Journal Browser Advanced Search Help
Journal Browser Advanced search HELP
-
J Korean Soc Med Inform. 2000 Mar;6(1):79-86. Korean. Original Article. https://doi.org/10.4258/jksmi.2000.6.1.79
Park HJ , Park KS , Jeong DU .
Institute of Biomedical Engineering, Seoul National University, College of Medicine, Korea. kspark@snuvh.snu.ac.kr
Department of Biomedical Engineering, Seoul National University, College of Medicine, Korea.
Department of Psychiatry, Seoul National University, College of Medicine, Korea.
Abstract

In order to increase the performance of automatic sleep stage scoring, we propose a hybrid neural-network and rule-based expert system taking advantages of each system. The suggesting hybrid system comprises signal cleaning. feature extraction, event detection, rule-based sleep scoring and neural network classification. We selected segment based EEG features. the state of EOG. and EMG tone as a major feature set. With the extracted features, the rule-based expert system classities the sleep stages by symbolic reasoning. The scoring process of rule-based expert system comprises the single epoch reasoning based on the typical events and the multi-epoch adjusting when no events are detected. If the decision of rule-based expert system is uncertain, then these features are fed into the neural network. We used a two hidden layer feed forward network using error hack propagation algorithm. The agreement rate between human scorer and automatic algorithm were evaluated. The neural network supplements the shortcomings of rule-based system by dealing with exceptions of rules. The result shows that the compuational ol computational and symbolic intelligence is promising approach sleep signal anal) sis.

Copyright © 2019. Korean Association of Medical Journal Editors.