Tutorials generating BIDS datasets

[bclenet]

As discussed with @robertoostenveld during OHBM brainhack 2026, it would be nice to have tutorials like the Event related averaging but generating a BIDS dataset as output.

This would help in creating pertinent examples for BIDS-Prov (a.k.a. BEP028 that extends BIDS with provenance metadata) on how to describe provenance of time-series modalities (MEG, EEG, iEEG, NIRS) in BIDS.

The BIDS-Prov example would be based on these tutorials, and I can help testing these.

[robertoostenveld]

The 4 subjects' data used in that event-related averaging tutorial is shared (with a DOI) here https://doi.org/10.34973/21pa-dg13

The paper for which that data was shared was also about capturing computational details, allowing to rerun the analysis. The paper is here https://doi.org/10.21105/joss.05566. Example 2 is the of interest, and aligns with the event-related averaging tutorial mentioned above.

The example from the "reproducescript" paper is documented on https://www.fieldtriptoolbox.org/example/other/reproducescript_group/

The specific code used in example 2 is here https://github.com/matsvanes/reproducescript

[arnodelorme]

@robertoostenveld would you mind posting the processed BIDS derivative dataset so I can look how the provenance was done?

[robertoostenveld]

oh, it is not that far yet. The original processing was not with BIDS-formatted data, so I have to write a script to bidsify the data and another script for the processing.

[bclenet]

Thank you @robertoostenveld for the information,

I'm starting a BIDS-Prov example for the FieldTrip reproducescript example 2 then. For now, I'll focus on the activities, software and environments. As the file names are subject to changes after bidsification.

It would be interesting to be able to have provenance records at the same level as commands in the project/example2/reproduce_Subject01/script.m ; meaning 1 activity per analysis step (e.g.: ft_preprocessing, ft_rejectartifact, etc.).
Otherwise, two other possible granularity levels would be:

• only 1 activity representing the analysis.m script
• 5 activities: 1 for each doSingleSubjectAnalysis execution, and 1 for the doGroupAnalysis
  All three levels can be described in the example, as separate provenance graphs.

I'll let you know as I move forward.

[bclenet]

Hi @robertoostenveld

I started a bids example for the FieldTrip / reproducescript example 2 here: https://github.com/bclenet/bids-examples/tree/BEP028_fieldtrip/provenance_fieldtrip

Eventually, I described provenance with 3 different granularity levels:

• dataset level overview (files with entity prov-overview): the activity is the matlab command line launching analysis.m
• dataset level script overview (files with entity prov-scripts): the activites are one for each doSingleSubjectAnalysis and one for doGroupAnalysis
• file level (files with entities prov-subject01 and prov-group): the activites are functions inside each script (e.g.: ft_timelockanalysis, ft_combineplanar, ft_multiplotER, etc.)

For now the file names (hence the Ids and Labels) correspond to the ones in the FieldTrip reproducescript example 2, meaning they are not BIDS compliant. But It will be easy to switch to BIDS URI once we get there.

I also created a first version of a parser to convert the script.m files generated by reproducescript into BIDS-Prov metadata. I'll integrate it in the bids_prov repository.