EEG Release Data🔗

The HBCD EEG data and EEG preprocessing outputs do not contain any personally identifiable information. It is important to consider that potentially stigmatizing conclusions could emerge from the inappropriate use of the EEG data together with available demographic information or questionnaires. Furthermore, all EEG tasks are all passive at the V03 age range and therefore conclusions should not be drawn about behavioral performance.

Methodologically, there are a number of best practices for responsible data use that will be included with each file. The first is selecting files that maintain a minimum trial threshold recommendation. For each task, there are three levels of quality control thresholds that can be used: (1) our QC thresholds used to provide feedback to sites on each upload, (2) a 30% trial retention threshold (which we use internally to indicate usability of an EEG recording), and (3) the reliability recommendations for each task.

Threshold recommendations by task:

  • RS 108 trials
  • FACE 15 trials (for each condition of interest)
  • MMN 30 trials (for each condition of interest)
  • VEP 36 trials

An additional consideration for responsible use of the HBCD EEG dataset applies to disproportionate missing data. It is possible that some participant data may be systematically missing from this dataset by virtue of not meeting the QC thresholds. For instance, with infants that are inattentive and prone to fussing out during the EEG recording, more data may be removed from their datasets by our preprocessing scripts. A similar risk holds with participants who have thick or dense hairstyling and hair texture, which may impact capping success, impedance, and data quality (Adams et al., 2024). The consortium has proactively worked to address this risk by using scheduling procedures that are flexible to participants hairstyling routines and by purchasing 3 long pedestal nets per site in sizes appropriate for the V03, V04, and V06 visits (Adams et al., 2024; Mlandu et al., 2024). Preliminary analyses indicate that capping quality for visits where the long pedestal net was used have been consistent with capping quality seen for the dataset at large.

It is important to use these data in a manner which takes into account that physical and neurological differences between groups are not necessarily representative of intrinsic qualities of a given demographic group. Discussions around data patterns should be sensitive to societal factors. In addition, it is important to note that variation within demographic populations is greater than variation across populations. Demographic markers are categorical proxies that cannot capture or explain the causal mechanisms that may account for evident differences.

Expected Differences Between Raw BIDS and Derivatives

Raw BIDS data are released for all collected EEG sessions, regardless of processing eligibility. Therefore, differences in session counts between raw BIDS and HBCD-MADE derivatives are expected. In addition to exclusion based on quality control, ~2% of sessions had technical or acquisition issues that prevented complete processing. These data may be made available in future releases as issues are resolved.

Stimtracker Artifact

The MMN, VEP, and FACE task data may contain an artifact originating from the StimTracker device used for stimulus timing - see more information on this here.

MMN: Change in ISI Between Visits

The interstimulus interval (ISI) for the Auditory Mismatch Negativity task changed between visits V03 and V04/V06 - see Fox et al. 2024 and Morr et al. 2002 for details.

RS: Change in Video Content Between Visits

The video content for the Resting State task changed between visits V03 and V04/V06 - see Fox et al. 2024 and Morr et al. 2002 for details. Also note that RS is not a true resting state as there is a visual stimulus present.

EEG release data include the following:

Raw EEG BIDS🔗

BIDS conversion was performed with the EEG2BIDS Wizard, a custom MATLAB application for HBCD EEG data management and formatting, installed at all HBCD sites. After each EEG session, raw data are uploaded to the Wizard, which converts them to the BIDS standard.

Each participant’s BIDS eeg/ folder contains task-specific .set and .fdt EEG recordings, along with channel metadata (channels and events TSV files). Electrodes are placed on either the head or chest (acq-eeg/ecg) and electrode placement information is stored in electrodes TSV files accompanied by coordsystem JSON files that define the Cartesian coordinates. Finally, the sourcedata/ subfolder includes impedance measurements (impedances JSON) used to ensure good electrode contact and task eventlogs txt files describing stimulus presentation timing.

 How To Read File Trees →
hbcd/
└── rawdata/
    └── sub-[ID]/
        └── ses-[V0X]/
            └── eeg/
              # Task Acquisitions
                ├── *_task-{FACE|MMN|RS|VEP}_acq-{eeg|ecg}_run-[X]_channels.tsv
                ├── *_task-{FACE|MMN|RS|VEP}_acq-{eeg|ecg}_run-[X]_eeg.set (+JSON)
                ├── *_task-{FACE|MMN|RS|VEP}_acq-{eeg|ecg}_run-[X]_events.tsv (+JSON)
                ├── *_task-{FACE|MMN|RS|VEP}_acq-eeg_run-[X]_eeg.fdt
              # Electrode Placement
                ├── *_acq-eeg_space-{CapTrak|CTF}_electrodes.tsv
                ├── *_acq-eeg_space-{CapTrak|CTF}_coordsystem.json
                └── sourcedata/
                    ├── *_acq-eeg_flags.json
                    ├── *_acq-eeg_impedances.json
                    └── *_task-{FACE|MMN|RS|VEP}_acq-eeg_run-[X]_eventlogs.txt

