HBN-EEG (Healthy Brain Network)

Description

The Healthy Brain Network (HBN) is an ongoing, large-scale community biobank run by the Child Mind Institute (New York), launched to advance transdiagnostic research in pediatric mental health and learning disorders (Alexander et al., 2017). It targets 10,000 New York-area participants aged 5—21 recruited under a self-referred, community-referred model: rather than enrolling a pre-diagnosed clinical cohort or a screened-healthy cohort, HBN advertises to families who have concerns about their child, yielding a heterogeneous sample that spans the full range from typically developing to clinically affected. Each participant contributes deep phenotyping (psychiatric/behavioral assessment, K-SADS clinical interview, cognitive and achievement testing) plus multimodal data: high-density EEG, eye-tracking, resting and naturalistic-viewing fMRI, diffusion and morphometric MRI, voice/video recordings, genetics and actigraphy.

Citation

• Paper DOI: 10.1101/2024.10.03.615261
• Dataset DOI: 10.18112/openneuro.ds005505.v1.0.1
• Reference: Shirazi, S. Y., Franco, A., Scopel Hoffmann, M., Esper, N. B., Truong, D., Delorme, A., Milham, M., & Makeig, S. (2024). HBN-EEG: The FAIR implementation of the Healthy Brain Network (HBN) electroencephalography dataset. bioRxiv. Alexander, L. M., et al. (2017). An open resource for transdiagnostic research in pediatric mental health and learning disorders. Scientific Data 4, 170181.

Catalog citation block

Primary (EEG curation): Shirazi, S. Y., Franco, A., Scopel Hoffmann, M., Esper, N. B., Truong, D., Delorme, A., Milham, M., & Makeig, S. (2024). HBN-EEG: The FAIR implementation of the Healthy Brain Network (HBN) electroencephalography dataset. bioRxiv. https://doi.org/10.1101/2024.10.03.615261

Original biobank: Alexander, L. M., Escalera, J., Ai, L., et al. (2017). An open resource for transdiagnostic research in pediatric mental health and learning disorders. Scientific Data, 4, 170181. https://doi.org/10.1038/sdata.2017.181

Cite both, plus the relevant OpenNeuro release DOI (e.g. R1: doi:10.18112/openneuro.ds005505.v1.0.1).

Download and access

Field	Value
Official source	https://openneuro.org/datasets/ds005505
Access class	open
Size	TODO(confirm)
Format	eeglab_set
BIDS	true
Mirrors	OpenNeuro ds005505–ds005512, ds005514–ds005516 (R1–R11); s3://fcp-indi/data/Projects/HBN/BIDS_EEG/; NEMAR

Source & access

The EEG data are distributed as HBN-EEG, a BIDS-formatted curation released across 11 numbered releases (R1—R11) plus a separate “Not-for-Commercial-Use” (NC) release, hosted on OpenNeuro and mirrored on NEMAR. Each release is a distinct OpenNeuro accession; the underlying raw data live in the FCP/INDI S3 bucket (s3://fcp-indi/data/Projects/HBN/BIDS_EEG/cmi_bids_R{n}). Release-to-accession mapping (from the R1 README, ds005505 v1.0.1):

Release	OpenNeuro	S3 source	Subjects
R1	ds005505	cmi_bids_R1	136
R2	ds005506	cmi_bids_R2	152
R3	ds005507	cmi_bids_R3	183
R4	ds005508	cmi_bids_R4	324
R5	ds005509	cmi_bids_R5	330
R6	ds005510	cmi_bids_R6	134
R7	ds005511	cmi_bids_R7	381
R8	ds005512	cmi_bids_R8	257
R9	ds005514	cmi_bids_R9	295
R10	ds005515	cmi_bids_R10	533
R11	ds005516	cmi_bids_R11	430
NC	(not on OpenNeuro)	cmi_bids_NC	458

The R9 accession (ds005514) skips ds005513, and R6’s README link mistypes the id as “DS05510” — the canonical id is ds005510. The OpenNeuro DOI for R1 is doi:10.18112/openneuro.ds005505.v1.0.1; each release carries its own openneuro.dsXXXXXX DOI. License: all public releases are CC-BY-SA 4.0; the NC release alone is CC-BY-NC-SA 4.0 (research-only, no commercial use). Ethics: approved by the Chesapeake Institutional Review Board. The original biobank’s phenotypic data sit behind a Data Usage Agreement via COINS/LORIS at data.healthybrainnetwork.org, but the EEG/imaging are openly downloadable through FCP/INDI and OpenNeuro.

Data structure

BIDS (BIDSVersion 1.9.0), generated by bids-matlab-tools v9.1. EEG is stored as EEGLAB .set (with companion .fdt) files; the descriptor notes both SET and BDF can be supported. Each recording carries *_eeg.json (sidecar metadata), *_channels.tsv (129 rows: 128 net electrodes named E1—E128 plus Cz as the reference/129th channel), and *_events.tsv. Events — both native EEG triggers and the merged behavioral events — are annotated with HED (Hierarchical Event Descriptors) version 8.3.0, making the corpus suitable for systematic meta-/mega-analysis. OpenNeuro serves each release as a DataLad / git-annex repository (the GitHub mirrors contain a .datalad/ tree and git-annex pointers rather than the binary EEG itself). A downsampled 100 Hz “mini” variant is provided for lightweight use. Planned future additions named in the descriptor include personalized electrode locations, per-subject lead-field matrices, and eye-tracking data.

