Re: [bids-discussion] BEP for allow phantom data into BIDS?

23 views
Skip to first unread message

Cyril Pernet

unread,
Aug 27, 2021, 1:00:25 AM8/27/21
to bids-di...@googlegroups.com, niQC
There are some requirements from a metadata perspective
- somehow indicate this is phantom data (the species tag won't really work?)
- how to relate a phantom dataset to a bunch of other datasets the phantom data can be used with
- date wise I guess the phantom data date can be shared (even for ABCD?) Isn't the issue about people scan dates and how to relate the phantom dates to people scans.

Cyril


On Thu, 26 Aug 2021, 20:08 Open MINDS Lab at Pitt, PI Pradeep Raamana, <openmi...@gmail.com> wrote:
Hi Everyone,

based on some recent discussions in the niQC SIG regarding phantom data in QA and beyond (see the video below), we feel there might need to a be a BEP to include phantom data directly into BIDS

One main reason is to get around the limitation of not being able to share dates to help match between subjects and the corresponding phantom data (e.g. from the day of subject's scanning). It appears to me that packaging them together while sharing, if possible, would get around that limitation while retaining the correspondence. Does that make sense?

If some or all of this functionality exists already, let us know. Or if we need to consider a new BEP to achieve it?

PS: we have an exciting meeting today in an hour (tuesday 11am EDT) focused on phantoms: tinyurl.com/QualityConvs (and other exciting talks in the future).. join us if you are interested!

Thanks,
Pradeep


--
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/22f076e4-e636-47de-bbca-e16bc22b92a6n%40googlegroups.com.

Robert Oostenveld

unread,
Aug 27, 2021, 4:05:52 AM8/27/21
to dhruv.sharma via bids-discussion
Hi Pradeep,

Although not exactly "phantom data", for MEG the specification includes instructions how to organize "empty room” recordings. See https://bids-specification.readthedocs.io/en/stable/04-modality-specific-files/02-magnetoencephalography.html#empty-room-meg-recordings. These are organized by sessions corresponding to dates, so that it is possible to find an empty room recording (which are often done once every morning) that matches an experimental recording with a subject later that day. This has similarities to organizing https://bids-specification.readthedocs.io/en/stable/06-longitudinal-and-multi-site-studies.html. If you would have different phantoms (e.g. bottle, belong, cauliflower), I would code their identity as the subject identifier.

Hope this helps
Robert

Remi Gau

unread,
Aug 27, 2021, 4:37:23 AM8/27/21
to bids-di...@googlegroups.com, Robert Oostenveld

Was trying to think of where this could fit in a BIDS dataset but Robert's suggestions seems a good starting point.

Not sure this requires a full fledged BEP if a pull-request will do, but this may be because I am not seeing the specifics issues.

If dates in the scans.tsv or sessions.tsv have to "anonymised" then I suspect they should be offsetted by the same amount for phantom and the subject data.

 @+

Dr Cyril, Pernet

unread,
Aug 27, 2021, 5:05:05 AM8/27/21
to bids-di...@googlegroups.com, Remi Gau, Robert Oostenveld

The issue is that one phantom BIDS dataset can fit many human/animal BIDS dataset, so we don't want to store many times the same data inside other datserts with specific subjects - while Robert suggestion works that is too cumbersome - also fMRI phantom is not on a daily basis (mote like once a week).

We would therefore rather create BIDS phantom datasets and human datasets can refer to it and conversely (while the later might not be needed, depends if we want to keep bidirectionality) - using interpolation of data (say a sine overtime) QC from phantom can be regressed out on group analyses.


To view this discussion on the web visit https://groups.google.com/d/msgid/bids-discussion/153f7118-e961-b5b5-6190-fa11007e98ce%40gmail.com.
-- 
Dr Cyril Pernet, PhD, OHBM fellow, SSI fellow
Neurobiology Research Unit, 
Building 8057, Blegdamsvej 9
Copenhagen University Hospital, Rigshospitalet
DK-2100 Copenhagen, Denmark

wamc...@gmail.com
https://cpernet.github.io/
https://orcid.org/0000-0003-4010-4632

Remi Gau

unread,
Aug 27, 2021, 5:15:39 AM8/27/21
to Dr Cyril, Pernet, bids-di...@googlegroups.com, Robert Oostenveld

Ha I see.

