Search for a command to run...
<ns3:p>Background Electroencephalography (EEG) measures brain activity. Different approaches to artifact identification (manual or automated) in previous EEG studies may affect reproducibility and consistency of results for scientific progress and biomarker research. We aimed to (a) compare the performance of the most common automated pipelines and (b) establish whether an automated pipeline can achieve comparable performance to a manual approach in a neurodivergent sample. Methods 131 participants with and without autism aged 6–31 years watched dynamic videos with varying social content while EEG was recorded. Spectral power and EEG functional connectivity were computed after pre-processing with a manual pipeline, the Maryland Analysis of Developmental EEG (MADE) pipeline, and the Harvard Automated Processing Pipeline for EEG (HAPPEv1), and adapted and updated versions of MADE and HAPPE on high- and low-density layouts (59 and 20 channels, resp.). We examined inclusion rates, consistency between the manual and other pipelines (intraclass correlations [ICCs]), and split-half reliability (for spectral power only, correlations – r’s). Results Inclusion rates were similar among high-density pipelines and between low-density HAPPILEE and miniMADE pipelines. Intraclass correlation coefficients between the manual pipeline and other pipelines showed excellent consistency, but were lower for manual vs. HAPPEv1. The ICCs between the high- and low-density versions of MADE were not significantly different from the ICCs between the density versions of HAPPE. Split-half reliability for power was high across all pipelines across all epochs, but lower for HAPPEv1 than for the other pipelines for condition differences. Conclusions The MADE and recent HAPPEv4 pipelines provide comparable data to a manual approach for samples with varying age, functioning level, and neurotype. This is encouraging, as it suggests that previous inconsistencies between results may arise from participant rather than from analytic heterogeneity. These findings will help improve data quality and provide guidance for biomarker research on neurodevelopmental conditions.</ns3:p>