A comparative study of different references for EEG default mode network: The use of the infinity reference

Choosing the reference electrode for EEG recordings substantially affects measures of brain activity, including the default mode network (DMN). Simulated data showed that the infinity reference obtained by the reference electrode standardisation technique (REST) exactly recovered the true network configuration. In real EEG data from 15 participants, different references—REST, average reference, linked mastoids, and left mastoid—produced significant differences in power spectra, coherence, and DMN connectivity. Compared with REST, the average reference strengthened long-distance connections between anterior and posterior brain areas, while linked mastoids and left mastoid references disrupted posterior connectivity. The effects depended on EEG frequency band. REST is a valid reference technique for objective comparisons across studies and clinical practice.

Study at a glance

Abstract

ObjectiveThe choice of electroencephalograph (EEG) reference is a critical issue for the study of brain activity. The present study addressed the use of the infinity reference obtained by the reference electrode standardisation technique (REST) in the study of EEG default mode network (DMN).MethodsA total of 100 randomly positioned source configurations, each consisting of two dipoles with coherent waveforms, were adopted for simulating EEG networks. Dense (129-channel), eyes-closed EEG was recorded from 15 subjects. Simulated data with infinity as reference and the real data were re-referenced to reconstructed infinity (REST), their average (AR), linked mastoids (LM) and left mastoid (L) references. For simulated data, the effects of different references on coherence and network were investigated. For real data, spectral properties of seven conventional EEG frequency bands were first analysed and then DMN was constructed based on the coherence.ResultsThe simulation showed that REST can exactly recover the true EEG network configuration. For real EEG data, significant differences among references were found for the power spectra, coherence and DMN configuration. Compared with REST, the long-distance connectivity between anterior and posterior areas was strengthened by AR, and the connectivity over posterior areas was destroyed when LM and L were employed. Moreover, all comparisons demonstrated frequency-dependent reference effects.ConclusionsNon-neutral reference influences the power spectra, coherence as well as the network analysis. REST demonstrates its validity in data referencing, and meanwhile, AR is much closer to REST than the other references in terms of spectra and coherence. However, the DMN alters a great deal with AR.SignificanceThe results underscore the importance of considering EEG reference effects in the functional connectivity studies. REST is a promising reference technique for objective comparison in cross-laboratory studies and clinical practices.