HBCD-MADE Derivatives🔗

EEG data were processed using HBCD-MADE, an adaptation of the Maryland Analysis of Developmental EEG (MADE) pipeline (Debnath et al., 2020) developed specifically for the HBCD Study. HBCD-MADE is implemented as a containerized BIDS App that adheres to HBCD processing and derivative data standards. Full documentation of processing parameters, pipeline configuration, and file selection logic is available at the official HBCD-MADE documentation site and external HBCD Processing website.

 How To Read File Trees →
hbcd/
└── derivatives/
    └── made/
        └── sub-[ID]/
            └── ses-[V0X]/
                └── eeg/
                    ├── filtered_data/
                    │   └── *_task-{FACE|MMN|RS|VEP}_acq-eeg_run-[X]_desc-filtered_eeg{.fdt|.set}
                    │
                    ├── ica_data/
                    │   ├── *_adjustReport.txt
                    │   └── *_desc-mergedICA_eeg{.fdt|.set}
                    │
                    ├── merged_data/
                    │   └── *_desc-merged_eeg{.fdt|.set}
                    │   └── *_desc-merged_eeg.json
                    │
                    ├── processed_data/
                    │   ├── *.jpg # Topographic and ERP plots- see details
                    │   ├── *_task-RS_{Log|db|Abs}PowerSpectra.csv
                    │   ├── *_task-RS_spectra.mat
                    │   ├── *_task-{FACE|MMN|VEP}_ERPSummaryStats.csv
                    │   ├── *_task-{FACE|MMN|VEP}_ERPTrialMeasures.csv
                    │   ├── *_task-{FACE|MMN|VEP}_acq-eeg_run-[X]_ERP.mat
                    │   └── *_task-{FACE|MMN|RS|VEP}_acq-eeg_run-[X]_desc-filteredprocessed_eeg{.fdt|.set}
                    │
                    ├── *_acq-eeg_preprocessingReport.csv
                    └── *_task-{FACE|MMN|RS|VEP}_acq-eeg_run-[X]_MADEspecification.json
File Selection For Processing🔗

Not all raw EEG sessions are eligible for processing. Inclusion in the HBCD-MADE pipeline requires that sessions meet predefined quality control criteria. Common causes of exclusion include missing EEG capping images or very poor capping quality. For participants with multiple EEG acquisition attempts within a session, only the final run was processed and included in the MADE derivatives.

Adams, E. J., Scott, M. E., Amarante, M., Ramírez, C. A., Rowley, S. J., Noble, K. G., & Troller-Renfree, S. V. (2024). Fostering inclusion in EEG measures of pediatric brain activity. Npj Science of Learning, 9(1), 27. https://doi.org/10.1038/s41539-024-00240-y

Debnath, R., Buzzell, G. A., Morales, S., Bowers, M. E., Leach, S. C., & Fox, N. A. (2020). The Maryland analysis of developmental EEG (MADE) pipeline. Psychophysiology, 57(6), e13580. https://doi.org/10.1111/psyp.13580

Fox, N. A., Pérez-Edgar, K., Morales, S., Brito, N. H., Campbell, A. M., Cavanagh, J. F., Gabard-Durnam, L. J., Hudac, C. M., Key, A. P., Larson-Prior, L. J., Pedapati, E. V., Norton, E. S., Reetzke, R., Roberts, T. P., Rutter, T. M., Scott, L. S., Shuffrey, L. C., Antúnez, M., Boylan, M. R., … Yoder, L. (2024). The development and structure of the Healthy Brain and Child Development (HBCD) study EEG Protocol. Developmental Cognitive Neuroscience, 69, 101447. https://doi.org/10.1016/j.dcn.2024.101447

Mlandu, N., McCormick, S. A., Davel, L., Zieff, M. R., Bradford, L., Herr, D., Jacobs, C. A., Khumalo, A., Knipe, C., Madi, Z., Mazubane, T., Methola, B., Mhlakwaphalwa, T., Miles, M., Nabi, Z. G., Negota, R., Nkubungu, K., Pan, T., Samuels, R., … Gabard-Durnam, L. J. (2024). Evaluating a novel high-density EEG sensor net structure for improving inclusivity in infants with curly or tightly coiled hair. Developmental Cognitive Neuroscience, 67(101396), 101396. https://doi.org/10.1016/j.dcn.2024.101396