Estimation of the independent components reliability of EEG signal in a clinical application

Independent Component Analysis (ICA) is a statistical method that has effective roles in processing the electroencephalography (EEG) signal, especially for finding artifacts. This research aimed to automatically estimate the validity of independent components in the EEG signal and extracting more brain-related sources. Data were obtained from adult attention deficit hyperactivity disorder (ADHD) participants and control groups during the continuous performance task (CPT). The reliability of ICs was estimated by different methods, including the statistical validity measurement for each component, their physiological plausibility and the equivalent dipole interpretation both in time and frequency domain. Group-ICA algorithm was applied to ADHD/control groups in order to compare the validity of components between subjects. Homogenous components in each group were determined by clustering. The results showed that in all algorithms, including the automatic algorithm and Group-ICA, the number of reliable components in the frequency domain is greater than (about twice) the number of estimated components in the time domain. Statistical results (Wilcoxon signed-rank test) for comparison the number of reliable components of degree 3 in both methods showed a significant difference (P-value: 0.0113) and the same result was observed for reliable components of degree 2 (P-value: 4.8e-04).

Automated summary

This study utilized Independent Component Analysis to automatically assess the reliability of components in EEG signals and found that the number of reliable components in the frequency domain is greater than that estimated in the time domain. Group-ICA algorithm was applied to ADHD/control groups to compare the validity of components between subjects, showing significant differences in the number of reliable components. Homogenous components in each group were determined by clustering.