[Dataset] Add AOMIC-PIOP1 (ds002785) - Amsterdam Multimodal MRI (216 subjects)

Dataset Info

Field	Value
Name	AOMIC-PIOP1 (Population Imaging of Psychology 1)
Source	OpenNeuro ds002785
Paper	Snoek et al., Scientific Data 2021
License	CC0 (Public Domain)
Subjects	216
Format	BIDS
HuggingFace Target	hugging-science/aomic-piop1

Description

AOMIC-PIOP1 is part of the Amsterdam Open MRI Collection - multimodal 3T MRI with detailed demographics and psychometric variables. This is the smallest of the 3 AOMIC datasets, making it ideal as a "tracer bullet" before tackling the larger ones.

Data includes:

• T1-weighted structural MRI
• Diffusion-weighted MRI
• Resting-state fMRI
• Task-based fMRI (emotion, working memory, faces, etc.)
• Physiological recordings (cardiac, respiratory)
• Demographics + psychometrics

Why This Matters

1. Tracer bullet - Smallest AOMIC, test the pattern before ID1000 (928 subjects)
2. Rich multimodal - Structural + diffusion + fMRI
3. Task fMRI - Valuable for cognitive neuroscience
4. Proves pipeline - Tests Sequence(Nifti()) for multiple runs

Exact Schema

from datasets import Features, Value
from datasets.features import Nifti, Sequence

def get_aomic_piop1_features() -> Features:
    """AOMIC-PIOP1 schema - one row per SUBJECT.

    Note: Following arc.py pattern, we use a single `bold` list for ALL
    functional runs (rest + tasks). Task info is in the filename (e.g.,
    _task-rest_, _task-workingmemory_). Separation can be done downstream.
    """
    return Features({
        "subject_id": Value("string"),
        # Structural
        "t1w": Nifti(),
        # Diffusion
        "dwi": Sequence(Nifti()),           # May have multiple runs
        # Functional (all runs: rest + tasks)
        "bold": Sequence(Nifti()),          # All BOLD runs (*_bold.nii.gz)
        # Metadata from participants.tsv
        "age": Value("float32"),
        "sex": Value("string"),
        "handedness": Value("string"),      # Verify column name (hand vs handedness)
    })

Implementation Note: Verify participants.tsv column names. BIDS uses handedness but some datasets use hand. Check actual file during implementation.

Directory Structure

ds002785/
├── participants.tsv
├── dataset_description.json
└── sub-XXXX/
    ├── anat/
    │   └── sub-XXXX_T1w.nii.gz
    ├── dwi/
    │   ├── sub-XXXX_dwi.nii.gz
    │   ├── sub-XXXX_dwi.bval
    │   └── sub-XXXX_dwi.bvec
    └── func/
        ├── sub-XXXX_task-rest_bold.nii.gz
        ├── sub-XXXX_task-workingmemory_bold.nii.gz
        ├── sub-XXXX_task-faces_bold.nii.gz
        └── ... (multiple tasks)

Files to Create

src/bids_hub/datasets/aomic_piop1.py       # Dataset module
src/bids_hub/validation/aomic.py           # Shared AOMIC validation
scripts/download_aomic_piop1.sh            # Download script
tests/test_aomic_piop1.py                  # Tests
docs/dataset-cards/aomic-piop1.md          # Dataset card (follow arc-aphasia-bids.md pattern)

Implementation Steps

1. Create download script (scripts/download_aomic_piop1.sh)

   aws s3 sync --no-sign-request s3://openneuro.org/ds002785 "$TARGET_DIR"
   

2. Create dataset module (src/bids_hub/datasets/aomic_piop1.py)

   • Implement build_aomic_piop1_file_table() - walk sub-*/anat/, dwi/, func/
   • Implement get_aomic_piop1_features() - schema above
   • Use find_all_niftis() for Sequence features (multiple runs)

3. Add CLI commands (in cli.py)

   @aomic.command()
   def piop1():
       """AOMIC-PIOP1 dataset commands."""
   

4. Add validation (src/bids_hub/validation/aomic.py)

   • Check participants.tsv exists
   • Check T1w exists per subject
   • Report available modalities

5. Upload to HuggingFace

   uv run bids-hub aomic piop1 build /path/to/ds002785 \
       --hf-repo hugging-science/aomic-piop1 --no-dry-run
   

   • Use num_shards=216 (one per subject)

Acceptance Criteria

• Download script works
• uv run bids-hub aomic piop1 validate <path> passes
• uv run bids-hub aomic piop1 build <path> --dry-run succeeds
• Tests pass
• Dataset uploaded to hugging-science/aomic-piop1
• HuggingFace README.md with proper frontmatter, usage examples, and citation
• docs/dataset-cards/aomic-piop1.md added (follow arc-aphasia-bids.md pattern)

Resources

• OpenNeuro ds002785
• AOMIC Website
• Scientific Data Paper

Citation

@article{snoek2021aomic,
  title={The Amsterdam Open MRI Collection, a set of multimodal MRI datasets for individual difference analyses},
  author={Snoek, Lukas and van der Miesen, Maite M and others},
  journal={Scientific Data},
  volume={8},
  number={1},
  pages={85},
  year={2021}
}

---

Related Issues

After this is complete, tackle:

• AOMIC-ID1000 (ds003097) - 928 subjects
• AOMIC-PIOP2 (ds002790) - 226 subjects