ICLabel

ICLabel classifies independent components from electroencephalogram (EEG) recordings to distinguish biological and non-biological source categories and enable automated interpretation of EEG source generators.

Key Features:

• ICLabel Dataset: Contains spatiotemporal measures for over 200,000 independent components (ICs) derived from more than 6,000 EEG recordings, including matching component labels for over 6,000 ICs and preprocessing using a common average reference.
• Automated Classifier: Implements automated classification of ICs into broad source categories with improved accuracy and up to a tenfold increase in processing speed compared to previously reported publicly available automated IC classification methods.
• Independent Component Analysis (ICA) integration: Leverages ICA to separate mixed EEG recordings into independent components corresponding to distinct source generator processes.
• Matlab implementation: Provides a MATLAB implementation of the automated IC classifier for computational use.

Scientific Applications:

• Large-scale EEG studies: Accelerates IC labeling and interpretation across numerous subjects for population-level and multi-subject analyses.
• Near-real-time and time-sensitive research: Supports applications requiring rapid IC classification, including near-real-time analysis and monitoring.
• Basic neuroscience research: Facilitates investigation of mesoscale brain dynamics by separating and labeling neural and non-neural sources.
• Clinical diagnostics and monitoring: Enables faster interpretation of EEG components for clinical diagnostics and real-time monitoring systems.

Methodology:

Preprocessing with a common average reference, decomposition using independent component analysis (ICA) to produce ICs, and automated classification of ICs into broad source categories based on spatiotemporal measures.