": "06/07/2021 20:36:48", "content": "I made a notebook a couple weeks ago that shows how to feed raw DICOM pixels into a CNN. In this case, I added a layer to InceptionV3.
\nhttps://www.kaggle.com/davidbroberts/dicom-full-range-pixels-as-cnn-input
", "rawMarkdown": "I made a notebook a couple weeks ago that shows how to feed raw DICOM pixels into a CNN. In this case, I added a layer to InceptionV3. \n\nhttps://www.kaggle.com/davidbroberts/dicom-full-range-pixels-as-cnn-input", "votes": null } ], "comments": [ { "id": 1338865, "author_name": "parzukrofila", "author_url": "", "post_date": "06/06/2021 19:41:05", "content": "After loading the DICOM file, you have the image as numpy array in ds.pixel_array. From here you can do anything you want - you can feed it to a CNN as it, resize it, or save it in another format.
", "votes": null, "replies": [] }, { "id": 1338994, "author_name": "faisalalsrheed", "author_url": "", "post_date": "06/06/2021 23:52:45", "content": "Hi @nur988
\nno need to use the raw files/ Dataset (dicom Images)
\nsee here some Kagglers made jpg , png images from the dataset.
\nhttps://www.kaggle.com/h053473666/siimcovid19-512-img-png-600-study-png
\nhttps://www.kaggle.com/awsaf49/siimcovid19-1024-jpg-image-dataset
\nhttps://www.kaggle.com/awsaf49/siimcovid19-256-jpg-image-dataset
More here
\nhttps://www.kaggle.com/search?q=siimcovid19+in%3Adatasets
To work with dicom images, there is a python package called pydicom.
\nThe pydicom package allows us to extract all of the metadata from a set of DICOM files and store them in a way that is easy to access.
Read a DICOM file
\nmy_dicom = pydicom.dcmread(“MY_DICOM.dcm”)
\nAccess the actual pixel data of the image
\nmy_dicom_image = my_dicom.pixel_array
\nShow the Image
\nmatplotlib.pyplot.imshow(my_dicom_image, cmap = ‘gray’)
P.S. : data returned in my_dicom.pixel_array will be in the coordinate format [y,x].
", "votes": null, "replies": [] }, { "id": 1340380, "author_name": "davidbroberts", "author_url": "", "post_date": "06/07/2021 20:36:48", "content": "I made a notebook a couple weeks ago that shows how to feed raw DICOM pixels into a CNN. In this case, I added a layer to InceptionV3.
\nhttps://www.kaggle.com/davidbroberts/dicom-full-range-pixels-as-cnn-input
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1338508": "How can I feed dicom Images to CNN", "1338865": "After loading the DICOM file, you have the image as numpy array in ds.pixel_array. From here you can do anything you want - you can feed it to a CNN as it, resize it, or save it in another format.", "1338994": "Hi @nur988 \n\nno need to use the raw files/ Dataset (dicom Images)\n\nsee here some Kagglers made jpg , png images from the dataset.\n\nhttps://www.kaggle.com/h053473666/siimcovid19-512-img-png-600-study-png\nhttps://www.kaggle.com/awsaf49/siimcovid19-1024-jpg-image-dataset\nhttps://www.kaggle.com/awsaf49/siimcovid19-256-jpg-image-dataset\n\nMore here\nhttps://www.kaggle.com/search?q=siimcovid19+in%3Adatasets", "1339106": "To work with dicom images, there is a python package called [pydicom](https://pypi.org/project/pydicom/).\nThe pydicom package allows us to extract all of the metadata from a set of DICOM files and store them in a way that is easy to access.\n\n**Read a DICOM file**\n`my_dicom = pydicom.dcmread(“MY_DICOM.dcm”)`\n**Access the actual pixel data of the image**\n`my_dicom_image = my_dicom.pixel_array `\n**Show the Image**\n`matplotlib.pyplot.imshow(my_dicom_image, cmap = ‘gray’)`\n\n**P.S.** : data returned in my_dicom.pixel_array will be in the coordinate format [y,x].", "1340380": "I made a notebook a couple weeks ago that shows how to feed raw DICOM pixels into a CNN. In this case, I added a layer to InceptionV3. \n\nhttps://www.kaggle.com/davidbroberts/dicom-full-range-pixels-as-cnn-input" }, "source": "meta" }