Eeg analysis in cerebral death condition using wavelet packet decomposition

This paper proposes a methodology for extracting features from EEG in suspected brain death conditions. Those signals are usually caracterized by a low SNR and in such situations distortions introduced by skull/electrode/cable interface become very relevant. After a calibration procedure, the analysis of EEG signals was performed with a Wavelet Packet Decomposition filter, where the main component of noise are removed with soft thresholding strategy. We propose an estimate of error probability leading to the choice of wavelet basis and threshold parameters. As result, in absence of brain activity, the output of this filter produces really flat signals. However the presence of some activity can be examinated in WP time-frequency domain revealing distinctions between real brain activity, narrow band noise and artifacts.

@InProceedings{SM03c,
  author       = {Santoni, F. and Marrosu, F.},
  title        = {Eeg analysis in cerebral death condition using wavelet packet decomposition},
  booktitle    = {Mathematical Modelling \& Computing In Biology And Medicine},
  volume       = {1},
  pages        = {110--117},
  year         = {2003},
  editor       = {Vincenzo Capasso},
  publisher    = {Societa editrice ESCULAPIO},
  organization = {European Society for mathematical and Theoretical Biology},
  address      = {Bologna},
  note         = {isbn: 88-7488-055-3idxproject: ?},
  keywords     = {EEG analysis, Wavelet Packet},
  url          = {https://publications.crs4.it/pubdocs/2003/SM03c},
}