vsoch
commented
3 years ago With this proposed new issue, I believe that I have an opportunity to build out our test image data set with a larger variety of PHI-safe test images from a variety of makes/models of scanners. Our department has an annual employee art show in which we can image (XR, CT, US) objects to serve as the basis for works of art.
I really like this idea. And actually, we can fairly easy separate the small example data (e.g., dicom cookies) from those that are just used for testing, which arguably shouldn't even be stored in the repository because they are so large. I think what would make sense to do is follow the convention of pydicom-data, and even first ask if there is room to accept files for deid here to that library (so pydicom can have a single, shared data library). @darcymason would this be something we could consider? If we want to keep the libraries separate, and you'd want to make a new dataset not just for deid but for other dicom enthusiasts then we could also have it branded differently and then used by the library here. Either way, we could definitely have a function to decompress after install.
wetzelj
Description
Related issues: #165
Modified pixel cleaning to accommodate various dimensions of stored pixels encountered when cleaning ultrasounds. Depending on the study, the pixel data array may be:
This commit includes a fix as well as tests for each of the above conditions.
Checklist
Open questions
The pixel data in the test images that I've uploaded with this PR have not been modified from the way they were sent from the modality. It was a struggle to find a RGB cine clip that was under the 100Mb file size limitation imposed by github. While the sample I provided was 99.6Mb (as reported by Win10), github detected this as >100Mb. To circumvent this, I added code to utils.py - get_files to detect and decompress files. It's not great, however, I'd like to open an issue to tackle test image storage as a future effort. It might be beneficial to compress everything for upload and decompress the images prior to use in the tests. With this proposed new issue, I believe that I have an opportunity to build out our test image data set with a larger variety of PHI-safe test images from a variety of makes/models of scanners. Our department has an annual employee art show in which we can image (XR, CT, US) objects to serve as the basis for works of art. Given that these images may not contain any body part and are tracked only to give back to the employee for use in creating their art work, they are not in any way tied to any patient or demographics. If you think this data set would be beneficial, I can start to work on the logistics and approvals for sharing this data. Currently the majority of the images are XR.