Semi-automatic identification of independent components representing EEG artifact

• Published in: Clinical neurophysiology Vol. 120; no. 5; pp. 868 - 877
• Main Authors: Campos Viola, Filipa, Thorne, Jeremy, Edmonds, Barrie, Schneider, Till, Eichele, Tom, Debener, Stefan
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ireland Ltd 01.05.2009; Elsevier
• Subjects: Algorithms; Artifact correction; Artificial Intelligence; Biological and medical sciences; Brain - physiology; Brain Mapping - methods; Computer Simulation; EEG; Electrodiagnosis. Electric activity recording; Electroencephalography - methods; Evoked Potentials - physiology; Eye and associated structures. Visual pathways and centers. Vision; Eye blinks; Eye Movements - physiology; Fundamental and applied biological sciences. Psychology; Heart Rate - physiology; Humans; ICA; Independent component analysis; Investigative techniques, diagnostic techniques (general aspects); Lateral eye movements; Medical sciences; Nervous system; Neurology; Pattern Recognition, Automated - methods; Signal Processing, Computer-Assisted; Software; Software Validation; Vertebrates: nervous system and sense organs; Artifact correction; Lateral eye movements; ICA; EEG; Eye blinks; Independent component analysis; Human; User; Decision making; Objective analysis; Central nervous system; Electrophysiology; Artefact; Electroencephalography; Interpretation; Eye movement; Encephalon; Eye; Visual system; Electrodiagnosis; Performance
• ISSN: 1388-2457; 1872-8952
• DOI: 10.1016/j.clinph.2009.01.015

Independent component analysis (ICA) can disentangle multi-channel electroencephalogram (EEG) signals into a number of artifacts and brain-related signals. However, the identification and interpretation of independent components is time-consuming and involves subjective decision making. We developed and evaluated a semi-automatic tool designed for clustering independent components from different subjects and/or EEG recordings. CORRMAP is an open-source EEGLAB plug-in, based on the correlation of ICA inverse weights, and finds independent components that are similar to a user-defined template. Component similarity is measured using a correlation procedure that selects components that pass a threshold. The threshold can be either user-defined or determined automatically. CORRMAP clustering performance was evaluated by comparing it with the performance of 11 users from different laboratories familiar with ICA. For eye-related artifacts, a very high degree of overlap between users (phi > 0.80), and between users and CORRMAP (phi > 0.80) was observed. Lower degrees of association were found for heartbeat artifact components, between users (phi < 0.70), and between users and CORRMAP (phi < 0.65). These results demonstrate that CORRMAP provides an efficient, convenient and objective way of clustering independent components. CORRMAP helps to efficiently use ICA for the removal EEG artifacts.