SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsbids-conversion

bids-conversion

Data & analytics

View repository

Updated 10-16 01:35

About this skill

BIDS Conversion

Convert neuroscience datasets to Brain Imaging Data Structure (BIDS) format. Supports EEG, EMG, MEG, fMRI, and behavioral data with proper file naming, JSON sidecars, and metadata.

When to Use

  • Converting raw data files to BIDS format
  • Creating or fixing BIDS metadata (JSON sidecars, TSV files)
  • Validating BIDS compliance
  • Setting up a new BIDS dataset from scratch
  • Converting between data formats (e.g., .set to .edf, .vhdr to .bdf)

BIDS Directory Structure

dataset/
  dataset_description.json
  participants.tsv
  participants.json
  README
  CHANGES
  sub-01/
    sub-01_scans.tsv
    eeg/
      sub-01_task-rest_eeg.set
      sub-01_task-rest_eeg.json
      sub-01_task-rest_channels.tsv
      sub-01_task-rest_electrodes.tsv
      sub-01_task-rest_coordsystem.json
      sub-01_task-rest_events.tsv
    emg/
      sub-01_task-grasp_emg.edf
      sub-01_task-grasp_emg.json
      sub-01_task-grasp_channels.tsv
      sub-01_task-grasp_events.tsv
    anat/
      sub-01_T1w.nii.gz
      sub-01_T1w.json
  derivatives/
    pipeline-name/
      sub-01/

File Naming Convention

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>]_<suffix>.<extension>
  • sub: subject identifier (required, alphanumeric, no special chars)
  • ses: session (optional, for longitudinal studies)
  • task: task name (required for functional data)
  • acq: acquisition parameters (optional)
  • run: run index (optional, for repeated acquisitions)
  • suffix: data type (eeg, emg, meg, bold, T1w, events, channels, electrodes)

Conversion Workflow

Step 1: Inventory Source Data

Identify:

  • Data format (BrainVision .vhdr, EEGLAB .set, EDF .edf, BDF .bdf, NIfTI .nii.gz)
  • Number of subjects and sessions
  • Task names and conditions
  • Channel types (EEG, EMG, EOG, ECG, misc)
  • Events/markers in the data
  • Coordinate system for electrode positions

Step 2: Create Dataset Scaffold

import json
from pathlib import Path

def create_bids_scaffold(root: str, subjects: list[str], tasks: list[str], modality: str = "eeg"):
    root = Path(root)
    root.mkdir(exist_ok=True)

    # dataset_description.json
    desc = {
        "Name": "Dataset Name",
        "BIDSVersion": "1.9.0",
        "DatasetType": "raw",
        "License": "CC0",
        "Authors": ["Last, First"],
        "DatasetDOI": "",
        "GeneratedBy": [{"Name": "Manual conversion"}]
    }
    (root / "dataset_description.json").write_text(json.dumps(desc, indent=2))

    # participants.tsv
    with open(root / "participants.tsv", "w") as f:
        f.write("participant_id\tage\tsex\thand\n")
        for sub in subjects:
            f.write(f"sub-{sub}\tn/a\tn/a\tn/a\n")

    # Create subject directories
    for sub in subjects:
        for task in tasks:
            (root / f"sub-{sub}" / modality).mkdir(parents=True, exist_ok=True)

Step 3: Convert Data Files

EEG (EEGLAB .set)

EEGLAB .set files are BIDS-compatible as-is. Copy and rename:

cp source.set sub-01/eeg/sub-01_task-rest_eeg.set
cp source.fdt sub-01/eeg/sub-01_task-rest_eeg.fdt  # if separate .fdt file

EEG (BrainVision .vhdr)

BrainVision files come in triplets (.vhdr, .vmrk, .eeg). All three must be renamed consistently:

cp source.vhdr sub-01/eeg/sub-01_task-rest_eeg.vhdr
cp source.vmrk sub-01/eeg/sub-01_task-rest_eeg.vmrk
cp source.eeg  sub-01/eeg/sub-01_task-rest_eeg.eeg

Update internal references in .vhdr and .vmrk to point to renamed files.

EEG (EDF/BDF)

Copy and rename:

cp source.edf sub-01/eeg/sub-01_task-rest_eeg.edf

EMG

EMG follows the same pattern but uses the emg directory and suffix:

cp source.edf sub-01/emg/sub-01_task-grasp_emg.edf

Step 4: Create JSON Sidecars

EEG sidecar (required fields)

{
  "TaskName": "rest",
  "TaskDescription": "Eyes-open resting state recording",
  "InstitutionName": "University Name",
  "InstitutionAddress": "Address",
  "Manufacturer": "BioSemi",
  "ManufacturersModelName": "ActiveTwo",
  "SamplingFrequency": 512,
  "EEGChannelCount": 64,
  "EOGChannelCount": 2,
  "EMGChannelCount": 0,
  "ECGChannelCount": 0,
  "MiscChannelCount": 0,
  "TriggerChannelCount": 1,
  "PowerLineFrequency": 60,
  "EEGPlacementScheme": "10-20",
  "EEGReference": "CMS/DRL",
  "EEGGround": "n/a",
  "SoftwareFilters": "n/a",
  "HardwareFilters": {"Highpass": {"HalfAmplitudeCutoffHz": 0.01}},
  "RecordingType": "continuous",
  "RecordingDuration": 300
}

EMG sidecar (required fields)

{
  "TaskName": "grasp",
  "TaskDescription": "Grasping task with force measurement",
  "SamplingFrequency": 2000,
  "EMGChannelCount": 8,
  "PowerLineFrequency": 60,
  "EMGPlacementScheme": "bipolar",
  "EMGReference": "differential",
  "Manufacturer": "Delsys",
  "ManufacturersModelName": "Trigno",
  "RecordingType": "continuous"
}

Step 5: Create TSV Files

channels.tsv

name	type	units	sampling_frequency	status	description
Fp1	EEG	uV	512	good	Frontal pole 1
Fp2	EEG	uV	512	good	Frontal pole 2
HEOG	EOG	uV	512	good	Horizontal EOG
VEOG	EOG	uV	512	good	Vertical EOG
EMG1	EMG	uV	2000	good	First dorsal interosseous

events.tsv

onset	duration	trial_type	value	sample
0.0	0.0	stimulus	1	0
1.5	0.5	response	2	768
3.0	0.0	stimulus	1	1536

electrodes.tsv (for EEG)

name	x	y	z
Fp1	-0.0294	0.0839	-0.0069
Fp2	0.0303	0.0835	-0.0083

coordsystem.json

{
  "EEGCoordinateSystem": "CapTrak",
  "EEGCoordinateUnits": "m",
  "EEGCoordinateSystemDescription": "Based on 10-20 system with 3D digitization"
}

Step 6: Validate

# Using the official BIDS validator (Node.js CLI)
bunx bids-validator /path/to/dataset

# Or install globally
bun install -g bids-validator
bids-validator /path/to/dataset

Modality-Specific Notes

Read the detailed references for modality-specific conversion guidance: