Best BIDS option for multisite study

For an upcoming multisite study, primarily anatomical and diffusion data, we have been exploring options to efficiently convert dicoms to BIDS, with dcm2bids being the primary option due to ease of use.

I have a few specific concerns:

  1. What is an efficient way to deal extract subject IDs from dicom directory names?

We’ve been experimenting with data from two sites, and the naming conventions for each subject dicom folder is quite heterogeneous.
I am thinking this might be mitigated using a subject-key spreadsheet that matches subject IDs with dicom folders.

  1. The config file for dcm2bids is quite easy to configure. While I anticipate having to update this file for every new site, is there a way to configure it more efficiently? E.g. fitting multiple “SeriesDescription” entries under a single category?

There is also another tool we’re thinking of exploring, heudiconv. The setup is a little more “complicated”; however would it potentially provide more flexibility needed in a multisite study?

Finally, those of you who have experience using BIDS format for multisite studies, what strategies have helped you make the BIDS conversion workflow more efficient?

For an upcoming study, I would strongly recommend harmonizing data prior to data acquisition. Given the investment in terms of time and cost, I would ensure that you have more than one MRI physicist evaluate your sequences, as these are always tradeoffs and there are a lot of intricacies (I always get 2-3 offsite experts to evaluate our sequences, you can see my summary for DWI and fMRI).

I would recommend consistently and meaningfully naming the DICOM series at each site. I would use the format of reproin, regardless if you use HeuDiConv or dcm2bids.

Thank you for the response @Chris_Rorden .
Unfortunately, the scans have already been acquired, so we will not be able to manage the name or type of sequences we receive.

But I will certainly keep that in mind going forward :slight_smile:

I don’t know if it will do the job in your case but in BIDScoin I have implemented regular expressions to extract the subject (and session) label from the path. BIDScoin comes with a GUI and you can use your bidsmap for multiple studies, which should give you an easy and sufficient workflow