Power spectral density (PSD) and network analysis performed on functional correlation (FC) patterns represent two common approaches used to characterize Electroencephalographic (EEG) data. Despite the two approaches are widely used, their possible association may need more attention. To investigate this question, we performed a comparison between PSD and some widely used nodal network metrics (namely strength, clustering coefficient and betweenness centrality), using two different publicly available resting-state EEG datasets, both at scalp and source levels, employing four different FC methods (PLV, PLI, AEC and AECC). Here we show that the two approaches may provide similar information and that their correlation depends on the method used to estimate FC. In particular, our results show a strong correlation between PSD and nodal network metrics derived from FC methods (pick at 0.736 for PLV and 0.530 for AEC) that do not limit the effects of volume conduction/signal leakage. The correlations are less relevant for more conservative FC methods (pick at 0.224 for AECC). These findings suggest that the results derived from the two different approaches may be not independent and should not be treated as distinct analyses. We conclude that it may represent good practice to report the findings from the two approaches in conjunction to have a more comprehensive view of the results.

A comparison between power spectral density and network metrics: an EEG study

Demuru, Matteo;La Cava, Simone Maurizio;Pani, Sara Maria;Fraschini, Matteo
2020-01-01

Abstract

Power spectral density (PSD) and network analysis performed on functional correlation (FC) patterns represent two common approaches used to characterize Electroencephalographic (EEG) data. Despite the two approaches are widely used, their possible association may need more attention. To investigate this question, we performed a comparison between PSD and some widely used nodal network metrics (namely strength, clustering coefficient and betweenness centrality), using two different publicly available resting-state EEG datasets, both at scalp and source levels, employing four different FC methods (PLV, PLI, AEC and AECC). Here we show that the two approaches may provide similar information and that their correlation depends on the method used to estimate FC. In particular, our results show a strong correlation between PSD and nodal network metrics derived from FC methods (pick at 0.736 for PLV and 0.530 for AEC) that do not limit the effects of volume conduction/signal leakage. The correlations are less relevant for more conservative FC methods (pick at 0.224 for AECC). These findings suggest that the results derived from the two different approaches may be not independent and should not be treated as distinct analyses. We conclude that it may represent good practice to report the findings from the two approaches in conjunction to have a more comprehensive view of the results.
2020
EEG; PSD; Network; Correlation
File in questo prodotto:
File Dimensione Formato  
1-s2.0-S1746809419303416-main.pdf

Solo gestori archivio

Tipologia: versione editoriale
Dimensione 2.24 MB
Formato Adobe PDF
2.24 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11584/280375
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 31
  • ???jsp.display-item.citation.isi??? 27
social impact