[ENH] Add generic metadata from BEP22 to MRI

[Remi-Gau]

Aim

Adds metadata to MRI and PET that either already exist in other modalities (BodyPart*) or are part of BEP22 (MRS).

See comment #1377 (comment)

This should:

• help make the MRS BEP is smaller by integrating a chunck of it early
• try to maximize metadata consistency between modalities

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 87.93%. Comparing base (83b12e9) to head (f3dd861).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1396   +/-   ##
=======================================
  Coverage   87.93%   87.93%           
=======================================
  Files          16       16           
  Lines        1351     1351           
=======================================
  Hits         1188     1188           
  Misses        163      163           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Remi-Gau]

I am not entirely sure if some of those have existing DICOM tags for MRI data.

BodyPart for PET mentions "DICOM Tag 0018, 0015 Body Part Examined", but I am not sure how it is for MRI data.

@neurolabusc if you have any hints on the DICOM fields for those metadata it'd be much appreciated.

[neurolabusc]

The DICOM Body Part Examined Attribute (0018,0015) is optional (Type 3). It is often included in MRI scanners with most vendors using BRAIN, but sometimes HEAD for Neuroimaging scans. If you need sourcedata/ for this, you could include the public data from dcm_qa_stc where GE and Siemens report (0018,0015) as [BRAIN] while UIH reports [HEAD].

[neurolabusc]

@Remi-Gau do you have source data exemplars for proposed tags like WaterSuppression, WaterSuppressionTechnique, B0ShimmingTechnique, B1ShimmingTechnique, and NumberReceiveCoilActiveElements? I do not think these have DICOM tags, so perhaps it is best to propose to the DICOM workgroup rather than BIDS.

[Remi-Gau]

@Remi-Gau do you have source data exemplars for proposed tags like WaterSuppression, WaterSuppressionTechnique, B0ShimmingTechnique, B1ShimmingTechnique, and NumberReceiveCoilActiveElements? I do not think these have DICOM tags, so perhaps it is best to propose to the DICOM workgroup rather than BIDS.

I actually do not.

Those are metadata being added to the BIDS MRS extension (see for example here:

bids-specification/src/schema/objects/metadata.yaml

Line 245 in c25573e

B0ShimmingTechnique:

)

I am just trying to also add them to the MRI datatypes to help have more internal consistency within BIDS.

Random question: would there be an issue in adding those to BIDS AND DICOM ?

[neurolabusc]

@Remi-Gau your question is timely. The DICOM Executive Committee is encouraging the collaboration between the BIDS steering committee and the DICOM Magnetic Resonance Work Group (WG-16). The Work Groups have representatives from all the manufacturers. While things take time to trickle through, it is an opportunity to create a virtuous loop between what users need and how different manufacturers refer to items.

[effigies]

@Remi-Gau How about we break out the sample components into a separate PR from the acquisition parameters? This will complement #1586.

[Remi-Gau]

@Remi-Gau How about we break out the sample components into a separate PR from the acquisition parameters? This will complement #1586.

oh yes!!! that would be nice

wanna open a PR or should I?

[effigies]

Go for it, or I can tomorrow.

[effigies]

Opened #1593.

[Remi-Gau]

removed sampled metadata as this is taken care of in #1593

[effigies]

Ah, I just rebased this on top of #1593. If you want it separate, just rebase 3664403 on master?

[effigies]

@markmikkelsen Trying to get a handle on this. Do we have example DICOMs that generate these additional terms? How do MRS converters determine them?

[markmikkelsen]

@effigies DICOM tag (0018,9025) fits these, but the value is not boolean as is proposed in #1377. The DICOM tag also covers OuterVolumeSuppression that the MRS BEP includes (which would be the FAT value in DICOM).

[effigies]

So do we want to add something like the following to WaterSuppression:

Corresponds to DICOM Tag 0018, 9025 SpectrallySelectedSuppression.
If present and not NONE, then this value SHOULD be True, otherwise False.

And the following to WaterSuppressionTechnique?

Corresponds to DICOM Tag 0018, 9025 SpectrallySelectedSuppression.
May be omitted if the value is NONE.

[markmikkelsen]

Makes sense to me! And the same would be the case for OuterVolumeSuppression.

[markmikkelsen]

I want to point out that most MR manufacturers don't even use the DICOM format to export/store MRS data. They use their own format. Hence why we designed the NIfTI-MRS format.

If MRS datasets are "DICOM" formatted, manufacturers alter them to their own specifications. So, I feel conforming MRS metadata in BIDS to DICOM is limiting what MRS users actually need for sharing data in BIDS format.

Is there a reason why BIDS should conform so much to DICOM?

[neurolabusc]

Perhaps @wtclarke can provide insight.

@markmikkelsen while BIDS is distinct from DICOM, it is useful to use the same terminology when it is already established in one standard. Furthermore, the BIDS steering group is actively collaborating with the relevant DICOM working groups (composed of engineers from the manufacturers) to establish common terminology where none exists. While this does slow down the process, the aim is for manufacturers to move from private DICOM tags or private manufacturer data formats to public DICOM tags. So while current MRS data may not use DICOM or public tags, the goal is in the future they will be able to. It makes sense to generate new terms when the existing ones are ambiguous or where manufacturers use competing terms. However, it makes sense to involve the manufacturers in the dialog to harmonize our terminology.

[markmikkelsen]

Thanks, @neurolabusc—that's very helpful to know! I'm glad the manufacturers are working together to better harmonize terminology. As you know, it's been a particularly challenging issue for MRS software developers.

[markmikkelsen]

Coming back to this from the MRS-BEP:

B0ShimmingTechnique
NumberReceiveCoilActiveElements

As far as I am aware, these have no equivalent DICOM tags but are very useful to the understanding of measurement of outcomes data outcomes of MRS users (e.g., see our multi-site paper)

B1ShimmingTechnique
NumberTransmitCoilActiveElements

These are relevant to parallel RF transmission in MRI and MRS, but they do not seem to have any DICOM tags.

[sappelhoff]

I am not a domain expert, but as far as I can tell this looks reasonable to me. Thanks!