Best practice for deep learning-based image conversion using DICOM file #1841
Closed
renxiangdai
started this conversation in
General
Replies: 2 comments 1 reply
-
Converting to discussion... |
Beta Was this translation helpful? Give feedback.
0 replies
-
I hope someone will chime in here, but I don't think there are very many people watching new pydicom issues/discussions, and probably only a subset of those would be involved in deep learning. Perhaps stack overflow, or AI/ML/kaggle forums might be a better place to get help. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I have been exploring the Pydicom repository and I am quite impressed with its capabilities for handling DICOM data. I am currently working on a project that involves training a deep learning model using DICOM images. I would like to seek guidance on the best practices for utilizing DICOM data in open-source frameworks for deep learning, specifically within the Pydicom library.
My main concern is whether it is recommended to directly use the pixel_array attribute provided by Pydicom to feed the DICOM data into a deep learning model or to convert the DICOM data into image files (.jpg or .png etc) first before training the model. A research article suggest "Rescale the CT values to (−1, 1) for training. Minimum and maximum values were stored in every slice image for denormalization", is this the best solution?
For my specific use case, I am trying to train a denoising deep learning network. During test, the input will be a DICOM file, and the desire output should also be a DICOM file, but with clearer image and other part of DICOM file remain the same. Considering many open-source framework for image denoising is designed for image files (.jpg or .png etc), it's easier for setup if converting DICOM to image files first, but I'm worried about the information loss during conversion.
Could you kindly provide insights on the most appropriate method for handling DICOM data when training deep learning models using Pydicom? Any advice, examples, or best practices would be greatly appreciated.
Thank you for your time and support!
Beta Was this translation helpful? Give feedback.
All reactions