Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia

Mark van Gils, H. Viertiö-Oja, Arvi Yli-Hankala, Ilkka Korhonen

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

Abstract

From a set of neurophysiologic variables an optimal combination was sought for determining the depth of anaesthesia. Performances of different comibations of variables were assessed on their ability to perdict the OAA/S score levels annotated by an anaesthesiologist. From the EEG of 32 patient nine features plus BIS-values were used. Factor analysis lead to 4 factors representing; spectral entropies, SynchFastSlow and betaratio; burst-suppression related variables; spectral edge frequency; and EMG information. The prediction probability of artificial neural network (ANN)-based classifiers trained with as inputs a) extracted factor variables and b) a combination of original variables was compared to the prediction probabilities of spectral entropy and BIS. The results indicated that when the performance was close to 100% this was so for all classifiers. For more difficult cases (performance <95 %) the ANNs perform better than the sole use of spectral entropy - BIS performs considerably worse.,
Original languageEnglish
Title of host publication2nd European Medical and Biomedical Engineering Conference, EMBEC02
Place of PublicationGraz, Austria
Pages390-391
Publication statusPublished - 2002
MoE publication typeA4 Article in a conference publication
Event2nd European Medical and Biomedical Engineering Conference - Vienna, Austria
Duration: 4 Dec 20028 Dec 2002

Publication series

SeriesIFMBE Proceedings
Volume3
ISSN1680-0737

Conference

Conference2nd European Medical and Biomedical Engineering Conference
CountryAustria
CityVienna
Period4/12/028/12/02

Fingerprint

entropy
prediction
factor analysis
artificial neural network
parameter

Keywords

  • depth of anaesthesia
  • EEG analysis
  • factor analysis
  • artificial neural networks
  • OAA/S

Cite this

van Gils, M., Viertiö-Oja, H., Yli-Hankala, A., & Korhonen, I. (2002). Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia. In 2nd European Medical and Biomedical Engineering Conference, EMBEC02 (pp. 390-391). Graz, Austria. IFMBE Proceedings, Vol.. 3
van Gils, Mark ; Viertiö-Oja, H. ; Yli-Hankala, Arvi ; Korhonen, Ilkka. / Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia. 2nd European Medical and Biomedical Engineering Conference, EMBEC02. Graz, Austria, 2002. pp. 390-391 (IFMBE Proceedings, Vol. 3).
@inproceedings{02c6a8e097174bf88eafc115fbabefb6,
title = "Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia",
abstract = "From a set of neurophysiologic variables an optimal combination was sought for determining the depth of anaesthesia. Performances of different comibations of variables were assessed on their ability to perdict the OAA/S score levels annotated by an anaesthesiologist. From the EEG of 32 patient nine features plus BIS-values were used. Factor analysis lead to 4 factors representing; spectral entropies, SynchFastSlow and betaratio; burst-suppression related variables; spectral edge frequency; and EMG information. The prediction probability of artificial neural network (ANN)-based classifiers trained with as inputs a) extracted factor variables and b) a combination of original variables was compared to the prediction probabilities of spectral entropy and BIS. The results indicated that when the performance was close to 100{\%} this was so for all classifiers. For more difficult cases (performance <95 {\%}) the ANNs perform better than the sole use of spectral entropy - BIS performs considerably worse.,",
keywords = "depth of anaesthesia, EEG analysis, factor analysis, artificial neural networks, OAA/S",
author = "{van Gils}, Mark and H. Vierti{\"o}-Oja and Arvi Yli-Hankala and Ilkka Korhonen",
year = "2002",
language = "English",
isbn = "3-901351-62-0",
series = "IFMBE Proceedings",
publisher = "Springer",
pages = "390--391",
booktitle = "2nd European Medical and Biomedical Engineering Conference, EMBEC02",

}

van Gils, M, Viertiö-Oja, H, Yli-Hankala, A & Korhonen, I 2002, Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia. in 2nd European Medical and Biomedical Engineering Conference, EMBEC02. Graz, Austria, IFMBE Proceedings, vol. 3, pp. 390-391, 2nd European Medical and Biomedical Engineering Conference, Vienna, Austria, 4/12/02.

Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia. / van Gils, Mark; Viertiö-Oja, H.; Yli-Hankala, Arvi; Korhonen, Ilkka.

2nd European Medical and Biomedical Engineering Conference, EMBEC02. Graz, Austria, 2002. p. 390-391 (IFMBE Proceedings, Vol. 3).

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

TY - GEN

T1 - Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia

AU - van Gils, Mark

AU - Viertiö-Oja, H.

AU - Yli-Hankala, Arvi

AU - Korhonen, Ilkka

PY - 2002

Y1 - 2002

N2 - From a set of neurophysiologic variables an optimal combination was sought for determining the depth of anaesthesia. Performances of different comibations of variables were assessed on their ability to perdict the OAA/S score levels annotated by an anaesthesiologist. From the EEG of 32 patient nine features plus BIS-values were used. Factor analysis lead to 4 factors representing; spectral entropies, SynchFastSlow and betaratio; burst-suppression related variables; spectral edge frequency; and EMG information. The prediction probability of artificial neural network (ANN)-based classifiers trained with as inputs a) extracted factor variables and b) a combination of original variables was compared to the prediction probabilities of spectral entropy and BIS. The results indicated that when the performance was close to 100% this was so for all classifiers. For more difficult cases (performance <95 %) the ANNs perform better than the sole use of spectral entropy - BIS performs considerably worse.,

AB - From a set of neurophysiologic variables an optimal combination was sought for determining the depth of anaesthesia. Performances of different comibations of variables were assessed on their ability to perdict the OAA/S score levels annotated by an anaesthesiologist. From the EEG of 32 patient nine features plus BIS-values were used. Factor analysis lead to 4 factors representing; spectral entropies, SynchFastSlow and betaratio; burst-suppression related variables; spectral edge frequency; and EMG information. The prediction probability of artificial neural network (ANN)-based classifiers trained with as inputs a) extracted factor variables and b) a combination of original variables was compared to the prediction probabilities of spectral entropy and BIS. The results indicated that when the performance was close to 100% this was so for all classifiers. For more difficult cases (performance <95 %) the ANNs perform better than the sole use of spectral entropy - BIS performs considerably worse.,

KW - depth of anaesthesia

KW - EEG analysis

KW - factor analysis

KW - artificial neural networks

KW - OAA/S

M3 - Conference article in proceedings

SN - 3-901351-62-0

T3 - IFMBE Proceedings

SP - 390

EP - 391

BT - 2nd European Medical and Biomedical Engineering Conference, EMBEC02

CY - Graz, Austria

ER -

van Gils M, Viertiö-Oja H, Yli-Hankala A, Korhonen I. Identification of a set of optimal EEG parameters for estimation of depth of anaesthesia. In 2nd European Medical and Biomedical Engineering Conference, EMBEC02. Graz, Austria. 2002. p. 390-391. (IFMBE Proceedings, Vol. 3).