NEAR

NEAR identifies and mitigates bad channels and non-stereotypical artifacts in neonatal EEG recordings to improve signal quality for developmental neurophysiology analyses.

Key Features:

• Bad Channel Detection: Uses the Local Outlier Factor (LOF) algorithm to identify and isolate problematic channels in neonatal EEG recordings.
• Artifact Removal via ASR: Adapts Artifacts Subspace Reconstruction (ASR) with a parameter calibration procedure tailored to newborn EEG to remove non-stereotypical artifacts.

Scientific Applications:

• Validation Datasets: Tested on simulated data and real-world recordings from newborns and 9-month-old infants using frequency-tagging and event-related potentials (ERP) experimental designs.
• Performance: Demonstrated superior removal of representative newborn non-stereotypical artifacts and reproduced established EEG responses from noisy datasets with higher statistical significance compared to existing methods.

Methodology:

Applies a two-step computational workflow: Step 1 uses Local Outlier Factor (LOF) for bad channel detection; Step 2 applies a parameter-calibrated Artifacts Subspace Reconstruction (ASR) to remove non-stereotypical artifacts.