Happy to present our novel pipeline for Newborns’ EEG Artifacts Removal (NEAR) from the joint work of E3DA (FBK) and CIMEC (UNITN).
Studying newborns in the first days of life is crucial to understanding the neurocognitive predispositions of humans. A primary hurdle in such studies is the presence of artifacts due to uncontrolled movements. We propose an automated EEGLAB-based pipeline: NEAR. NEAR is both for single-subject and batch processing. NEAR provides a summary report for quality assessment. NEAR’s artifacts preprocessing consists of 2 steps: 1) Bad Channel Detection using a novel outlier detection algorithm Local Outlier Factor (LOF). 2) Bad Segments Detection using the most well-known Artifacts Subspace Reconstruction (ASR).
You can find all details about NEAR in our paper and how we compared it with existing works and how we validated it on simulated data, real data from newborns, infants and adults and on both Frequency-Tagging Paradigm and ERP design.
We made our source code freely available along with a dataset in the spirit of Open Science in GitHub (source code) and OSF
(sample newborn EEG data).
Oh! Don’t worry about defining thresholds for your datasets. We provide template scripts that you can easily modify to find the best thresholds for your dataset/application needs. Look for \TuneASR and \TuneLOF folders in the code repo.
