ICLabel: An automated electroencephalographic independent component classifier, dataset, and website

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Electroencephalogram (EEG) offers high temporal resolution for brain dynamics but generates complex, correlated signals.
• Independent Component Analysis (ICA) isolates signal sources, but manual interpretation of independent components (ICs) is time-consuming.
• Automated classifiers can expedite EEG analysis and enable real-time applications.

Purpose of the Study:

• To present the ICLabel project, including a large dataset, educational website, and an automated classifier for EEG independent components (ICs).
• To improve the accuracy and computational efficiency of automated IC classification for EEG data.

Main Methods:

• Developed the ICLabel dataset with over 200,000 ICs from 6000+ EEG recordings and crowdsourced labels.
• Created the ICLabel website for data collection and researcher education.
• Implemented the ICLabel automated classifier in MATLAB, optimizing for accuracy and speed.

Main Results:

• The ICLabel classifier demonstrates improved accuracy in estimating IC labels compared to existing methods.
• ICLabel achieves significantly enhanced computational efficiency, running ten times faster than previous top classifiers.
• Performance is comparable or superior across all measured IC categories.

Conclusions:

• ICLabel provides a highly accurate and efficient automated solution for classifying EEG independent components.
• The ICLabel project facilitates faster, more accessible EEG data analysis for researchers and practitioners.
• This tool supports broader adoption of ICA in EEG studies and real-time applications.