hcpep-skill

The skill generates a numbered execution plan covering ConnectomeDB download, BIDS staging, and sMRI, fMRI, and dMRI processing, then delegates each step to the appropriate modality skill after your explicit confirmation. Outputs are organized under an hcpep_output directory.

Run the reorganize_hcpep.py script with input and output directories to map HCP native layouts into BIDS structure. It normalizes subject IDs, routes T1w, T2w, task-fMRI, resting-state, and diffusion files, and generates dataset_description.json, with a dry-run mode for previewing.

The phenotype extraction script merges diagnosis, PANSS, BPRS, SAPS, SANS, GAF scores, medication and chlorpromazine equivalents, MATRICS cognitive domains, and illness duration measures. It merges multiple source files by subject ID and can filter to subjects with imaging data.

Yes, downloading HCP-EP data requires a ConnectomeDB account and acceptance of the HCP data use agreement. If raw data already exists locally, the download stage can be skipped and the workflow proceeds directly to BIDS staging.

The QC script flags subjects whose mean framewise displacement exceeds a default threshold of 0.3 mm, applied equally to patients and controls. It also pulls FreeSurfer eTIV from aseg.stats and can attach diagnosis labels to the QC summary.

The skill is orchestration-only and contains no preprocessing implementation; all modality processing is delegated to separate smri, fmri, and dwi skills. Clinical patient data also requires careful de-identification, and motion thresholds may need adjustment for early psychosis cohorts.