Independent Component Analysis (ICA) With Covariates Strengthens Behavioral Links in Electroencephalography (EEG) Connectivity

The search for reliable biomarkers of clinical and cognitive deficits remains the holy grail of clinical neuroscience, with improved predictive methods and neural correlates paving the way for more effective treatments. Independent component analysis (ICA) has been widely applied in electroencephalography (EEG) signal processing to isolate neural activity from artifacts and noise. This study introduces a novel approach by integrating clinical covariates - behavioral assessments from the Woodcock-Johnson Cognitive Abilities Test III (WJ) Tests - into ICA, enabling the simultaneous decomposition of EEG connectivity patterns and cognitive performance metrics. Using functional connectivity measures as input, we applied two ICA methodologies to a dataset of 175 patients: (1) conventional ICA on EEG connectivity data, followed by correlation analysis with WJ scores, and (2) an augmented ICA approach incorporating both EEG connectivity and WJ measures. Our findings demonstrate that integrating behavioral data into ICA decomposition enhances the significance and robustness of correlations between EEG connectivity and cognitive performance in independent test datasets. These results underscore the potential of ICA with integrated covariates as a powerful multivariate framework for uncovering brain-behavior relationships, offering new insights for clinical and cognitive neuroscience research.