Skip to content

/neuroflow:data

Data intake — locate, inventory, validate, and convert your raw data.

/data takes raw recorded data and gets it ready for preprocessing. It covers locating and inventorying the data, checking BIDS compliance, and running conversion scripts if needed.

When to use it

  • You've finished data collection and need to organize raw files
  • You want to check if your data is BIDS-compliant
  • You need to convert from a proprietary format (BrainVision, EDF, etc.) to MNE/BIDS
  • You're starting fresh with an existing dataset

Three steps

Claude works through these in order:

Step 1 — Locate and inventory

Claude uses Glob and file inspection to understand your data:

  • How many subjects / sessions / runs
  • File formats present
  • Whether it looks BIDS-compliant already
Claude: I found 24 subject folders (sub-01 through sub-24), each with:
        - sub-XX_task-oddball_eeg.vhdr
        - sub-XX_task-oddball_eeg.vmrk
        - sub-XX_task-oddball_eeg.eeg

        No dataset_description.json or participants.tsv found.
        This is not yet BIDS-compliant.

        Saving inventory to .neuroflow/data/data-inventory.md

Output: data-inventory.md saved to .neuroflow/data/

Step 2 — Validate BIDS structure

Checks for BIDS compliance: - Naming conventions (sub-XX_task-YY_eeg.vhdr) - Required files (dataset_description.json, participants.tsv, events.tsv) - Sidecar JSON completeness

If data is not in BIDS format, Claude asks whether you want to convert or proceed as-is.

Step 3 — Convert

Runs or writes conversion scripts for common cases:

Source format Target format
BrainVision (.vhdr/.vmrk/.eeg) MNE Raw → FIF, or BIDS via MNE-BIDS
EDF / EDF+ MNE Raw → FIF
Raw folder structure BIDS layout
.set (EEGLAB) MNE Raw

Conversion scripts are saved in your project repository (not inside .neuroflow/).

Output: data-intake.md saved to .neuroflow/data/

Example session

/neuroflow:data

You: Data is in data/raw/ — BrainVision format, 24 subjects.

Claude: Found 24 subjects in data/raw/. BrainVision format (.vhdr/.vmrk/.eeg).
        No BIDS structure detected.

        Do you want to convert to BIDS format?
        1. Yes — convert with MNE-BIDS
        2. No — proceed as-is (I'll note the raw paths)

You: Yes — convert to BIDS.

Claude: I'll write a conversion script using mne_bids.write_raw_bids().
        The script will:
        - Create dataset_description.json with your study metadata
        - Create participants.tsv from your subject list
        - Convert each .vhdr file to BIDS naming convention
        - Add sidecar JSONs with EEG metadata

        [generates scripts/convert_to_bids.py]

Files read and written

Direction Files
Reads .neuroflow/project_config.md, .neuroflow/flow.md, .neuroflow/data/flow.md
Writes .neuroflow/data/, .neuroflow/data/flow.md, .neuroflow/sessions/YYYY-MM-DD.md, scripts/ (conversion code)