License

• Name: CC-BY-SA-4.0
• Note (verbatim from the directory): All public releases CC BY-SA 4.0; the NC release (458 subjects) is CC BY-NC-SA 4.0 and is not on OpenNeuro
• Snippets on this site: no — the site ships no snippet or precomputed product from this dataset; lessons link the official access page.

Acquisition

Field	Value
Device	128-channel EGI HydroCel Geodesic Sensor Net (+ Cz)
Device class	research-cap
Channels	129 (128 net electrodes named E1–E128 plus Cz (reference, channel 129))
Sampling rate (Hz)	500 (registry value: raw 500)
Online filters	0.1–100 Hz acquisition band-pass
Reference	Cz
Mains frequency (Hz)	60
Paradigms	rest-eo, rest-ec, surround-suppression, movie-watching, contrast-change-detection, sequence-learning, symbol-search
Participants	3155 (3,155 public across R1–R11; a separate NC release adds 458 (CC BY-NC-SA, not on OpenNeuro))
Population	3,155 children and adolescents 5–21 across releases R1–R11; transdiagnostic community sample (not screened-healthy)
Clinical groups	transdiagnostic
Sessions	TODO(confirm)
Durations	varies by task (RestingState ~403 s in a sampled subject); ~13 BIDS task/run files per fully-recorded subject

Acquisition

Recorded at Child Mind Institute sites in a sound-shielded room on a 128-channel EGI (Electrical Geodesics, Inc.) HydroCel Geodesic Sensor Net. EEG is digitized at 500 Hz with an online acquisition bandpass of 0.1—100 Hz and referenced to Cz (the vertex); the per-task eeg.json reports EEGChannelCount: 129, EEGReference: Cz, SamplingFrequency: 500, PowerLineFrequency: 60, and SoftwareFilters: "n/a". The net is sized to head circumference; electrode impedances are checked before recording and roughly every 30 minutes, kept below 40 kΩ. Power line is 60 Hz (USA). Recording length varies by task — e.g., RestingState is ~403 s in the sampled subject.

Participants

Children and adolescents aged 5—21 years. The sample is transdiagnostic by design and is not split into fixed “healthy vs. clinical” groups; most participants carry one or more DSM diagnoses, but enrollment does not require a diagnosis. Inclusion: ages 5—21, English fluency, and the capacity to provide assent/consent. Exclusion: serious neurological disease precluding participation, recent/acute psychosis, acute safety risk, IQ < 66, hearing or visual impairment that prevents task performance, substance dependence requiring chemical replacement, and acute intoxication (Alexander et al., 2017). Dimensional psychopathology is provided rather than categorical labels — see Phenotyping below.

Phenotyping

participants.tsv carries demographics (sex, age, handedness/ehq_total) plus per-task availability and quality-control flags from the minimal QC pass (file integrity, presence of required events, preprocessing-readiness), and commercial_use / full_pheno flags. Crucially, HBN-EEG ships dimensional CBCL-derived psychopathology scores — a general p-factor plus attention, internalizing, and externalizing dimensions — enabling brain—behavior analyses directly against the EEG without a separate phenotypic download.

Tasks / conditions

The HBN-EEG protocol comprises six paradigms (Shirazi et al., 2024), which expand to ~13 BIDS task/run files per fully-recorded subject. Three are passive and three are active:

• RestingState — participants rest with chin on a chin-rest, alternating eyes-open / eyes-closed while fixating a central cross.
• Surround Suppression (surroundSupp, 2 runs) — flashing peripheral disks on contrasting backgrounds; an inhibition/excitation probe.
• Movie Watching — four short films of differing theme: DespicableMe, DiaryOfAWimpyKid, FunwithFractals, and ThePresent.
• Contrast Change Detection (contrastChangeDetection, 3 runs) — perceptual decision-making with feedback on flickering disks.
• Sequence Learning (seqLearning6target / seqLearning8target, age-dependent) — memorize and reproduce flashed circle sequences.
• Symbol Search (symbolSearch) — a computerized WISC-style processing-speed task.

Behavioral responses (reaction times, accuracy) originally stored in the HBN behavior directory are folded into the BIDS events.tsv files alongside the EEG event markers.

Used on this site

L7.6, L7.8

Loader

Fetches only the stated subset. Recorded loader: DataLad (one OpenNeuro accession per release).

# One OpenNeuro accession per release (R1 = ds005505 … R11 = ds005516).
datalad install https://github.com/OpenNeuroDatasets/ds005505.git
cd ds005505
datalad get sub-*/eeg/*task-RestingState*   # resting runs of R1 only

Caveats worth knowing

• Channels are E1–E128 + Cz, not 10-20 names.
• Each release is a separate accession.
• The sample is transdiagnostic despite the name.

Notable caveats

• Channel labels are electrode indices E1—E128 plus Cz (the reference, channel 129), not standard 10—20 names; loaders that assume 10—20 montages must map via the EGI HydroCel coordinates.
• Heterogeneous, not screened-healthy: despite the “Healthy Brain Network” name, the sample is transdiagnostic — treat it as a community sample with high clinical loading, not a control cohort.
• The NC release (458 subjects) is CC-BY-NC-SA 4.0 and must be excluded from any commercial-use analysis; it is not hosted on OpenNeuro/NEMAR.

Related datasets

• ERP CORE (Compendium of Open Resources and Experiments)
• EEG Motor Movement/Imagery Dataset (EEGMMIDB)
• LEMON (MPI-Leipzig Mind-Brain-Body)
• Dortmund Vital Study resting EEG
• SRM Resting-state EEG

# Loader recorded in the catalog: DataLad (one OpenNeuro accession per release)
# TODO(confirm) the exact call and the subset to fetch. Official source: https://openneuro.org/datasets/ds005505