Hyperscanning-BIDS?

[Kaare Mikkelsen]

Our group is considering how to transition our recordings to the BIDS structure. However, we are wondering what to do about 'hyperscanning' experiments, where multiple subjects are recorded simultaneously (in EEG)?

 The specification seems to be assume sessions inside subjects, but in a hyperscanning setup, sessions are at a higher level than subjects. What is the recommended way to deal with this?

I hope the question makes sense, and that it is ok to ask it in here.

Best

Kaare Mikkelsen

[Robert Oostenveld]

Hi Kaare,

Yes, this is a perfectly fine place to ask this. 

When extending BIDS to include EEG, we did discuss hyperscanning, and decided to not deal with it explicitly to keep the specification simple. But that does not mean it cannot be represented. If you search this google discussion group on “hyperscanning”, you can also find some previous messages on this.

I think there are two specific challenges to deal with:

1) the pairing of the subjects

2) the temporal synchronisation of the recordings

Re 1: if you have simple pairs where each subject is part of a pair only once, you can code those pairs in an additional column in the participants.tsv. E.g. sub-01 and sub-02 form pair 1, etc. You could additionally also use subject identifiers to code for the pairs: sub-01a and sub-01b forming pair 1, etc. In short: the pairing is represented in the metadata and/or the file names, not in the directory structure. The same would apply when recording data from twin-pairs, or sibblings, or parent-infant dyads, regardless whether it is instantaneous or not.

Re 2: subsequent analysis will be slightly easier if the data (and events) of the two subjects in each pair is aligned. You could trim the start of each recording such that data and triggers are aligned, and also cut off the end of the recording to make them exactly the same length. However, I think that most EEG researchers (who would potentially reanalyze the data) won’t require that and are also able to process the data pair-wise if the temporal synchronization were not so strict. And considering that one amplifier might run at 499.9Hz and the other at 500.1Hz, over a long recording you would not maintain sample-by-sample synchonization anyway. So I would not bother, but just make clear via the events.tsv of both subjects forming a pair which events are synchronized over the two recordings.

best regards

Robert

--
We are all colleagues working together to shape brain imaging for tomorrow, please be respectful, gracious, and patient with your fellow group members.
---
You received this message because you are subscribed to the Google Groups "bids-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bids-discussi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bids-discussion/b4ae9126-8bc2-42d2-b0d2-bf4b3cae18d7n%40googlegroups.com.

[Remi Gau]

Hey Kaare,

Can only give you a pointer for now.

There is an open discussion thread on the specification repo: https://github.com/bids-standard/bids-specification/issues/402

@+

Rémi

--
We are all colleagues working together to shape brain imaging for tomorrow, please be respectful, gracious, and patient with your fellow group members.
---
You received this message because you are subscribed to the Google Groups "bids-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bids-discussi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bids-discussion/b4ae9126-8bc2-42d2-b0d2-bf4b3cae18d7n%40googlegroups.com.

--

Rémi Gau

[Kaare Mikkelsen]

thank you both for the quick responses :) we were mainly worried about the grouping, but I think we can store that in the meta data. 

the github discussion is interesting reading :)

Best

Kaare