Comparison between Conventional and Neural Network Classifiers for Rat Sleep-Wake Stage Discrimination

Abstract

This article describes an approach to selecting the most efficient classifier for rat sleep staging (waking, REM sleep and NREM sleep) discrimination. Three conventional (bayesian, linear, and euclidean) and two neural network (multilayer perceptrons integrating or not integrating contextual information) classifiers were compared. For each classifier, performances were presented in the form of a statistical concordance matrix comparing classifier results versus human expert results on 6 24-hour records (1 record per animal). Comparison between classifiers was based on the estimation and accuracy of global agreements. Interest was also focused on REM sleep state discrimination. The results show that neural network classifiers are appropriate tools to be integrated in an automatic rat sleep-wake stage system. The approach presented should help scientists in choosing a method of data classification.

Automated analysis of sleep–wake state in rats

“…one cannot expect a computer to agree with two human scorers who disagree." In other words, the best one can reasonably expect is for the computer-human agreement to approach inter-rater agreement (Gottesmann et al, 1977), which is generally reported to be within the range 85-95% in rat sleep scoring (Crisler et al, 2008;Louis et al, 2004;Neckelmann et al, 1994;Robert et al, 1997;Ruigt et al, 1989b). This problem was highlighted in the present study when two raters were instructed to rate their scores as "high confidence" or "equivocal".…”

Sleep Impairments in Rats Implanted with Morphine Pellets

“…A postprocessing procedure based on logical rules was also used to enhance PS discrimination. Further details concerning the EEG recording technique and the neural network classifier developed can be found in [14][15][16].…”

Comparison Between Five Classifiers for Automatic Scoring of Human Sleep Recordings

“…These results can be explained by the fact that the neural network is not sensitive to the distributions of the data; transformations have no effect on its ability to separate space into subspaces [26]. But, the effect of the transformation leads to an improvement of the speed of convergence for the optimization of the backpropagation during the learning process explained by a better homogeneity in the distribution of the weights and features in the input layer.…”