Advanced

Detection of 'EEG bursts' in the early preterm EEG: Visual vs. automated detection

Palmu, Kirsi; Wikstrom, Sverre; Hippelainen, Eero; Boylan, Geraldine; Hellström-Westas, Lena LU and Vanhatalo, Sampsa (2010) In Clinical Neurophysiology 121(7). p.1015-1022
Abstract
Objective: To describe the characteristics of activity bursts in the early preterm EEG, to assess inter-rater agreement of burst detection by visual inspection, and to determine the performance of an automated burst detector that uses non-linear energy operator (NLEO). Methods: EEG recordings from extremely preterm (n = 12) and very preterm (n = 6) infants were analysed. Three neurophysiologists independently marked bursts in the EEG, the characteristics of bursts were analyzed and inter-rater agreement determined. Unanimous detections were used as the gold standard in estimating the performance of an automated burst detector. In addition, some details of this automated detector were revised in an attempt to improve performance. Results:... (More)
Objective: To describe the characteristics of activity bursts in the early preterm EEG, to assess inter-rater agreement of burst detection by visual inspection, and to determine the performance of an automated burst detector that uses non-linear energy operator (NLEO). Methods: EEG recordings from extremely preterm (n = 12) and very preterm (n = 6) infants were analysed. Three neurophysiologists independently marked bursts in the EEG, the characteristics of bursts were analyzed and inter-rater agreement determined. Unanimous detections were used as the gold standard in estimating the performance of an automated burst detector. In addition, some details of this automated detector were revised in an attempt to improve performance. Results: Overall, inter-rater agreement was 86% for extremely preterm infants and 81% for very preterm infants. In visual markings, bursts had variable lengths (similar to 1-10 s) and increased amplitudes (and power) throughout the frequency spectrum. Accuracy of the original detection algorithm was 87% and 79% and accuracy of the revised algorithm 93% and 87% for extremely preterm and very preterm babies, respectively. Conclusion: Visual detection of bursts from the early preterm EEG is comparable albeit not identical between raters. The original automated detector underestimates the amount of burst occurrence, but can be readily improved to yield results comparable to visual detection. Further clinical studies are warranted to assess the optimal descriptors of burst detection for monitoring and prognostication. Significance: Validation of a burst detector offers an evidence-based platform for further development of brain monitors in very preterm babies. (C) 2010 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
Brain monitoring, Neonatal intensive care unit, SAT, Premature EEG, Neonatal EEG
in
Clinical Neurophysiology
volume
121
issue
7
pages
1015 - 1022
publisher
Elsevier
external identifiers
  • wos:000278222300007
  • scopus:77952741790
ISSN
1872-8952
DOI
10.1016/j.clinph.2010.02.010
language
English
LU publication?
yes
id
b6879c16-a1c7-4db0-847c-8e2193f8dbbc (old id 1632244)
date added to LUP
2010-07-21 09:17:11
date last changed
2018-05-29 10:07:41
@article{b6879c16-a1c7-4db0-847c-8e2193f8dbbc,
  abstract     = {Objective: To describe the characteristics of activity bursts in the early preterm EEG, to assess inter-rater agreement of burst detection by visual inspection, and to determine the performance of an automated burst detector that uses non-linear energy operator (NLEO). Methods: EEG recordings from extremely preterm (n = 12) and very preterm (n = 6) infants were analysed. Three neurophysiologists independently marked bursts in the EEG, the characteristics of bursts were analyzed and inter-rater agreement determined. Unanimous detections were used as the gold standard in estimating the performance of an automated burst detector. In addition, some details of this automated detector were revised in an attempt to improve performance. Results: Overall, inter-rater agreement was 86% for extremely preterm infants and 81% for very preterm infants. In visual markings, bursts had variable lengths (similar to 1-10 s) and increased amplitudes (and power) throughout the frequency spectrum. Accuracy of the original detection algorithm was 87% and 79% and accuracy of the revised algorithm 93% and 87% for extremely preterm and very preterm babies, respectively. Conclusion: Visual detection of bursts from the early preterm EEG is comparable albeit not identical between raters. The original automated detector underestimates the amount of burst occurrence, but can be readily improved to yield results comparable to visual detection. Further clinical studies are warranted to assess the optimal descriptors of burst detection for monitoring and prognostication. Significance: Validation of a burst detector offers an evidence-based platform for further development of brain monitors in very preterm babies. (C) 2010 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.},
  author       = {Palmu, Kirsi and Wikstrom, Sverre and Hippelainen, Eero and Boylan, Geraldine and Hellström-Westas, Lena and Vanhatalo, Sampsa},
  issn         = {1872-8952},
  keyword      = {Brain monitoring,Neonatal intensive care unit,SAT,Premature EEG,Neonatal EEG},
  language     = {eng},
  number       = {7},
  pages        = {1015--1022},
  publisher    = {Elsevier},
  series       = {Clinical Neurophysiology},
  title        = {Detection of 'EEG bursts' in the early preterm EEG: Visual vs. automated detection},
  url          = {http://dx.doi.org/10.1016/j.clinph.2010.02.010},
  volume       = {121},
  year         = {2010},
}