Fundamentals of Spectral Clustering for Extracting Microstates of EEG

EasyChair Preprint 10449, version 2

Abstract

The present work proposes a novel approach for multichan-nel Electroencephalogram (EEG) microstates extraction based on the fundamentals of spectral clustering algorithm as a lower cost alternative technique from the classical model. This approach involves microstates generated from the La-placian matrix spectrum with special attention to the graph metric distances effects on the model performances. Results have demonstrated the potential of the technique to soften the clustering stages and encompass the variation present on the EEG. Two groups of subjects EEG have been used in this work, control and schizophrenic adolescents, and the experi-ments have presented the minimum of 79.64% explained variance for 6 microstates.

Keyphrases: EEG, Microstates, graphs, spectral clustering

@booklet{EasyChair:10449,
  author    = {Vladimir da Rocha Junior and Patrick Marques Ciarelli},
  title     = {Fundamentals of Spectral Clustering for Extracting Microstates of EEG},
  howpublished = {EasyChair Preprint 10449},
  year      = {EasyChair, 2023}}