# Data — sources, scope, and terms

This repository redistributes **only derived, de-identified features** computed from two
public neuroimaging consortia. No raw imaging and no personally identifying information
is included.

## Sources

### ABIDE I (Autism Brain Imaging Data Exchange I)
- Resting-state fMRI + structural MRI, aggregated across ~20 international sites.
- We use the **Preprocessed Connectomes Project (PCP)** release: **C-PAC** pipeline,
  **CC200** functional atlas (200 ROIs).
- Cohort used: **937** strict-QC subjects (ASD 424 / Normal-control 513), mean
  framewise displacement ≤ 0.20. A relaxed-QC tier adds 165 more for the clinic view.
- Attribution: Di Martino et al., *Molecular Psychiatry* (2014); Craddock et al. (PCP).
- Terms: ABIDE data are released under **Creative Commons BY-NC-SA**; use is for
  non-commercial research. Honor the original ABIDE/PCP data-use terms.

### ADHD-200
- Resting-state fMRI used here only to populate the clinic view's extended cohort
  (**155** subjects), projected through the ABIDE strict reference.
- Amplitude-family markers (global/DMN ALFF) are **neutralized** for ADHD-200 because
  of preprocessing scale differences; they are not comparable across the two releases.
- Attribution: The ADHD-200 Consortium (2012). Honor the ADHD-200 data-use terms.

## What ships in this repo (sample)

| Artifact | Scope | Note |
|---|---|---|
| `clinic/clinic_data.js` | all **1257** subjects | single bundle, derived features only |
| `braintwin/data/braintwin_data.js` | meshes + reference + **16 curated** subjects | inlined |
| `braintwin/data/subjects/*.json` | **24** additional subjects | lazy-loaded sample |
| `braintwin/data/subject_index.js` | **40**-subject light index | trimmed from 1257 |

Not redistributed: the full 213 MB / 1257-subject per-subject library, the 4.2 GB of
raw ABIDE/ADHD timeseries, and the private internal model (AUROC 0.91) weights/data.

## Regenerating the full set

All artifacts are reproducible from the public releases with `pipeline/`. You must
obtain ABIDE I (PCP) and ADHD-200 yourself under their data-use terms. See
[docs/REPRODUCIBILITY.md](docs/REPRODUCIBILITY.md).

## De-identification

Subject identifiers are the public ABIDE/ADHD site IDs. No names, dates, free text, or
images are present. Only numeric derived features (functional connectivity summaries,
structural marker z-scores, graph metrics, normative-trajectory coefficients).