Data Record

All CNeuroMod data are distributed as DataLad repositories hosted on GitHub under the courtois-neuromod organisation. The master meta-repository, cneuromod.all, tracks all datasets and their derivatives as git submodules following YODA principles. Full technical documentation is available at docs.cneuromod.ca. Data are released under a CC0 license.

Repository structure¶

cneuromod.all is a DataLad YODA meta-repository comprising 43 git submodules. Each experimental paradigm occupies a top-level folder (e.g., friends/, hcptrt/, anat/), which in turn contains independent submodules for each data component:

Most data files are git-annex symlinks and must be explicitly retrieved with datalad get. Cloning the meta-repository fetches only metadata (git history, file pointers); data files are downloaded on demand. Recursive installation of submodules should be avoided, as submodules re-expose their own sub-submodules at differing versions for provenance tracking.

For example, the friends dataset (participants watching the sitcom Friends) exposes friends/bids (raw data), friends/fmriprep and friends/mriqc (preprocessing and quality-control derivatives), and friends/physprep (processed physiological recordings). Dataset-specific submodules follow the same convention — e.g. shinobi/training for behavioural logs, or things/glm and things/glmsingle for first-level model estimates. A full description of what each dataset contains is given in Data Overview.

BIDS compliance¶

All functional and anatomical neuroimaging data are formatted according to the Brain Imaging Data Structure (BIDS) specification. Deviations from the core specification are:

• Multi-contrast anatomical sequences follow BEP001.

• Spinal cord images use the Body Part tag proposed in BEP025 (bp-cspine) to distinguish them from brain anatomical images acquired with the same contrasts.

Session indices (ses-001, ses-002, …) reflect the order of data acquisition; the number of runs, tasks, and their order within a session vary across participants. A small number of session indices are skipped where an entire session was discarded for scanning issues.

All images covering the face — collected as part of the longitudinal anatomical protocol (anat/bids) — were anonymised by zeroing the face, teeth, and ear regions with a custom mask warped from MNI space.

Preprocessing derivatives¶

fMRIPrep¶

Functional data were preprocessed using fMRIPrep 20.2.5 [MISSING REF: Esteban, O. et al. (2019). “fMRIPrep: a robust preprocessing pipeline for functional MRI.” Nature Methods, 16, 111–116. doi: 10.1038/s41592-018-0235-4 — add to references.bib] via an anatomical fast-track using sMRIPrep output (--anat-derivatives), ensuring a consistent anatomical basis across all functional datasets. The --ignore slicetiming flag was used. Outputs are provided in three spaces: native T1w, volumetric MNI152NLin2009cAsym, and surface-based fsLR-den-91k (grayordinates). Each functional run yields:

• *_desc-preproc_bold.nii.gz — preprocessed BOLD timeseries

• *_boldref.nii.gz — single-volume BOLD reference

• *_desc-brain_mask.nii.gz — brain mask

• *_desc-confounds_timeseries.tsv — confound regressors (motion parameters, CompCor components, global signals, framewise displacement, DVARS)

sMRIPrep / FreeSurfer¶

Anatomical processing used sMRIPrep and FreeSurfer in both cross-sectional and longitudinal modes, providing cortical surface reconstructions, segmentations, and subject-specific atlases. Cortical flat maps are available via PyCortex surfaces (anat/pycortex).

PhysPrep¶

Physiological signals were preprocessed using the CNeuroMod PhysPrep pipeline [MISSING REF: PhysPrep — github.com/courtois-neuromod/physprep — add citation to references.bib], integrating Phys2Bids, NeuroKit2, and Systole. Outputs per run include:

• *_physio.tsv.gz / *_physio.json — raw segmented biosignals (in <dataset>/bids)

• *_desc-preproc_physio.tsv.gz — filtered and cleaned timeseries

• *_desc-physio_events.tsv — sparse extracted features (peaks, troughs, SCRs)

• *_desc-quality.json — run-level quality assessment (pass/fail per modality)

Data access and versioning¶

Five of the six participants have consented to fully open sharing; access to the complete databank requires a registered-access data transfer agreement. cneuromod.all uses yearly release tags to pin a specific state of all submodule pointers, enabling exact reproduction of prior analyses. See Data Availability for deposit locations, license terms, access procedure, and download instructions.