Hi All,
I would like to use de-identified medical dicom datasets for my segmentation process. Any recommendations in this space? Are there standard library available for achieving it?
Thanks,
Jiten
Hello @Jigabytes,
If you don’t need any of the DICOM tags, the simplest way is to save the volumes to another format (e.g. mha, nrrd, nifti).
If you need to keep the images in DICOM format, take a look at the RSNA Clinical trial processor (CTP) and possibly read the guide from The Cancer Imaging Archive which also discusses private tags.
2 Likes
Hi @zivy,
I will look into CTP. if I have DICOM series I read the series into one image and then save it as one of the file formats right ? And then I can directly read for example .mha as inage from itk reader. Have I understood correctly?
Thanks,
Jiten
Hi @Jigabytes, yes you understood correctly.
If you only need to de-identify DICOM files before processing them we have a de-identifier in our software, it works with a free license but requires windows. It’s quite strict but it keeps all UID connections etc.
Thanks @mattias for ur response. Yes I want to de-identify prior to segmentation processing. Currently I am trying out on Mac OS based system. I will keep u posted in case my other options does not workout. Apart from option to save as .mha I am also exploring Google Clouds Healthcare De-identification API. I am looking for something that does this preprocessing in over all workflow that I am trying to achieve. Thanks, Jiten
You can try DICOM Anonymizer Adage, easily available and for free on Adage - AI Medical