EEG functional connectivity methods for seizure-related disorders

EEG functional connectivity analysis is increasingly used to study neurological disorders. However, the lack of standardization in connectivity estimation limits its clinical translation. This work addresses the need for more guidance in connectivity pipeline design by evaluating numerous techniques in two seizure-related disorders to identify optimal approaches. Retrospective EEG data were analyzed from two cohorts: acutely ill adults stratified by seizure occurrence (140 patients; 70 with seizures), and children with self-limited focal epilepsy with centro-temporal spikes (SeLECTS) stratified by cognitive regression (32 patients; 16 with regression). In each cohort, 960 connectivity pipelines were evaluated for their ability to discriminate between sub-groups using statistical testing and logistic regression. Pipelines were constructed by systematically varying re-referencing schemes, frequency bands, association measures, and network features. Significant connectivity differences were observed between sub-groups in both cohorts. Pipelines using corrected association measures (e.g., corrected correlation, weighted phase lag index) and network features capturing integration and segregation (e.g., characteristic path length, clustering coefficient) consistently showed superior performance. Pipeline rankings differed substantially across disorders, underscoring the importance of clinical context on the choice of parameters for connectivity analysis. These findings demonstrate the potential of EEG functional connectivity as a biomarker in seizure-related conditions, while emphasizing the critical impact of methodological choices. This study provides quantitative guidelines for informed pipeline design and supports the development of automated EEG connectivity analyses in clinical practice.