Auditory brainstem evoked potential classification for threshold detection by neural networks. I. Network design, similarities between human-expert and network classification, feasibility

Dogan Alpsan, Ozcan Ozdamar

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

An artificial neural network method to classify auditory brainstem evoked potentials (BAEPs) for automated response detection is developed. Three layer feedforward networks trained with backpropagation method were used to classify the BAEP signals into `response' or `no-response' classes. Temporal waveforms were used as inputs after various coding procedures. The effects of hidden layer size and training set size on the recognition rate of neural networks were studied. The simulations showed that the correct recognition rates of the networks were moderate, approaching the lower boundary of an optimum Bayesian classifier in several cases. However, when the contextual information available to the human experts during classification but not to the networks was removed, by letting the human experts classify the same signals under conditions similar to which the networks operated, the networks performed nearly as good as the experts. The results also suggest that the features selected and represented by the networks resemble those features of the signals used by experts in making their classifications.

Original languageEnglish
Pages (from-to)67-82
Number of pages16
JournalAutomedica
Volume15
Issue number1
StatePublished - Dec 1 1992
Externally publishedYes

Fingerprint

Brain Stem Auditory Evoked Potentials
Bioelectric potentials
Neural networks
Backpropagation
Classifiers
Evoked Potentials
Brain Stem

ASJC Scopus subject areas

  • Biophysics
  • Bioengineering

Cite this

@article{81b3d6949b4c4f93b588268be1aedcb3,
title = "Auditory brainstem evoked potential classification for threshold detection by neural networks. I. Network design, similarities between human-expert and network classification, feasibility",
abstract = "An artificial neural network method to classify auditory brainstem evoked potentials (BAEPs) for automated response detection is developed. Three layer feedforward networks trained with backpropagation method were used to classify the BAEP signals into `response' or `no-response' classes. Temporal waveforms were used as inputs after various coding procedures. The effects of hidden layer size and training set size on the recognition rate of neural networks were studied. The simulations showed that the correct recognition rates of the networks were moderate, approaching the lower boundary of an optimum Bayesian classifier in several cases. However, when the contextual information available to the human experts during classification but not to the networks was removed, by letting the human experts classify the same signals under conditions similar to which the networks operated, the networks performed nearly as good as the experts. The results also suggest that the features selected and represented by the networks resemble those features of the signals used by experts in making their classifications.",
author = "Dogan Alpsan and Ozcan Ozdamar",
year = "1992",
month = "12",
day = "1",
language = "English",
volume = "15",
pages = "67--82",
journal = "Automedica",
issn = "0095-0963",
publisher = "Gordon and Breach Science Publishers",
number = "1",

}

TY - JOUR

T1 - Auditory brainstem evoked potential classification for threshold detection by neural networks. I. Network design, similarities between human-expert and network classification, feasibility

AU - Alpsan, Dogan

AU - Ozdamar, Ozcan

PY - 1992/12/1

Y1 - 1992/12/1

N2 - An artificial neural network method to classify auditory brainstem evoked potentials (BAEPs) for automated response detection is developed. Three layer feedforward networks trained with backpropagation method were used to classify the BAEP signals into `response' or `no-response' classes. Temporal waveforms were used as inputs after various coding procedures. The effects of hidden layer size and training set size on the recognition rate of neural networks were studied. The simulations showed that the correct recognition rates of the networks were moderate, approaching the lower boundary of an optimum Bayesian classifier in several cases. However, when the contextual information available to the human experts during classification but not to the networks was removed, by letting the human experts classify the same signals under conditions similar to which the networks operated, the networks performed nearly as good as the experts. The results also suggest that the features selected and represented by the networks resemble those features of the signals used by experts in making their classifications.

AB - An artificial neural network method to classify auditory brainstem evoked potentials (BAEPs) for automated response detection is developed. Three layer feedforward networks trained with backpropagation method were used to classify the BAEP signals into `response' or `no-response' classes. Temporal waveforms were used as inputs after various coding procedures. The effects of hidden layer size and training set size on the recognition rate of neural networks were studied. The simulations showed that the correct recognition rates of the networks were moderate, approaching the lower boundary of an optimum Bayesian classifier in several cases. However, when the contextual information available to the human experts during classification but not to the networks was removed, by letting the human experts classify the same signals under conditions similar to which the networks operated, the networks performed nearly as good as the experts. The results also suggest that the features selected and represented by the networks resemble those features of the signals used by experts in making their classifications.

UR - http://www.scopus.com/inward/record.url?scp=0027002257&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0027002257&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:0027002257

VL - 15

SP - 67

EP - 82

JO - Automedica

JF - Automedica

SN - 0095-0963

IS - 1

ER -