Managing BIDS versions in an MR facility

Hi there!

(new to neurostars… reposting here a question asked to Chris in a different context, following his advice :wink: )

We’ve had our MR facility (a research dedicated one) delivering BIDS dataset to the users for more than a year now; however, we’ve realized when attempting to upload a dataset to OpenNeuro that a dataset that was BIDS-validated (i.e by the validator) a year ago (when it was acquired) does not pass the validation anymore. This raises the question of BIDS versions, and how fast they change etc., and we havn’t followed all this too much. Our goal (on the MR facility side) would be to stick with a “stable” version of BIDS so that they (at the MR facility) don’t have to change the way they deliver data to the users too often… So my question is a fairly general one: do you have any advice on what to do? Or… how do you handle this on your side? Any hint would be much helpful!



Hi Sylvain,

Thanks for riposting this here.

Let me clarify something. Since 1.0.0 we put a lot of effort not to introduce any backward incompatible changes - this was to avoid the situation where someone had a valid BIDS 1.x.x and it overnight became invalid.

What you are experiencing (most likely) are not incompatible changes to the standard, but the validator tool getting better at checking adherence to the standard. In other words the early version of the validator did not catch some of the issues with your dataset - however, the issues were always there - the standard did not change.

As the validator matures it will asymptote towards catching majority of issues so the situation you run into should happen less often in the future. I would, however, recommend periodically revisiting your data curation pipelines with new version of the validator to make sure everything is valid.