I was wondering if anyone has any guidance on how to fully deidentify imaging data. I have a set of T1 structural images that I need to remove all participant information from- beyond the file names, when you open the images there is personal info in the corner that would prevent my colleague from analyzing them while being fully blinded.
Are there any scripts or programs anyone can recommend to accomplish this?
Would the following work for you?
3dNotes -HH “*” yourImageFile
Are you converting the data from DICOM to NIFTI?
If you’re still in DICOM space, you can use the anonymize function in mricron/dcm2nii (https://www.nitrc.org/projects/mricron) to anonymize the files. If you’re still seeing text on the images themselves, there are some solutions, but would be good to know if you’re converting from DICOM to NIFTI.
Hi, thank you for the reply- I tried the mricron/dcm2nii (https://www.nitrc.org/projects/mricron) and was able to anonymize the dicom files. I see that the program outputs the anonymized dicom files into the same folder as the identifiable ones. I’ll look into how to sort them out and have them saved into another folder.
To answer your question- I am working with dicoms and want to convert them to nifti, because ultimately, my colleague will be working with the nifti files. By anonymizing the dicom files before converting them into niftis, I am assuming this ensures the data is de-identified. Would you recommend this process to get my intended result? Thank you.
NIFTI files are almost inherently de-identified as most of the tags in the NIFTI files do not contain PII. You could try just converting them to NIFTI and then check the headers using 3dinfo[/url] or [url=https://afni.nimh.nih.gov/pub/dist/doc/program_help/nifti_tool.html]nifti_tool to check to make sure nothing slips in the NIFTI file that you’re worried about. If this works, you could skip the anonymize dicom step.
A growing number of sites are recommending that you remove the face on the structural T1 images. There are a variety of ways of doing this, but some of them cause issues for further processing (e.g. freesurfer). In an ideal world, I’d recommend using pydeface[/url] or a full skull strip ([url=https://afni.nimh.nih.gov/pub/dist/doc/program_help/3dSkullStrip.html]3dSkullStrip[/url] or [url=https://afni.nimh.nih.gov/pub/dist/doc/program_help/@SSwarper.html]@SSwarper).
If you’re feeling kind, it would also be nice to run the basic Freesurfer pipeline (recon-all -autorecon1) removing the T1.mgz and orig.mgz files and then sharing the rest of the Freesurfer output directory.
Hi Gang- thank you for the response. I tried using the 3Notes command but I wasn’t able to use it to anonymize the data. I am working with dicoms and plan on anonymizing them and then converting them to niftis. Would this command accomplish this?