Reliability and state-dependency of EEG connectivity, complexity and network characteristics

Le résumé fourni par la source

Abstract Resting-state electroencephalography (EEG) metrics are influenced by task instructions and momentary factors such as cognitive state, complicating their use as biomarkers. We assessed how strongly three classes of resting-state EEG metrics depend on time and state: functional connectivity (FC; amplitude-envelope correlation, AECc, and phase-lag index, PLI), signal complexity (inverted joint permutation entropy, JPE INV , and permutation entropy, PE) and network topology derived from minimum spanning trees (MST). Sixty-four-channel EEG was recorded in healthy adults during two sessions six weeks apart (n = 42) and during semi-resting-state epochs embedded in a P50-gating task (n = 24). Reliability for repeated-resting recordings and resting-state versus semi-resting-state comparisons was quantified with intraclass correlation coefficients (ICC) at sensor and source level. PE showed consistently good-to-excellent reliability (ICC > 0.75–0.90). FC metrics ranged from poor to excellent, and MST metrics from poor to good. Across analyses, theta and alpha bands outperformed delta and beta bands. Alpha and theta PE and alpha PLI were the most robust, whereas MST and AECc require caution, especially outside theta and alpha bands. Our results identify theta and alpha PE and alpha PLI as robust measures suitable for biomarker development, while urging caution for MST and AECc due to their limited stability across time and state.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.