If I am understanding correctly, my gut feeling is that there is a PR in the pipeline that might help formalize reference to BIDS dataset outside of the current dataset: https://github.com/bids-standard/bids-specification/pull/820

But I can see that there will be an issue if you want to:

  • keep the acquisition date of the phantom "as is" in a standalone dataset,
  • keep track when each scan was acquired respective to the phantom,
  • AND anonymize the acquisition dates
--

Rémi Gau

Dr Cyril, Pernet

unread,
Aug 27, 2021, 5:44:57 AM8/27/21
to Remi Gau, bids-di...@googlegroups.com, Robert Oostenveld, niQC, Raamana, Pradeep Reddy

yes I think we should try to build a repo as for humans and use the updated common principles from that PR (to be merged?) -- and see how it goes from there

Open Minds Lab

unread,
Sep 13, 2021, 1:31:36 PM9/13/21
to bids-discussion
Thanks everyone - I agree with the suggestions and comments so far (e.g. building on what'd been developed for and learnt from empty room recordings etc).

The PR linked to by Remi #820 may form part of the answer, but some related aspects are
  • defining different types of Datasets (human, phantom, empty room MEG etc)
  • modify or develop associated validations (e.g. in pybids)
perhaps a good start might be try build an example dataset with both human and phantom data available on openneuro and datalad?

Pradeep




Open Minds Lab

unread,
Sep 15, 2021, 9:40:29 PM9/15/21
to bids-discussion
perhaps that PR already addresses the validation part, I didn't review it in detail.

Paola Di Maio

unread,
Sep 16, 2021, 1:33:27 AM9/16/21
to bids-di...@googlegroups.com
Greetings
can you perhaps provide a reference and definition for phantom data
and the other concepts in the proposed approach
is it used as a non-null db?
I  am interested in relation to data/model validation 
thanks
PDM

Dr Cyril, Pernet

unread,
Sep 16, 2021, 2:33:20 AM9/16/21
to bids-di...@googlegroups.com, Open Minds Lab

Hi Pradeep,

Empty room MEG can be acquired for each subject, while we record an MRI phantom once a week and those data should be associated with many subjects across multiple studies - very different setting. While as an exercise I agree we can try to prepare something like an empty room recording, I do not agree with the concept since it won't generalize to want we want to achieve with the MRI phantom.

Cyril

Open Minds Lab

unread,
Sep 16, 2021, 9:13:14 AM9/16/21
to Dr Cyril, Pernet, bids-di...@googlegroups.com
I didn't mean to say the existing solution for the MEG empty room recordings would work as is, but a good initial draft to start off with for our needs. I agree with you (Cyril) that we need to hash out and implement the 1 to Many mappings needed well. I think we should start by making an example/representative dataset, and figure out what extensions to BIDS are needed (if any) to realize them. 

Paola, great questions! On the "reference and definition for phantom data", I think of the phantom data as regular medical images lacking any internal activity or structure (except some noise) or PHI/identifiable information. Perhaps there is a clear definition elsewhere (i need to look it up)

Thanks,
Pradeep

Paola Di Maio

unread,
Sep 16, 2021, 10:00:20 PM9/16/21
to bids-di...@googlegroups.com, Dr Cyril, Pernet
Thank Pradeep
it looks like you guys are having a really interesting conversation about phantom data
so I d like to add it to the glossary of bioinformatics terminology, I wait for a definition from this group
with thanks

he phantom data as regular medical images lacking any internal activity or structure (except some noise) or PHI/identifiable information. [#uh?]



Virus-free. www.avast.com


Virus-free. www.avast.com
Reply all
Reply to author
Forward
0 new messages