Request for bad quality rodent MRI data

[matthew-brett]

From @ eilidhmacnicol April 7, 2022, 1:12pm

I am part of a team putting an educational session together for this
year's ISMRM regarding the importance of data quality. We are seeking
examples of bad quality data to highlight the importance of
standardised acquisition and data processing, particularly in small
animal imaging .

It is our intention to collate and distribute examples of structural
(T1w and T2w) and functional (bold) rodent MRI data with questionable
quality (i.e., that would be discarded at some point before the final
analysis) and make it available via OpenNeuro as an educational
resource.

If you are interested in contributing to this resource, you must:

• agree to make the data freely available under a CC0 license (as
  required by
  OpenNeuro.org)
• provide information about the ethical approvals required for the
  data acquisition (e.g., ethical board approvals, home office
  approval in the UK, etc.)
• provide the relevant information about funding and potential
  conflicts of interest

In return:

• you and anyone associated with the data acquisition can be listed as
  an author on the citeable OpenNeuro repository
• the rubbish data you thought would never go anywhere will see the
  light of day -- meaning that those hours of scanning were not in
  vain!
• you will be able to contribute to the (probable) "publication bias"
  in MRI data quality :)

Additionally, there may be some scope for turning this work (perhaps
with key outcomes from the educational symposium) into a manuscript,
provided there is enough interest.

Please contact me if you would like to contribute. We would appreciate
it if you can provide us with your examples as soon as possible, but
no later than 29th of April 2022 to be considered for the event.

[arokem]

This seems to be from a while back. Does anyone know whether this data was ever published?

[matthew-brett]

@eilidhmacnicol - just copying these across from the Nipy Discourse forum (that got shut down by Discourse). Did you publish these in the end?

[yarikoptic]

@TheChymera do you have any "bad" data?

[TheChymera]

@yarikoptic yes, I do.
@matthew-brett how “bad” do you want it? do you do some SNR metric to determine badness, or are you looking for severe artifacts?

[matthew-brett]

@TheChymera - not me - but I think @arokem was interested - is that right?

@eilidhmacnicol - are you still in the market for bad data?

[arokem]

Curious moreso than interested.

[eilidhmacnicol]

Hi all -

We made this call for a session in ISMRM '22 and some examples were put together in the animal QA/QC section of this online resource, but due to low* engagement with the call, they were never published as described in the original call.
*(To be fair, there was good interest in the initial call, but it seems digging up the data was a task that put folks off the commitment. All the images we ended up using were from our scanner at KCL.)

@matthew-brett to answer your question, I still think a collection of bad data would be a useful resource, but I don't currently have the capacity to take the lead. Maybe sometime...

how “bad” do you want it? do you do some SNR metric to determine badness, or are you looking for severe artifacts?

@TheChymera Since achievable SNR is relative to the acquisition, we were mostly looking for artifacts and problems that arise during image processing (abnormal anatomy causing registration problems, over-correction of motion in fixed-head rodents, brain masking that still includes the skull or takes the oblong rodent brain and makes it a sphere like a human, etc). I imagined that ideally we would have several examples of each with a continuum of severity with some more obvious than others since, in a real-world scenario, the presence of some artifacts will not hit you over the head.