Hidden Markov models to improve the performance of an artificial neural network in sleep stage scoring

Authors:
A. J. Serrano;R. Magdalena;J. D. Martín;M. Bataller;M. Martínez;E. Soria
Affiliations:
Department of Electronics Engineering, University of Valencia, Valencia, Spain;Department of Electronics Engineering, University of Valencia, Valencia, Spain;Department of Electronics Engineering, University of Valencia, Valencia, Spain;Department of Electronics Engineering, University of Valencia, Valencia, Spain;Department of Electronics Engineering, University of Valencia, Valencia, Spain;Department of Electronics Engineering, University of Valencia, Valencia, Spain
Venue:
ACS'06 Proceedings of the 6th WSEAS international conference on Applied computer science
Year:
2006

Citing 4
Cited 2

A model-based monitor of human sleep stages

Biological Cybernetics
Neural network model: application to automatic analysis of human sleep

Computers and Biomedical Research
Neural networks for pattern recognition

Neural networks for pattern recognition
A review of evidence of health benefit from artificial neural networks in medical intervention

Neural Networks

Advanced methods for data mining

AIKED'09 Proceedings of the 8th WSEAS international conference on Artificial intelligence, knowledge engineering and data bases
A website structure optimization model

ACS'10 Proceedings of the 10th WSEAS international conference on Applied computer science

Quantified Score

Hi-index	0.00

Visualization

Abstract

It is common to use classifiers on polisomnographic records in order to determine the different stages during sleep. Most of the times the results yielded by this systems are not coherent with physiological aspects of the sleep. This work uses the Hidden Markov Models as a modeller of the physiological act of sleeping, and uses it as a corrector of the classification yielded by an artificial neural network. It has been tested on polisomnographic records from the MIT database. Results confirm an improvement of 0,17±0,05 in the Kappa coefficient of agreement and an improvement of 12,51±4,09% in success during test set.