brainjar 0.1.0

Reproducible processing pipelines and uniform loaders for brain imaging datasets.

brainjar

Reproducible processing pipelines and uniform loaders for brain imaging datasets.

Data access

brainjar ships code, not data. For most datasets you must obtain the raw data yourself under the dataset's own Data Use Agreement; redistribution is not permitted and process() cannot download anything — you point at your own copy:

process(download=False, raw_dir='/data/hcp_ya_open/raw')   # pipeline runs locally

The exceptions, where the processed derivative is openly redistributable and process(download=True) will fetch it from Zenodo:

• HCP-YA Open (HCP Consortium Open Access Data Use Terms)

Install

pip install brainjar

That gives you every dataset loader. To re-run a pipeline, install its extra into a dedicated venv:

python -m venv .venv && source .venv/bin/activate
pip install "brainjar[hcp_ya_open-pipeline]"

Pipeline extras install the exact pins recorded in each dataset's manifest.yaml. Different datasets may have conflicting pins — install one at a time.

Use

from brainjar.hcp_ya_open import process, get_df_image, get_df_xfeat, LABELS

process()                      # ensures the processed derivative exists
                               # (prompts: download the deposited derivative
                               # from Zenodo, or run the pipeline locally)
df_image = get_df_image()      # index: subject_id; cols: fa, md -> absolute Paths
df_xfeat = get_df_xfeat()      # index: subject_id; cols: age, sex,
                               # Release, ... (~580 columns from ConnectomeDB)

LABELS['age']                  # 'Age (years, 5-yr bucket)'
LABELS['fa']                   # 'Fractional Anisotropy'

process() is the entry point that gets data into place. Pass download=True / False to skip the prompt, or raw_dir=... to point at raw data when running locally.

Every dataset module exposes the same names: process, get_df_image, get_df_xfeat, LABELS.

Datasets

Access procedure, DUA, provenance, and pipeline extra for each are in the subpackage README:

• brainjar.hcp_ya_open
• brainjar.oasis3
• brainjar.camcan
• brainjar.hcp_ya_restricted
• brainjar.hcp_aging
• brainjar.hcp_development

Cache

Default: platformdirs.user_data_dir('brainjar') / <dataset>. Override per call (process(dest=...)) or globally (BRAINJAR_<DATASET>_PATH). A .complete sentinel marks a finished run.