RELICA: a method for estimating the reliability of independent components

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Independent Component Analysis (ICA) is crucial for dissecting electroencephalography (EEG) signals into independent sources.
• Signal instability in ICA can arise from data noise, complicating the interpretation of brain and artifact components.
• Reliable identification of physiologically plausible components is essential for accurate EEG analysis.

Purpose of the Study:

• To introduce RELICA (RELiable ICA), a novel method for quantifying within-subject reliability of Independent Components (ICs) derived from EEG data.
• To provide a robust measure of IC stability and physiological plausibility, independent of cross-subject comparisons.
• To improve the confidence in interpreting identified brain and artifactual sources in EEG.

Main Methods:

• RELICA calculates IC "dipolarity" for physiological plausibility and assesses IC consistency across bootstrap data decompositions.
• These measures are used to visualize and cluster ICs, generating a within-subject reliability score.
• The method was validated on EEG data from 14 subjects during a working memory task.

Main Results:

• RELICA successfully classified many brain and ocular artifact ICs as "stable" (highly repeatable).
• Identified stable ICs corresponded to both genuine brain activity and non-brain artifacts like line noise.
• The findings demonstrate RELICA's effectiveness in distinguishing reliable components from unstable ones.

Conclusions:

• RELICA offers a reliable within-subject method to assess the stability and physiological relevance of ICs from EEG.
• This approach can enhance the interpretability of ICA results and reduce reliance on unstable or uninterpretable components.
• RELICA is adaptable to various linear blind source separation algorithms for improved EEG data analysis.