get skull pixel data

[NirutaDhimal]

I am trying to mask everything outside skull region in brain. Is there a way I specify skull in the recipe without having to give the actual coordinates?

[vsoch]

Are you using fMRI or nifti? If you are using dicom from brain imaging your best bet is to use more well established tools like FSL, SPM, or AFNI especially for masking. The tools here are primarily for other dicom like CT and ultrasound.

[vsoch]

Take a look at BET in FSL as a good start.

[NirutaDhimal]

Thank you. I am using dicom for brain imaging

[vsoch]

Gotcha! So I’d check out those software suites and the nipy suite of tools. Most convert dicom to nifti and go from there.

[NirutaDhimal]

Thanks. Is there a way to mask everything outside skull in deid module or to get the skull pixels or any pixels having certain value?

[vsoch]

That’s what a tool like BET is for.

[NirutaDhimal]

Got it. Thanks for the help

[NirutaDhimal]

Can I ask you something? I may sound stupid. What are we anonymizing while cleaning pixel data? Are we just removing the annotations?

[vsoch]

Yes exactly. For modality like CT and ultrasound patient personal information and metadata is often burned into the images, and the machines do it in predictable locations so we can erase it. That’s also more typical in a clinical context so the images have patient information. At least for brain imaging studies, if there are any burnt pixels (it’s been a long time but I don’t remember the post processed nifti having them) they would be removed in the analysis pipeline in an early step when you generate a brain mask.

If you are new to brain imaging the community is hugely active and vibrant and there are a ton of tools and methods for learning! Generally pydicom and deid are more catered toward clinical CT and ultrasound (and similar) that remain in dicom space and the communities overlap but I think it’s a bit smaller than neuroimaging primarily.

[NirutaDhimal]

Thank you so much. It's really helpful and I am new to medical images.

[vsoch]

Sure thing! Good luck and have fun. The tools are very good - most have both command line and GUI interfaces and pipelines (check out nipype for example) so you can script or click through multiple pipelines very easily. My biggest advice is to make sure to look at your data every step of the way, and (at least when I was learning) I liked to try something in a GUI first and then reproduce on the command line to run automated on many more subjects, usually on HPC.

[vsoch]

Going to close this @NirutaDhimal - thanks for stopping by!