ds000001 example processed with SPM

[cmaumet]

This PR includes an example of BIDS-derivatives dataset: ds000001 processed using SPM (using code available in https://github.com/NISOx-BDI/Software_Comparison/tree/bids_derivatives).

This dataset will be used as an example use-case for BEP028 (provenance)

---

• [x] README
• [x] dataset_description.json

[Remi-Gau]

Couple of comments on the preprocessed data only as BIDS has nothing official to say on stats output (yet).

Also this is an example of derivative data, so I might be extra annoying to make it "extra BIDS compliant" because it could definitely help implementing validation later on. Feel free to push back on my nitpicking.

• Should the derivatives include the raw files? (IMHO: no)

I know that this is a very SPM thing to do but I feel this goes against the modularisation of "raw" and "derivatives" that BIDS tends to promote. Also this also leads to data duplication.

ds000001-spm/sub-01/anat
├── sub-01_from-T1w_to-IXI549Space_mode-image_xfm.nii.gz
├── sub-01_label-T1w_segparam.mat
├── sub-01_space-individual_desc-biascor_T1w.nii.gz
├── sub-01_space-individual_desc-skullstripped.gz
├── sub-01_space-individual_label-CSF_probseg.nii.gz
├── sub-01_space-individual_label-GM_probseg.nii.gz
├── sub-01_space-individual_label-WM_probseg.nii.gz
├── sub-01_space-IXI549Space_desc-preproc_T1w.nii.gz
└── sub-01_T1w.nii.gz                                                         <-- original raw file?

• Mix of .nii and .nii.gz

find ds000001-spm/ -name '*.nii' shows that subject 15 and 16 (and few others) have .nii files while most others have .nii.gz.

Might be better to make everything .nii.gz.

• missing sidecar JSON

Only highlighting the missing REQUIRED bits.

For preprocessed data, it should be mentioned if they are skullstripped or not https://bids-specification.readthedocs.io/en/latest/05-derivatives/03-imaging.html#preprocessed-coregistered-andor-resampled-volumes

When using the individual-space, the reference MUST me mentoned: https://bids-specification.readthedocs.io/en/latest/05-derivatives/02-common-data-types.html#examples_1

REQUIRED fields for the RAW should be propagated to the derivatives: https://bids-specification.readthedocs.io/en/latest/05-derivatives/01-introduction.html#metadata-conventions

In practice this mostly means TaskName and RepetitionTime for the func bold timeseries.

[Remi-Gau]

Also added some 'tick box' in the top message for other things TODO

[cmaumet]

Thanks a bunch @Remi-Gau!

The previous commits includes:

• Deletion of raw anat and func files
• All nifti files in .nii.gz
• Adding missing README + dataset_description.json

I'll work on the JSON files next!

[Remi-Gau]

I'll work on the JSON files next!

Suspect you can make your life easier by using the inheritance principle for many of the bold files.

[effigies]

Is it possible to submit a small subset of subjects? 2000+ files is hard on the browser, so I can't really review what's in here.

[Remi-Gau]

trying to come up with a decent .bidsignore to make things BIDS derivatives compatible if we removed the .SKIP_VALIDATION and had a validator that can handle derivatives.

distinguishing:

• valid BIDS derivatives files (currently described in the specs) that by definition are not in the .bidsignore
• BIDS derivatives files described in BEP in the bidsingore for now
• typical SPM flles whose name can be bidsiied without breaking SPM internal machinery
• typical SPM flles whose name CANNOT be bidsiied without breaking SPM internal machinery

Suggested .bidsignore

### SPM specific output 

## BIDSified filenames

# anat

# deformation fields not supported yet (see BEP 14 https://bids.neuroimaging.io/bep014)
*xfm.nii.gz

# segmentation parameters
sub-*/anat/sub-*_label-T1w_segparam.mat

# func

# realignment parameters
sub-*/func/sub-*_task-*_desc-confounds_regressors.txt

## non BIDSified names (renaming those would break SPM internal machinery)

# stats

SPM.mat
beta_*.nii*
RPV.nii*
ResMS.nii*
mask.nii*
con_*.nii*
spmT_*.nii*
spmF_*.nii*

*nidm.zip    # filename could be bidsified
*.pdf        # filename could be bidsified
*.ps         # filename could be bidsified

@cmaumet

I am having second thought on the Might be better to make *everything* .nii.gz.

For example zipping beta*.nii.gz will make SPM scream. From a data management perspective having .gz is preferable. As a SPM user having to unzipping things before doing stats is inconvenient.

What kind of examples should we set ?

[Remi-Gau]

@effigies would having this kind of .bidsignore help non-SPMers review this PR?

[Remi-Gau]

Random thought:

sub-01_task-balloonanalogrisktask_run-01_desc-confounds_regressors.txt

should probably be

sub-01_task-balloonanalogrisktask_run-01_motion.txt

according to the BEP on funcitonal derivatives: https://github.com/bids-standard/bids-specification/pull/519

ideally it should even be a TSV files with proper headers