A direct comparison of laboratory and community EEG recordings for neurodevelopmental research

Abstract

Leveraging portable electroencephalography (EEG) to measure brain function in community settings offers a promising strategy to improve the scalability and accessibility of developmental neuroscience research. To encourage broader adoption of these methods, it is important to demonstrate that data quality and neural signal integrity are comparable to gold-standard lab-based recordings. In this study, we directly compared EEG data collected in laboratory and home environments using portable EEG systems in a developmentally diverse group of young children under four years of age (n = 10). Despite differences in equipment and setting, our results showed comparable data quality and signal characteristics across conditions. Specifically, EEG data retention rates, noise levels, and spectral power measures were highly consistent at the group level, with no systematic differences between lab- and home-based recordings. To assess individual-level consistency, we calculated intraclass correlation coefficients (ICCs) for spectral power across brain regions and frequency bands. Most region-by-band combinations showed good to excellent consistency across settings; however, lower consistency was observed for some lower-frequency metrics, such as delta power in parietal regions. This suggests that certain individual features may be more sensitive to contextual or developmental factors. Overall, our findings demonstrate that portable, community-based EEG maintains data quality and neural signal integrity comparable to laboratory systems. Broader use of portable EEG may enhance scalability, increase participation, and promote greater inclusion in neurodevelopmental research.

Fig. 1

Boxplots display (A) recording lengths and (B) 60 Hz power estimates for each condition, with individual data points overlaid. (C) Power spectral density plotted illustrate the 60 Hz signal in each condition. Power spectra represent the average data from all 32 channels, prior to low-pass filtering and artifact removal. Shaded areas represent 95% confidence intervals. (D,E) Ridge plots depict the proportion of (D) seconds and (E) channels retained for each of the three cleaning algorithms used to assess data quality. Individual data points are overlaid to depict individual values for the primary cleaning method (ASR).

Fig. 2

PSD plots provide participant-level comparisons of spectral characteristics. Each subplot shows the power spectra for a single participant, with lab recordings in blue and community recordings in red. The power spectra are averaged across all 32 channels, with shaded regions representing confidence intervals calculated from the standard deviation of channel values.

Fig. 3

Boxplots illustrating power values across conditions. Panels (A–E) show power values for each spectral band (averaged across all brain regions), with individual participant data points overlaid. Panels (F–H) display power values for three specific regions (Frontal, Central, and Occipital), with power values averaged across all spectral bands, and individual participant data points overlaid for clarity.