OpenNeuro and BIDS

Hello,
My group is about to upload a dataset to OpenNeuro. Per our IRB requirements and the OpenNeuro guidelines, we will be uploading the skull-stripped data used in our study. Technically, per BIDS structure, the skull-stripped data should go into a derivatives folder because the data in the root --> subject directory should be just the raw .nii files converted from Dicoms. Is this correct? It seems like most studies posted to OpenNeuro include the .nii files with the skull intact, so I want to confirm that we are preparing the right data and putting into into the appropriate bids format.
Thank you!
Yelena

Hi Yelena,

Strictly speaking, yes, BIDS does say minimal modifications should be done, but defacing is recognized as a reason to have a “derivative” inside the “raw” dataset, as data would not otherwise be shareable. I think most people who need to skull-strip their data to share it consider it a form of defacing. As a result, all datasets I’m aware of that only have skull-stripped T1w images available place them as if they were standard, T1w images.

This will also allow them to work best with most tools, which will expect to find images there.

That said, I can see organizational reasons to put your skull-stripped data in derivatives – for example, you have an original BIDS dataset, and now you’re constructing a shareable version that otherwise keeps everything in place – and if that makes the most sense with how you’ve organized your data up to this point, by all means, do that.

One approach to consider is to provide your skull-stripped data as a standard T1w image, and a binary mask in derivatives. For example:

bids-root/
    derivatives/
        skullstrip/
            sub-01/
                anat/
                    sub-01_desc-brain_mask.nii.gz
            ...
    sub-01/
        anat/
            sub-01_T1w.nii.gz
    ...

This would allow tools that would otherwise perform skull-stripping to query for a pre-defined mask, and just use that. We’re working on making this work in fMRIPrep, and it’s the best compromise we’ve been able to come up with that balances BIDS principles with historic usage. (We’re also happy to hear other suggestions.)

Best,
Chris

Thanks Chris! So, just be sure I understand - in your example above, the file called " sub-01_T1w.nii.gz" has the skull-stripped data in it, correct? We also skull-strip our bold data, so we would put that into the root directory as well?
Best,
Yelena

Hi, sorry for not noticing your response. Yes, that’s correct.