Multi-site EEG studies in early infancy: Methods to enhance data quality

Abstract

Brain differences linked to autism spectrum disorder (ASD) can manifest before observable symptoms. Studying these early neural precursors in larger and more diverse cohorts is crucial for advancing our understanding of developmental pathways and potentially facilitating earlier identification. EEG is an ideal tool for investigating early neural differences in ASD, given its scalability and high tolerability in infant populations. In this context, we integrated EEG into an existing multi-site MRI study of infants with a higher familial likelihood of developing ASD. This paper describes the comprehensive protocol established to collect longitudinal, high-density EEG data from infants across five sites as part of the Infant Brain Imaging Study (IBIS) Network and reports interim feasibility and data quality results. We evaluated feasibility by measuring the percentage of infants from whom we successfully collected each EEG paradigm. The quality of task-free data was assessed based on the duration of EEG recordings remaining after artifact removal. Preliminary analyses revealed low data loss, with average in-session loss rates at 4.16 % and quality control loss rates at 11.66 %. Overall, the task-free data retention rate, accounting for both in-session issues and quality control, was 84.16 %, with high consistency across sites. The insights gained from this preliminary analysis highlight key sources of data attrition and provide practical considerations to guide similar research endeavors.

Keywords: Autism; Early identification; Electrophysiology; Multi-site; Multimodal.

Fig. 1

Schematic diagram illustrating the structured three-phased approach used to integrate EEG data collection into IBIS.

Fig. 2

Schematics figures detailing A) The IBIS-EP protocol, with new additions (EEG and eye tracking) in red, and B) the paradigm-specific protocol for EEG recordings.

Fig. 3

A) Site-specific EEG system details B) Standardized laptop implemented across sites (1) Standardized IBIS Laptop; 2) Integrated photocell for monitoring visual stimuli; 3) Auxiliary connection transmits auditory timing from laptop to AV device; 4) AV Device relays stimulus timing to DAC; 5) Soundbar for auditory stimuli; 6) Guidelines for pre-session laptop checks).

Fig. 4

Paradigm-Specific EEG Completion Rates Across Sites and Time Points. (A) Completion rates for Task-Free, AEP, and VEP paradigms aggregated across all five sites at 6 and 12 months. (B-C) Detailed completion rates for each paradigm, broken down by individual site at the B) 6-month, and (C) 12-month timepoint.

Fig. 5

EEG Protocol Completion Rates Across Sites and Time Points. (A) Total protocol completion rates aggregated across all five sites at 6 and 12 months. (B) Protocol completion rates broken down by site for the 6-month time point. (C) Protocol completion rates broken down by site for the 12-month time point.

Fig. 6

A-B) Box plots showing the retained data duration (in seconds) at each site, with a total average plotted in gray for A) 6 and B) 12 months. C-D) Box plots showing the proportion of channels retained at each site, with a total average plotted in gray for C) 6 and D) 12 months.

Fig. 7

Power Spectral Densities (PSDs) for 6-month and 12-month EEG data, averaged across all scalp channels. Shaded regions represent 95 % confidence intervals. A) Overlaying PSDs from each site reveals high consistency in signal characteristics at both 6- and 12-month timepoints. B) PSDs show consistent age-related changes in signal characteristics between 6 and 12 months at each site, aligning with anticipated developmental trends.